1
0
mirror of https://github.com/gorhill/uBlock.git synced 2024-11-17 07:52:42 +01:00
uBlock/src/js/biditrie.js

933 lines
34 KiB
JavaScript
Raw Normal View History

/*******************************************************************************
uBlock Origin - a browser extension to block requests.
Copyright (C) 2019-present Raymond Hill
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see {http://www.gnu.org/licenses/}.
Home: https://github.com/gorhill/uBlock
*/
/* globals WebAssembly, vAPI */
'use strict';
/*******************************************************************************
A BidiTrieContainer is mostly a large buffer in which distinct but related
tries are stored. The memory layout of the buffer is as follow:
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
0-2047: haystack section
2048-2051: number of significant characters in the haystack
2052-2055: offset to start of trie data section (=> trie0)
2056-2059: offset to end of trie data section (=> trie1)
2060-2063: offset to start of character data section (=> char0)
2064-2067: offset to end of character data section (=> char1)
2068: start of trie data section
+--------------+
Normal cell: | And | If "Segment info" matches:
(aka CELL) +--------------+ Goto "And"
| Or | Else
+--------------+ Goto "Or"
| Segment info |
+--------------+
+--------------+
Boundary cell: | Right And | "Right And" and/or "Left And"
(aka BCELL) +--------------+ can be 0 in last-segment condition.
| Left And |
+--------------+
| 0 |
+--------------+
Given following filters and assuming token is "ad" for all of them:
-images/ad-
/google_ad.
/images_ad.
_images/ad.
We get the following internal representation:
+-----------+ +-----------+ +---+
| |---->| |---->| 0 |
+-----------+ +-----------+ +---+ +-----------+
| 0 | +--| | | |---->| 0 |
+-----------+ | +-----------+ +---+ +-----------+
| ad | | | - | | 0 | | 0 |
+-----------+ | +-----------+ +---+ +-----------+
| | -images/ |
| +-----------+ +---+ +-----------+
+->| |---->| 0 |
+-----------+ +---+ +-----------+ +-----------+
| 0 | | |---->| |---->| 0 |
+-----------+ +---+ +-----------+ +-----------+
| . | | 0 | +--| | +--| |
+-----------+ +---+ | +-----------+ | +-----------+
| | _ | | | /google |
| +-----------+ | +-----------+
| |
| | +-----------+
| +->| 0 |
| +-----------+
| | 0 |
| +-----------+
| | /images |
| +-----------+
|
| +-----------+
+->| 0 |
+-----------+
| 0 |
+-----------+
| _images/ |
+-----------+
*/
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const PAGE_SIZE = 65536*2;
const HAYSTACK_START = 0;
const HAYSTACK_SIZE = 2048; // i32 / i8
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const HAYSTACK_SIZE_SLOT = HAYSTACK_SIZE >>> 2; // 512 / 2048
const TRIE0_SLOT = HAYSTACK_SIZE_SLOT + 1; // 513 / 2052
const TRIE1_SLOT = HAYSTACK_SIZE_SLOT + 2; // 514 / 2056
const CHAR0_SLOT = HAYSTACK_SIZE_SLOT + 3; // 515 / 2060
const CHAR1_SLOT = HAYSTACK_SIZE_SLOT + 4; // 516 / 2064
const RESULT_L_SLOT = HAYSTACK_SIZE_SLOT + 5; // 517 / 2068
const RESULT_R_SLOT = HAYSTACK_SIZE_SLOT + 6; // 518 / 2072
const RESULT_IU_SLOT = HAYSTACK_SIZE_SLOT + 7; // 519 / 2076
const TRIE0_START = HAYSTACK_SIZE_SLOT + 8 << 2; // 2080
const CELL_BYTE_LENGTH = 12;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const MIN_FREE_CELL_BYTE_LENGTH = CELL_BYTE_LENGTH * 8;
const CELL_AND = 0;
const CELL_OR = 1;
const SEGMENT_INFO = 2;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const BCELL_NEXT_AND = 0;
const BCELL_ALT_AND = 1;
const BCELL_EXTRA = 2;
const BCELL_EXTRA_MAX = 0x00FFFFFF;
const toSegmentInfo = (aL, l, r) => ((r - l) << 24) | (aL + l);
const roundToPageSize = v => (v + PAGE_SIZE-1) & ~(PAGE_SIZE-1);
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
class BidiTrieContainer {
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
constructor(extraHandler) {
const len = PAGE_SIZE * 4;
this.buf8 = new Uint8Array(len);
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
this.buf32 = new Uint32Array(this.buf8.buffer);
this.buf32[TRIE0_SLOT] = TRIE0_START;
this.buf32[TRIE1_SLOT] = this.buf32[TRIE0_SLOT];
this.buf32[CHAR0_SLOT] = len >>> 1;
this.buf32[CHAR1_SLOT] = this.buf32[CHAR0_SLOT];
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
this.haystack = this.buf8.subarray(
HAYSTACK_START,
HAYSTACK_START + HAYSTACK_SIZE
);
this.extraHandler = extraHandler;
this.textDecoder = null;
this.wasmMemory = null;
this.lastStored = '';
this.lastStoredLen = this.lastStoredIndex = 0;
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
//--------------------------------------------------------------------------
// Public methods
//--------------------------------------------------------------------------
get haystackLen() {
return this.buf32[HAYSTACK_SIZE_SLOT];
}
set haystackLen(v) {
this.buf32[HAYSTACK_SIZE_SLOT] = v;
}
reset(details) {
if (
details instanceof Object &&
typeof details.byteLength === 'number' &&
typeof details.char0 === 'number'
) {
if ( details.byteLength > this.buf8.byteLength ) {
this.reallocateBuf(details.byteLength);
}
this.buf32[CHAR0_SLOT] = details.char0;
}
this.buf32[TRIE1_SLOT] = this.buf32[TRIE0_SLOT];
this.buf32[CHAR1_SLOT] = this.buf32[CHAR0_SLOT];
this.lastStored = '';
this.lastStoredLen = this.lastStoredIndex = 0;
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
createTrie() {
// grow buffer if needed
if ( (this.buf32[CHAR0_SLOT] - this.buf32[TRIE1_SLOT]) < CELL_BYTE_LENGTH ) {
this.growBuf(CELL_BYTE_LENGTH, 0);
}
const iroot = this.buf32[TRIE1_SLOT] >>> 2;
this.buf32[TRIE1_SLOT] += CELL_BYTE_LENGTH;
this.buf32[iroot+CELL_OR] = 0;
this.buf32[iroot+CELL_AND] = 0;
this.buf32[iroot+SEGMENT_INFO] = 0;
return iroot;
}
matches(icell, ai) {
const buf32 = this.buf32;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const buf8 = this.buf8;
const char0 = buf32[CHAR0_SLOT];
const aR = buf32[HAYSTACK_SIZE_SLOT];
let al = ai, x = 0, y = 0;
for (;;) {
x = buf8[al];
al += 1;
// find matching segment
for (;;) {
y = buf32[icell+SEGMENT_INFO];
let bl = char0 + (y & 0x00FFFFFF);
if ( buf8[bl] === x ) {
y = (y >>> 24) - 1;
if ( y !== 0 ) {
x = al + y;
if ( x > aR ) { return 0; }
for (;;) {
bl += 1;
if ( buf8[bl] !== buf8[al] ) { return 0; }
al += 1;
if ( al === x ) { break; }
}
}
break;
}
icell = buf32[icell+CELL_OR];
if ( icell === 0 ) { return 0; }
}
// next segment
icell = buf32[icell+CELL_AND];
x = buf32[icell+BCELL_EXTRA];
if ( x <= BCELL_EXTRA_MAX ) {
if ( x !== 0 && this.matchesExtra(ai, al, x) !== 0 ) {
return 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
x = buf32[icell+BCELL_ALT_AND];
if ( x !== 0 && this.matchesLeft(x, ai, al) !== 0 ) {
return 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
icell = buf32[icell+BCELL_NEXT_AND];
if ( icell === 0 ) { return 0; }
}
if ( al === aR ) { return 0; }
}
return 0; // eslint-disable-line no-unreachable
}
matchesLeft(icell, ar, r) {
const buf32 = this.buf32;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const buf8 = this.buf8;
const char0 = buf32[CHAR0_SLOT];
let x = 0, y = 0;
for (;;) {
if ( ar === 0 ) { return 0; }
ar -= 1;
x = buf8[ar];
// find first segment with a first-character match
for (;;) {
y = buf32[icell+SEGMENT_INFO];
let br = char0 + (y & 0x00FFFFFF);
y = (y >>> 24) - 1;
br += y;
if ( buf8[br] === x ) { // all characters in segment must match
if ( y !== 0 ) {
x = ar - y;
if ( x < 0 ) { return 0; }
for (;;) {
ar -= 1; br -= 1;
if ( buf8[ar] !== buf8[br] ) { return 0; }
if ( ar === x ) { break; }
}
}
break;
}
icell = buf32[icell+CELL_OR];
if ( icell === 0 ) { return 0; }
}
// next segment
icell = buf32[icell+CELL_AND];
x = buf32[icell+BCELL_EXTRA];
if ( x <= BCELL_EXTRA_MAX ) {
if ( x !== 0 && this.matchesExtra(ar, r, x) !== 0 ) {
return 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
icell = buf32[icell+BCELL_NEXT_AND];
if ( icell === 0 ) { return 0; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
}
return 0; // eslint-disable-line no-unreachable
}
matchesExtra(l, r, ix) {
let iu = 0;
if ( ix !== 1 ) {
iu = this.extraHandler(l, r, ix);
if ( iu === 0 ) { return 0; }
} else {
iu = -1;
}
this.buf32[RESULT_IU_SLOT] = iu;
this.buf32[RESULT_L_SLOT] = l;
this.buf32[RESULT_R_SLOT] = r;
return 1;
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
get $l() { return this.buf32[RESULT_L_SLOT] | 0; }
get $r() { return this.buf32[RESULT_R_SLOT] | 0; }
get $iu() { return this.buf32[RESULT_IU_SLOT] | 0; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
add(iroot, aL0, n, pivot = 0) {
const aR = n;
if ( aR === 0 ) { return 0; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// Grow buffer if needed. The characters are already in our character
// data buffer, so we do not need to grow character data buffer.
if (
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
(this.buf32[CHAR0_SLOT] - this.buf32[TRIE1_SLOT]) <
MIN_FREE_CELL_BYTE_LENGTH
) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
this.growBuf(MIN_FREE_CELL_BYTE_LENGTH, 0);
}
const buf32 = this.buf32;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const char0 = buf32[CHAR0_SLOT];
let icell = iroot;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
let aL = char0 + aL0;
// special case: first node in trie
if ( buf32[icell+SEGMENT_INFO] === 0 ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
buf32[icell+SEGMENT_INFO] = toSegmentInfo(aL0, pivot, aR);
return this.addLeft(icell, aL0, pivot);
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const buf8 = this.buf8;
let al = pivot;
let inext;
// find a matching cell: move down
for (;;) {
const binfo = buf32[icell+SEGMENT_INFO];
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// length of segment
const bR = binfo >>> 24;
// skip boundary cells
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( bR === 0 ) {
icell = buf32[icell+BCELL_NEXT_AND];
continue;
}
let bl = char0 + (binfo & 0x00FFFFFF);
// if first character is no match, move to next descendant
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( buf8[bl] !== buf8[aL+al] ) {
inext = buf32[icell+CELL_OR];
if ( inext === 0 ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
inext = this.addCell(0, 0, toSegmentInfo(aL0, al, aR));
buf32[icell+CELL_OR] = inext;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return this.addLeft(inext, aL0, pivot);
}
icell = inext;
continue;
}
// 1st character was tested
let bi = 1;
al += 1;
// find 1st mismatch in rest of segment
if ( bR !== 1 ) {
for (;;) {
if ( bi === bR ) { break; }
if ( al === aR ) { break; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( buf8[bl+bi] !== buf8[aL+al] ) { break; }
bi += 1;
al += 1;
}
}
// all segment characters matched
if ( bi === bR ) {
// needle remainder: no
if ( al === aR ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return this.addLeft(icell, aL0, pivot);
}
// needle remainder: yes
inext = buf32[icell+CELL_AND];
if ( buf32[inext+CELL_AND] !== 0 ) {
icell = inext;
continue;
}
// add needle remainder
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
icell = this.addCell(0, 0, toSegmentInfo(aL0, al, aR));
buf32[inext+CELL_AND] = icell;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return this.addLeft(icell, aL0, pivot);
}
// some characters matched
// split current segment
bl -= char0;
buf32[icell+SEGMENT_INFO] = bi << 24 | bl;
inext = this.addCell(
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
buf32[icell+CELL_AND], 0, bR - bi << 24 | bl + bi
);
buf32[icell+CELL_AND] = inext;
// needle remainder: no = need boundary cell
if ( al === aR ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return this.addLeft(icell, aL0, pivot);
}
// needle remainder: yes = need new cell for remaining characters
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
icell = this.addCell(0, 0, toSegmentInfo(aL0, al, aR));
buf32[inext+CELL_OR] = icell;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return this.addLeft(icell, aL0, pivot);
}
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
addLeft(icell, aL0, pivot) {
const buf32 = this.buf32;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const char0 = buf32[CHAR0_SLOT];
let aL = aL0 + char0;
// fetch boundary cell
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
let iboundary = buf32[icell+CELL_AND];
// add boundary cell if none exist
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if (
iboundary === 0 ||
buf32[iboundary+SEGMENT_INFO] > BCELL_EXTRA_MAX
) {
const inext = iboundary;
iboundary = this.allocateCell();
buf32[icell+CELL_AND] = iboundary;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
buf32[iboundary+BCELL_NEXT_AND] = inext;
if ( pivot === 0 ) { return iboundary; }
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// shortest match with no extra conditions will always win
if ( buf32[iboundary+BCELL_EXTRA] === 1 ) {
return iboundary;
}
// bail out if no left segment
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( pivot === 0 ) { return iboundary; }
// fetch root cell of left segment
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
icell = buf32[iboundary+BCELL_ALT_AND];
if ( icell === 0 ) {
icell = this.allocateCell();
buf32[iboundary+BCELL_ALT_AND] = icell;
}
// special case: first node in trie
if ( buf32[icell+SEGMENT_INFO] === 0 ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
buf32[icell+SEGMENT_INFO] = toSegmentInfo(aL0, 0, pivot);
iboundary = this.allocateCell();
buf32[icell+CELL_AND] = iboundary;
return iboundary;
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
const buf8 = this.buf8;
let ar = pivot, inext;
// find a matching cell: move down
for (;;) {
const binfo = buf32[icell+SEGMENT_INFO];
// skip boundary cells
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( binfo <= BCELL_EXTRA_MAX ) {
inext = buf32[icell+CELL_AND];
if ( inext !== 0 ) {
icell = inext;
continue;
}
iboundary = this.allocateCell();
buf32[icell+CELL_AND] =
this.addCell(iboundary, 0, toSegmentInfo(aL0, 0, ar));
// TODO: boundary cell might be last
// add remainder + boundary cell
return iboundary;
}
const bL = char0 + (binfo & 0x00FFFFFF);
const bR = bL + (binfo >>> 24);
let br = bR;
// if first character is no match, move to next descendant
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( buf8[br-1] !== buf8[aL+ar-1] ) {
inext = buf32[icell+CELL_OR];
if ( inext === 0 ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
iboundary = this.allocateCell();
inext = this.addCell(
iboundary, 0, toSegmentInfo(aL0, 0, ar)
);
buf32[icell+CELL_OR] = inext;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
return iboundary;
}
icell = inext;
continue;
}
// 1st character was tested
br -= 1;
ar -= 1;
// find 1st mismatch in rest of segment
if ( br !== bL ) {
for (;;) {
if ( br === bL ) { break; }
if ( ar === 0 ) { break; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( buf8[br-1] !== buf8[aL+ar-1] ) { break; }
br -= 1;
ar -= 1;
}
}
// all segment characters matched
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// a: ...vvvvvvv
// b: vvvvvvv
if ( br === bL ) {
inext = buf32[icell+CELL_AND];
// needle remainder: no
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// a: vvvvvvv
// b: vvvvvvv
// r: 0 & vvvvvvv
if ( ar === 0 ) {
// boundary cell already present
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( buf32[inext+BCELL_EXTRA] <= BCELL_EXTRA_MAX ) {
return inext;
}
// need boundary cell
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
iboundary = this.allocateCell();
buf32[iboundary+CELL_AND] = inext;
buf32[icell+CELL_AND] = iboundary;
return iboundary;
}
// needle remainder: yes
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// a: yyyyyyyvvvvvvv
// b: vvvvvvv
else {
if ( inext !== 0 ) {
icell = inext;
continue;
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// TODO: we should never reach here because there will
// always be a boundary cell.
// eslint-disable-next-line no-debugger
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
debugger; // jshint ignore:line
// boundary cell + needle remainder
inext = this.addCell(0, 0, 0);
buf32[icell+CELL_AND] = inext;
buf32[inext+CELL_AND] =
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
this.addCell(0, 0, toSegmentInfo(aL0, 0, ar));
}
}
// some segment characters matched
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// a: ...vvvvvvv
// b: yyyyyyyvvvvvvv
else {
// split current cell
buf32[icell+SEGMENT_INFO] = (bR - br) << 24 | (br - char0);
inext = this.addCell(
buf32[icell+CELL_AND],
0,
(br - bL) << 24 | (bL - char0)
);
// needle remainder: no = need boundary cell
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// a: vvvvvvv
// b: yyyyyyyvvvvvvv
// r: yyyyyyy & 0 & vvvvvvv
if ( ar === 0 ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
iboundary = this.allocateCell();
buf32[icell+CELL_AND] = iboundary;
buf32[iboundary+CELL_AND] = inext;
return iboundary;
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
// needle remainder: yes = need new cell for remaining
// characters
// a: wwwwvvvvvvv
// b: yyyyyyyvvvvvvv
// r: (0 & wwww | yyyyyyy) & vvvvvvv
else {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
buf32[icell+CELL_AND] = inext;
iboundary = this.allocateCell();
buf32[inext+CELL_OR] = this.addCell(
iboundary, 0, toSegmentInfo(aL0, 0, ar)
);
return iboundary;
}
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
//debugger; // jshint ignore:line
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
getExtra(iboundary) {
return this.buf32[iboundary+BCELL_EXTRA];
}
setExtra(iboundary, v) {
this.buf32[iboundary+BCELL_EXTRA] = v;
}
optimize(shrink = false) {
if ( shrink ) {
this.shrinkBuf();
}
return {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
byteLength: this.buf8.byteLength,
char0: this.buf32[CHAR0_SLOT],
};
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
serialize(encoder) {
if ( encoder instanceof Object ) {
return encoder.encode(
this.buf32.buffer,
this.buf32[CHAR1_SLOT]
);
}
return Array.from(
new Uint32Array(
this.buf32.buffer,
0,
this.buf32[CHAR1_SLOT] + 3 >>> 2
)
);
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
unserialize(selfie, decoder) {
const shouldDecode = typeof selfie === 'string';
let byteLength = shouldDecode
? decoder.decodeSize(selfie)
: selfie.length << 2;
if ( byteLength === 0 ) { return false; }
this.reallocateBuf(byteLength);
if ( shouldDecode ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
decoder.decode(selfie, this.buf8.buffer);
} else {
this.buf32.set(selfie);
}
return true;
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
storeString(s) {
const n = s.length;
if ( n === this.lastStoredLen && s === this.lastStored ) {
return this.lastStoredIndex;
}
this.lastStored = s;
this.lastStoredLen = n;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
if ( (this.buf8.length - this.buf32[CHAR1_SLOT]) < n ) {
this.growBuf(0, n);
}
const offset = this.buf32[CHAR1_SLOT];
this.buf32[CHAR1_SLOT] = offset + n;
const buf8 = this.buf8;
for ( let i = 0; i < n; i++ ) {
buf8[offset+i] = s.charCodeAt(i);
}
return (this.lastStoredIndex = offset - this.buf32[CHAR0_SLOT]);
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
extractString(i, n) {
if ( this.textDecoder === null ) {
this.textDecoder = new TextDecoder();
}
const offset = this.buf32[CHAR0_SLOT] + i;
return this.textDecoder.decode(
this.buf8.subarray(offset, offset + n)
);
}
// WASMable.
startsWith(haystackLeft, haystackRight, needleLeft, needleLen) {
if ( haystackLeft < 0 || (haystackLeft + needleLen) > haystackRight ) {
return 0;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
const charCodes = this.buf8;
needleLeft += this.buf32[CHAR0_SLOT];
const needleRight = needleLeft + needleLen;
while ( charCodes[haystackLeft] === charCodes[needleLeft] ) {
needleLeft += 1;
if ( needleLeft === needleRight ) { return 1; }
haystackLeft += 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
return 0;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
// Find the left-most instance of substring in main string
// WASMable.
indexOf(haystackLeft, haystackEnd, needleLeft, needleLen) {
if ( needleLen === 0 ) { return haystackLeft; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
haystackEnd -= needleLen;
if ( haystackEnd < haystackLeft ) { return -1; }
needleLeft += this.buf32[CHAR0_SLOT];
const needleRight = needleLeft + needleLen;
const charCodes = this.buf8;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
for (;;) {
let i = haystackLeft;
let j = needleLeft;
while ( charCodes[i] === charCodes[j] ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
j += 1;
if ( j === needleRight ) { return haystackLeft; }
i += 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
haystackLeft += 1;
if ( haystackLeft > haystackEnd ) { break; }
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
return -1;
}
// Find the right-most instance of substring in main string.
// WASMable.
lastIndexOf(haystackBeg, haystackEnd, needleLeft, needleLen) {
if ( needleLen === 0 ) { return haystackBeg; }
let haystackLeft = haystackEnd - needleLen;
if ( haystackLeft < haystackBeg ) { return -1; }
needleLeft += this.buf32[CHAR0_SLOT];
const needleRight = needleLeft + needleLen;
const charCodes = this.buf8;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
for (;;) {
let i = haystackLeft;
let j = needleLeft;
while ( charCodes[i] === charCodes[j] ) {
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
j += 1;
if ( j === needleRight ) { return haystackLeft; }
i += 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
if ( haystackLeft === haystackBeg ) { break; }
haystackLeft -= 1;
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
}
return -1;
}
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
dumpTrie(iroot) {
for ( const s of this.trieIterator(iroot) ) {
console.log(s);
}
}
trieIterator(iroot) {
return {
value: undefined,
done: false,
next() {
if ( this.icell === 0 ) {
if ( this.forks.length === 0 ) {
this.value = undefined;
this.done = true;
return this;
}
this.charPtr = this.forks.pop();
this.icell = this.forks.pop();
}
for (;;) {
const idown = this.container.buf32[this.icell+CELL_OR];
if ( idown !== 0 ) {
this.forks.push(idown, this.charPtr);
}
const v = this.container.buf32[this.icell+SEGMENT_INFO];
let i0 = this.container.buf32[CHAR0_SLOT] + (v & 0x00FFFFFF);
const i1 = i0 + (v >>> 24);
while ( i0 < i1 ) {
this.charBuf[this.charPtr] = this.container.buf8[i0];
this.charPtr += 1;
i0 += 1;
}
this.icell = this.container.buf32[this.icell+CELL_AND];
if ( this.icell === 0 ) {
return this.toPattern();
}
if ( this.container.buf32[this.icell+SEGMENT_INFO] === 0 ) {
this.icell = this.container.buf32[this.icell+CELL_AND];
return this.toPattern();
}
}
},
toPattern() {
this.value = this.textDecoder.decode(
new Uint8Array(this.charBuf.buffer, 0, this.charPtr)
);
return this;
},
container: this,
icell: iroot,
charBuf: new Uint8Array(256),
charPtr: 0,
forks: [],
textDecoder: new TextDecoder(),
[Symbol.iterator]() { return this; },
};
}
async enableWASM(wasmModuleFetcher, path) {
if ( typeof WebAssembly !== 'object' ) { return false; }
if ( this.wasmMemory instanceof WebAssembly.Memory ) { return true; }
const module = await getWasmModule(wasmModuleFetcher, path);
if ( module instanceof WebAssembly.Module === false ) { return false; }
const memory = new WebAssembly.Memory({
initial: roundToPageSize(this.buf8.length) >>> 16
});
const instance = await WebAssembly.instantiate(module, {
imports: { memory, extraHandler: this.extraHandler }
});
if ( instance instanceof WebAssembly.Instance === false ) {
return false;
}
this.wasmMemory = memory;
const curPageCount = memory.buffer.byteLength >>> 16;
const newPageCount = roundToPageSize(this.buf8.byteLength) >>> 16;
if ( newPageCount > curPageCount ) {
memory.grow(newPageCount - curPageCount);
}
const buf8 = new Uint8Array(memory.buffer);
buf8.set(this.buf8);
this.buf8 = buf8;
this.buf32 = new Uint32Array(this.buf8.buffer);
this.haystack = this.buf8.subarray(
HAYSTACK_START,
HAYSTACK_START + HAYSTACK_SIZE
);
this.matches = instance.exports.matches;
this.startsWith = instance.exports.startsWith;
this.indexOf = instance.exports.indexOf;
this.lastIndexOf = instance.exports.lastIndexOf;
return true;
}
dumpInfo() {
return [
`Buffer size (Uint8Array): ${this.buf32[CHAR1_SLOT].toLocaleString('en')}`,
`WASM: ${this.wasmMemory === null ? 'disabled' : 'enabled'}`,
].join('\n');
}
//--------------------------------------------------------------------------
// Private methods
//--------------------------------------------------------------------------
allocateCell() {
let icell = this.buf32[TRIE1_SLOT];
this.buf32[TRIE1_SLOT] = icell + CELL_BYTE_LENGTH;
icell >>>= 2;
this.buf32[icell+0] = 0;
this.buf32[icell+1] = 0;
this.buf32[icell+2] = 0;
return icell;
}
addCell(iand, ior, v) {
const icell = this.allocateCell();
this.buf32[icell+CELL_AND] = iand;
this.buf32[icell+CELL_OR] = ior;
this.buf32[icell+SEGMENT_INFO] = v;
return icell;
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
growBuf(trieGrow, charGrow) {
const char0 = Math.max(
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
roundToPageSize(this.buf32[TRIE1_SLOT] + trieGrow),
this.buf32[CHAR0_SLOT]
);
const char1 = char0 + this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT];
const bufLen = Math.max(
Expand bidi-trie usage in static network filtering engine Related issues: - https://github.com/uBlockOrigin/uBlock-issues/issues/761 - https://github.com/uBlockOrigin/uBlock-issues/issues/528 The previous bidi-trie code could only hold filters which are plain pattern, i.e. no wildcard characters, and which had no origin option (`domain=`), right and/or left anchor, and no `csp=` option. Example of filters that could be moved into a bidi-trie data structure: &ad_box_ /w/d/capu.php?z=$script,third-party ||liveonlinetv247.com/images/muvixx-150x50-watch-now-in-hd-play-btn.gif Examples of filters that could NOT be moved to a bidi-trie: -adap.$domain=~l-adap.org /tsc.php?*&ses= ||ibsrv.net/*forumsponsor$domain=[...] @@||imgspice.com/jquery.cookie.js|$script ||view.atdmt.com^*/iview/$third-party ||postimg.cc/image/$csp=[...] Ideally the filters above should be able to be moved to a bidi-trie since they are basically plain patterns, or at least partially moved to a bidi-trie when there is only a single wildcard (i.e. made of two plain patterns). Also, there were two distinct bidi-tries in which plain-pattern filters can be moved to: one for patterns without hostname anchoring and another one for patterns with hostname-anchoring. This was required because the hostname-anchored patterns have an extra condition which is outside the bidi-trie knowledge. This commit expands the number of filters which can be stored in the bidi-trie, and also remove the need to use two distinct bidi-tries. - Added ability to associate a pattern with an integer in the bidi-trie [1]. - The bidi-trie match code passes this externally provided integer when calling an externally provided method used for testing extra conditions that may be present for a plain pattern found to be matching in the bidi-trie. - Decomposed existing filters into smaller logical units: - FilterPlainLeftAnchored => FilterPatternPlain + FilterAnchorLeft - FilterPlainRightAnchored => FilterPatternPlain + FilterAnchorRight - FilterExactMatch => FilterPatternPlain + FilterAnchorLeft + FilterAnchorRight - FilterPlainHnAnchored => FilterPatternPlain + FilterAnchorHn - FilterWildcard1 => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] - FilterWildcard1HnAnchored => FilterPatternPlain + [ FilterPatternLeft or FilterPatternRight ] + FilterAnchorHn - FilterGenericHnAnchored => FilterPatternGeneric + FilterAnchorHn - FilterGenericHnAndRightAnchored => FilterPatternGeneric + FilterAnchorRight + FilterAnchorHn - FilterOriginMixedSet => FilterOriginMissSet + FilterOriginHitSet - Instances of FilterOrigin[...], FilterDataHolder can also be added to a composite filter to represent `domain=` and `csp=` options. - Added a new filter class, FilterComposite, for filters which are a combination of two or more logical units. A FilterComposite instance is a match when *all* filters composing it are a match. Since filters are now encoded into combination of smaller units, it becomes possible to extract the FilterPatternPlain component and store it in the bidi-trie, and use the integer as a handle for the remaining extra conditions, if any. Since a single pattern in the bidi-trie may be a component for different filters, the associated integer points to a sequence of extra conditions, and a match occurs as soon as one of the extra conditions (which may itself be a sequence of conditions) is fulfilled. Decomposing filters which are currently single instance into sequences of smaller logical filters means increasing the storage and CPU overhead when evaluating such filters. The CPU overhead is compensated by the fact that more filters can now moved into the bidi-trie, where the first match is efficiently evaluated. The extra conditions have to be evaluated if and only if there is a match in the bidi-trie. The storage overhead is compensated by the bidi-trie's intrinsic nature of merging similar patterns. Furthermore, the storage overhead is reduced by no longer using JavaScript array to store collection of filters (which is what FilterComposite is): the same technique used in [2] is imported to store sequences of filters. A sequence of filters is a sequence of integer pairs where the first integer is an index to an actual filter instance stored in a global array of filters (`filterUnits`), while the second integer is an index to the next pair in the sequence -- which means all sequences of filters are encoded in one single array of integers (`filterSequences` => Uint32Array). As a result, a sequence of filters can be represented by one single integer -- an index to the first pair -- regardless of the number of filters in the sequence. This representation is further leveraged to replace the use of JavaScript array in FilterBucket [3], which used a JavaScript array to store collection of filters. Doing so means there is no more need for FilterPair [4], which purpose was to be a lightweight representation when there was only two filters in a collection. As a result of the above changes, the map of `token` (integer) => filter instance (object) used to associate tokens to filters or collections of filters is replaced with a more efficient map of `token` (integer) to filter unit index (integer) to lookup a filter object from the global `filterUnits` array. Another consequence of using one single global array to store all filter instances means we can reuse existing instances when a logical filter instance is parameter-less, which is the case for FilterAnchorLeft, FilterAnchorRight, FilterAnchorHn, the index to these single instances is reused where needed. `urlTokenizer` now stores the character codes of the scanned URL into a bidi-trie buffer, for reuse when string matching methods are called. New method: `tokenHistogram()`, used to generate histograms of occurrences of token extracted from URLs in built-in benchmark. The top results of the "miss" histogram are used as "bad tokens", i.e. tokens to avoid if possible when compiling filter lists. All plain pattern strings are now stored in the bidi-trie memory buffer, regardless of whether they will be used in the trie proper or not. Three methods have been added to the bidi-trie to test stored string against the URL which is also stored in then bidi-trie. FilterParser is now instanciated on demand and released when no longer used. *** [1] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/strie.js#L120 [2] https://github.com/gorhill/uBlock/commit/e94024d350b066e4e04a772b0a3dbc69daab3fb7 [3] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1630 [4] https://github.com/gorhill/uBlock/blob/135a45a878f5b93bc538f822981e3a42b1e9073f/src/js/static-net-filtering.js#L1566
2019-10-21 14:15:58 +02:00
roundToPageSize(char1 + charGrow),
this.buf8.length
);
if ( bufLen > this.buf8.length ) {
this.reallocateBuf(bufLen);
}
if ( char0 !== this.buf32[CHAR0_SLOT] ) {
this.buf8.copyWithin(
char0,
this.buf32[CHAR0_SLOT],
this.buf32[CHAR1_SLOT]
);
this.buf32[CHAR0_SLOT] = char0;
this.buf32[CHAR1_SLOT] = char1;
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
shrinkBuf() {
const char0 = this.buf32[TRIE1_SLOT] + MIN_FREE_CELL_BYTE_LENGTH;
const char1 = char0 + this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT];
const bufLen = char1 + 256;
if ( char0 !== this.buf32[CHAR0_SLOT] ) {
this.buf8.copyWithin(
char0,
this.buf32[CHAR0_SLOT],
this.buf32[CHAR1_SLOT]
);
this.buf32[CHAR0_SLOT] = char0;
this.buf32[CHAR1_SLOT] = char1;
}
if ( bufLen < this.buf8.length ) {
this.reallocateBuf(bufLen);
}
}
reallocateBuf(newSize) {
newSize = roundToPageSize(newSize);
if ( newSize === this.buf8.length ) { return; }
if ( this.wasmMemory === null ) {
const newBuf = new Uint8Array(newSize);
newBuf.set(
newBuf.length < this.buf8.length
? this.buf8.subarray(0, newBuf.length)
: this.buf8
);
this.buf8 = newBuf;
} else {
const growBy =
((newSize + 0xFFFF) >>> 16) - (this.buf8.length >>> 16);
if ( growBy <= 0 ) { return; }
this.wasmMemory.grow(growBy);
this.buf8 = new Uint8Array(this.wasmMemory.buffer);
}
this.buf32 = new Uint32Array(this.buf8.buffer);
this.haystack = this.buf8.subarray(
HAYSTACK_START,
HAYSTACK_START + HAYSTACK_SIZE
);
Add HNTrie-based filter classes to store origin-only filters Related issue: - https://github.com/uBlockOrigin/uBlock-issues/issues/528#issuecomment-484408622 Following STrie-related work in above issue, I noticed that a large number of filters in EasyList were filters which only had to match against the document origin. For instance, among just the top 10 most populous buckets, there were four such buckets with over hundreds of entries each: - bits: 72, token: "http", 146 entries - bits: 72, token: "https", 139 entries - bits: 88, token: "http", 122 entries - bits: 88, token: "https", 118 entries These filters in these buckets have to be matched against all the network requests. In order to leverage HNTrie for these filters[1], they are now handled in a special way so as to ensure they all end up in a single HNTrie (per bucket), which means that instead of scanning hundreds of entries per URL, there is now a single scan per bucket per URL for these apply-everywhere filters. Now, any filter which fulfill ALL the following condition will be processed in a special manner internally: - Is of the form `|https://` or `|http://` or `*`; and - Does have a `domain=` option; and - Does not have a negated domain in its `domain=` option; and - Does not have `csp=` option; and - Does not have a `redirect=` option If a filter does not fulfill ALL the conditions above, no change in behavior. A filter which matches ALL of the above will be processed in a special manner: - The `domain=` option will be decomposed so as to create as many distinct filter as there is distinct value in the `domain=` option - This also apply to the `badfilter` version of the filter, which means it now become possible to `badfilter` only one of the distinct filter without having to `badfilter` all of them. - The logger will always report these special filters with only a single hostname in the `domain=` option. *** [1] HNTrie is currently WASM-ed on Firefox.
2019-04-19 22:33:46 +02:00
}
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
}
/******************************************************************************/
// Code below is to attempt to load a WASM module which implements:
//
// - BidiTrieContainer.startsWith()
//
// The WASM module is entirely optional, the JS implementations will be
// used should the WASM module be unavailable for whatever reason.
const getWasmModule = (( ) => {
let wasmModulePromise;
return async function(wasmModuleFetcher, path) {
if ( wasmModulePromise instanceof Promise ) {
return wasmModulePromise;
}
if ( typeof WebAssembly !== 'object' ) { return; }
// Soft-dependency on vAPI so that the code here can be used outside of
// uBO (i.e. tests, benchmarks)
if ( typeof vAPI === 'object' && vAPI.canWASM !== true ) { return; }
// The wasm module will work only if CPU is natively little-endian,
// as we use native uint32 array in our js code.
const uint32s = new Uint32Array(1);
const uint8s = new Uint8Array(uint32s.buffer);
uint32s[0] = 1;
if ( uint8s[0] !== 1 ) { return; }
wasmModulePromise = wasmModuleFetcher(`${path}biditrie`).catch(reason => {
console.info(reason);
});
return wasmModulePromise;
};
})();
/******************************************************************************/
Refactoring work in static network filtering engine The original motivation is to further speed up launch time for either non-selfie-based and selfie-based initialization of the static network filtering engine (SNFE). As a result of the refactoring: Filters are no longer instance-based, they are sequence-of- integer-based. This eliminates the need to create instances of filters at launch, and consequently eliminates all the calls to class constructors, the resulting churning of memory, and so forth. All the properties defining filter instances are now as much as possible 32-bit integer-based, and these are allocated in a single module-scoped typed array -- this eliminates the need to allocate memory for every filter being instantiated. Not all filter properties can be represented as a 32-bit integer, and in this case a filter class can allocate slots into another module-scoped array of references. As a result, this eliminates a lot of memory allocations when the SNFE is populated with filters, and this makes the saving and loading of selfie more straightforward, as the operation is reduced to saving/loading two arrays, one of 32-bit integers, and the other, much smaller, an array JSON-able values. All filter classes now only contain static methods, and all of these methods are called with an index to the specific filter data in the module-scoped array of 32-bit integers. The filter sequences (used to avoid the use of JS arrays) are also allocated in the single module-scoped array of 32-bit integers -- they used to be stored in their own dedicated array. Additionally, some filters are now loaded more in a deferred way, so as reduce uBO's time-to-readiness -- the outcome of this still needs to be evaluated, time-to-readiness is especially a concern in Firefox for Android or less powerful computers.
2021-12-04 17:16:44 +01:00
export default BidiTrieContainer;