mirror of
https://github.com/mikf/gallery-dl.git
synced 2024-11-25 20:22:36 +01:00
[twitter] add extractor
This commit is contained in:
parent
e3926bf351
commit
bf8d88499a
16
README.rst
16
README.rst
@ -40,22 +40,24 @@ Supported Sites
|
||||
behoimi.org, danbooru.donmai.us, e621.net, gelbooru.com, konachan.com,
|
||||
rule34.xxx, safebooru.org, chan.sankakucomplex.com, yande.re
|
||||
* Manga:
|
||||
bato.to, dynasty-scans.com, kissmanga.com, mangahere.co, mangamint.com,
|
||||
mangapanda.com, mangapark.me, mangareader.net, mangashare.com,
|
||||
mangastream.com, powermanga.org, raw.senmanga.com, thespectrum.net
|
||||
bato.to, dynasty-scans.com, kissmanga.com, kobato.hologfx.com,
|
||||
mangahere.co, mangamint.com, mangapanda.com, mangapark.me, mangareader.net,
|
||||
mangashare.com, mangastream.com, powermanga.org, raw.senmanga.com,
|
||||
thespectrum.net
|
||||
* Hentai:
|
||||
doujinmode.net, exhentai.org, hbrowse.com, hentai2read.com,
|
||||
hentaibox.net, hitomi.la, luscious.net, nhentai.net
|
||||
hentaibox.net, hentaihere.com, hitomi.la, luscious.net, nhentai.net
|
||||
* Japanese:
|
||||
pixiv.net, nijie.info, seiga.nicovideo.jp
|
||||
* Western:
|
||||
deviantart.com, hentai-foundry.com, imagefap,com, imgth.com, imgur.com,
|
||||
pinterest.com, tumblr.com
|
||||
pinterest.com, tumblr.com, twitter.com
|
||||
* Futaba Channel-like:
|
||||
4chan.org, 8ch.net
|
||||
* Image Hosts:
|
||||
chronos.to, coreimg.net, imagebam.com, imagetwist.com, img.yt, imgbox.com,
|
||||
imgcandy.net, imgchili.net, imgtrex.com, turboimagehost.com
|
||||
imgcandy.net, imgchili.net, imgtrex.com, pic-maniac.com, rapidimg.net,
|
||||
turboimagehost.com
|
||||
|
||||
|
||||
Configuration
|
||||
@ -88,7 +90,7 @@ You can set the necessary information in your configuration file
|
||||
{
|
||||
"extractor": {
|
||||
...
|
||||
"pixiv:" {
|
||||
"pixiv": {
|
||||
"username": "<username>",
|
||||
"password": "<password>"
|
||||
}
|
||||
|
@ -67,6 +67,7 @@ modules = [
|
||||
"spectrumnexus",
|
||||
"tumblr",
|
||||
"turboimagehost",
|
||||
"twitter",
|
||||
"yandere",
|
||||
"recursive",
|
||||
]
|
||||
|
56
gallery_dl/extractor/twitter.py
Normal file
56
gallery_dl/extractor/twitter.py
Normal file
@ -0,0 +1,56 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
|
||||
# Copyright 2016 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.
|
||||
|
||||
"""Extract images from https://twitter.com/"""
|
||||
|
||||
from .common import Extractor, Message
|
||||
from .. import text
|
||||
|
||||
class TwitterTweetExtractor(Extractor):
|
||||
"""Extractor for images from twitter tweets"""
|
||||
category = "twitter"
|
||||
subcategory = "tweet"
|
||||
directory_fmt = ["{category}", "{user}"]
|
||||
filename_fmt = "{tweet-id}_{num}.{extension}"
|
||||
pattern = [r"(?:https?://)?(?:www\.|mobile\.)?twitter\.com/(([^/]+)/status/(\d+))"]
|
||||
test = [("https://twitter.com/PicturesEarth/status/672897688871018500", {
|
||||
"url": "ec37d74e63fc751b9d1838415e3fe60d4e8cbe43",
|
||||
"keyword": "3cd8e27026a2112008985b1b53f5e4baf4616177",
|
||||
"content": "f0fe8f9cd428580cf082fda3e31c0cc436764693",
|
||||
})]
|
||||
|
||||
def __init__(self, match):
|
||||
Extractor.__init__(self)
|
||||
self.path, self.user, self.tid = match.groups()
|
||||
|
||||
def items(self):
|
||||
self.session.headers["User-Agent"] = (
|
||||
"Mozilla/5.0 (X11; Linux x86_64; rv:48.0) "
|
||||
"Gecko/20100101 Firefox/48.0"
|
||||
)
|
||||
page = self.request("https://mobile.twitter.com/" + self.path).text
|
||||
data = self.get_job_metadata()
|
||||
imgs = self.get_image_urls(page)
|
||||
data["count"] = len(imgs)
|
||||
yield Message.Version, 1
|
||||
yield Message.Directory, data
|
||||
for num, url in enumerate(imgs, 1):
|
||||
data["num"] = num
|
||||
yield Message.Url, url + ":large", text.nameext_from_url(url, data)
|
||||
|
||||
def get_job_metadata(self):
|
||||
"""Collect metadata for extractor-job"""
|
||||
return {
|
||||
"user": self.user,
|
||||
"tweet-id": self.tid,
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def get_image_urls(page):
|
||||
"""Extract and return a list of all image-urls"""
|
||||
return list(text.extract_iter(page, 'alt="Embedded image" src="', '"'))
|
Loading…
Reference in New Issue
Block a user