1
0
mirror of https://github.com/mikf/gallery-dl.git synced 2024-11-23 11:12:40 +01:00
gallery-dl/gallery_dl/extractor/slickpic.py
Mike Fährmann bd08ee2859
remove most 'yield Message.Version' statements
only leave them in oauth.py as noop results
2021-08-16 03:10:48 +02:00

141 lines
4.5 KiB
Python

# -*- coding: utf-8 -*-
# Copyright 2019-2020 Mike Fährmann
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License version 2 as
# published by the Free Software Foundation.
"""Extractors for https://www.slickpic.com/"""
from .common import Extractor, Message
from .. import text
import time
BASE_PATTERN = r"(?:https?://)?([^.]+)\.slickpic\.com"
class SlickpicExtractor(Extractor):
"""Base class for slickpic extractors"""
category = "slickpic"
def __init__(self, match):
Extractor.__init__(self, match)
self.user = match.group(1)
self.root = "https://{}.slickpic.com".format(self.user)
class SlickpicAlbumExtractor(SlickpicExtractor):
"""Extractor for albums on slickpic.com"""
subcategory = "album"
directory_fmt = ("{category}", "{user[name]}",
"{album[id]} {album[title]}")
filename_fmt = "{num:>03}_{id}{title:?_//}.{extension}"
archive_fmt = "{id}"
pattern = BASE_PATTERN + r"/albums/([^/?#]+)"
test = (
("https://mattcrandall.slickpic.com/albums/LamborghiniMurcielago/", {
"url": "58bd94ebc80fd906e9879826970b408d54c6da07",
"keyword": "54a9d6f9e42ae43c644aa9316186fb9d9955fe53",
}),
("https://mattcrandall.slickpic.com/albums/LamborghiniMurcielago/", {
"range": "34",
"content": ("cec6630e659dc72db1ee1a9a6f3b525189261988",
"6f81e1e74c6cd6db36844e7211eef8e7cd30055d",
"22e83645fc242bc3584eca7ec982c8a53a4d8a44"),
}),
)
def __init__(self, match):
SlickpicExtractor.__init__(self, match)
self.album = match.group(2)
def items(self):
data = self.metadata()
imgs = self.images(data)
data = {
"album": {
"id" : text.parse_int(data["aid"]),
"title": text.unescape(data["title"]),
},
"user": {
"id" : text.parse_int(data["uid"]),
"name": text.unescape(data["user"]),
"nick": self.user
},
"count": len(imgs),
}
yield Message.Directory, data
for num, img in enumerate(imgs, 1):
url = img["url_rsz"] + "/o/" + img["fname"]
img = text.nameext_from_url(img["fname"], {
"url" : url,
"num" : num,
"id" : text.parse_int(img["id"]),
"width" : text.parse_int(img["width"]),
"height" : text.parse_int(img["height"]),
"title" : img["title"],
"description": img["descr"],
})
img.update(data)
yield Message.Url, url, img
def metadata(self):
url = "{}/albums/{}/?wallpaper".format(self.root, self.album)
extr = text.extract_from(self.request(url).text)
title = text.unescape(extr("<title>", "</title>"))
title, _, user = title.rpartition(" by ")
return {
"title": title,
"user" : user,
"tk" : extr('tk = "', '"'),
"shd" : extr('shd = "', '"'),
"aid" : extr('data-aid="', '"', ),
"uid" : extr('data-uid="', '"', ),
}
def images(self, data):
url = self.root + "/xhr/photo/get/list"
data = {
"tm" : time.time(),
"tk" : data["tk"],
"shd" : data["shd"],
"aid" : data["aid"],
"uid" : data["uid"],
"col" : "0",
"sys" : self.album,
"vw" : "1280",
"vh" : "1024",
"skey" : "",
"viewer": "false",
"pub" : "1",
"sng" : "0",
"whq" : "1",
}
return self.request(url, method="POST", data=data).json()["list"]
class SlickpicUserExtractor(SlickpicExtractor):
subcategory = "user"
pattern = BASE_PATTERN + r"(?:/gallery)?/?(?:$|[?#])"
test = (
("https://mattcrandall.slickpic.com/gallery/", {
"count": ">= 358",
"pattern": SlickpicAlbumExtractor.pattern,
}),
("https://mattcrandall.slickpic.com/"),
)
def items(self):
page = self.request(self.root + "/gallery?viewer").text
data = {"_extractor": SlickpicAlbumExtractor}
base = self.root + "/albums/"
for album in text.extract_iter(page, 'href="' + base, '"'):
yield Message.Queue, base + album, data