1
0
mirror of https://github.com/l1ving/youtube-dl synced 2024-12-01 00:23:03 +08:00
youtube-dl/youtube_dl/extractor/twitch.py
Sergey M? 5c2266df4b Switch codebase to use sanitized_Request instead of
compat_urllib_request.Request

[downloader/dash] Use sanitized_Request

[downloader/http] Use sanitized_Request

[atresplayer] Use sanitized_Request

[bambuser] Use sanitized_Request

[bliptv] Use sanitized_Request

[brightcove] Use sanitized_Request

[cbs] Use sanitized_Request

[ceskatelevize] Use sanitized_Request

[collegerama] Use sanitized_Request

[extractor/common] Use sanitized_Request

[crunchyroll] Use sanitized_Request

[dailymotion] Use sanitized_Request

[dcn] Use sanitized_Request

[dramafever] Use sanitized_Request

[dumpert] Use sanitized_Request

[eitb] Use sanitized_Request

[escapist] Use sanitized_Request

[everyonesmixtape] Use sanitized_Request

[extremetube] Use sanitized_Request

[facebook] Use sanitized_Request

[fc2] Use sanitized_Request

[flickr] Use sanitized_Request

[4tube] Use sanitized_Request

[gdcvault] Use sanitized_Request

[extractor/generic] Use sanitized_Request

[hearthisat] Use sanitized_Request

[hotnewhiphop] Use sanitized_Request

[hypem] Use sanitized_Request

[iprima] Use sanitized_Request

[ivi] Use sanitized_Request

[keezmovies] Use sanitized_Request

[letv] Use sanitized_Request

[lynda] Use sanitized_Request

[metacafe] Use sanitized_Request

[minhateca] Use sanitized_Request

[miomio] Use sanitized_Request

[meovideo] Use sanitized_Request

[mofosex] Use sanitized_Request

[moniker] Use sanitized_Request

[mooshare] Use sanitized_Request

[movieclips] Use sanitized_Request

[mtv] Use sanitized_Request

[myvideo] Use sanitized_Request

[neteasemusic] Use sanitized_Request

[nfb] Use sanitized_Request

[niconico] Use sanitized_Request

[noco] Use sanitized_Request

[nosvideo] Use sanitized_Request

[novamov] Use sanitized_Request

[nowness] Use sanitized_Request

[nuvid] Use sanitized_Request

[played] Use sanitized_Request

[pluralsight] Use sanitized_Request

[pornhub] Use sanitized_Request

[pornotube] Use sanitized_Request

[primesharetv] Use sanitized_Request

[promptfile] Use sanitized_Request

[qqmusic] Use sanitized_Request

[rtve] Use sanitized_Request

[safari] Use sanitized_Request

[sandia] Use sanitized_Request

[shared] Use sanitized_Request

[sharesix] Use sanitized_Request

[sina] Use sanitized_Request

[smotri] Use sanitized_Request

[sohu] Use sanitized_Request

[spankwire] Use sanitized_Request

[sportdeutschland] Use sanitized_Request

[streamcloud] Use sanitized_Request

[streamcz] Use sanitized_Request

[tapely] Use sanitized_Request

[tube8] Use sanitized_Request

[tubitv] Use sanitized_Request

[twitch] Use sanitized_Request

[twitter] Use sanitized_Request

[udemy] Use sanitized_Request

[vbox7] Use sanitized_Request

[veoh] Use sanitized_Request

[vessel] Use sanitized_Request

[vevo] Use sanitized_Request

[viddler] Use sanitized_Request

[videomega] Use sanitized_Request

[viewvster] Use sanitized_Request

[viki] Use sanitized_Request

[vk] Use sanitized_Request

[vodlocker] Use sanitized_Request

[voicerepublic] Use sanitized_Request

[wistia] Use sanitized_Request

[xfileshare] Use sanitized_Request

[xtube] Use sanitized_Request

[xvideos] Use sanitized_Request

[yandexmusic] Use sanitized_Request

[youku] Use sanitized_Request

[youporn] Use sanitized_Request

[youtube] Use sanitized_Request

[patreon] Use sanitized_Request

[extractor/common] Remove unused import

[nfb] PEP 8
2015-11-23 21:56:23 +06:00

457 lines
16 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
import itertools
import re
import random
from .common import InfoExtractor
from ..compat import (
compat_parse_qs,
compat_str,
compat_urllib_parse,
compat_urllib_parse_urlparse,
compat_urlparse,
)
from ..utils import (
encode_dict,
ExtractorError,
int_or_none,
parse_duration,
parse_iso8601,
sanitized_Request,
)
class TwitchBaseIE(InfoExtractor):
_VALID_URL_BASE = r'https?://(?:www\.)?twitch\.tv'
_API_BASE = 'https://api.twitch.tv'
_USHER_BASE = 'http://usher.twitch.tv'
_LOGIN_URL = 'http://www.twitch.tv/login'
_NETRC_MACHINE = 'twitch'
def _handle_error(self, response):
if not isinstance(response, dict):
return
error = response.get('error')
if error:
raise ExtractorError(
'%s returned error: %s - %s' % (self.IE_NAME, error, response.get('message')),
expected=True)
def _download_json(self, url, video_id, note='Downloading JSON metadata'):
headers = {
'Referer': 'http://api.twitch.tv/crossdomain/receiver.html?v=2',
'X-Requested-With': 'XMLHttpRequest',
}
for cookie in self._downloader.cookiejar:
if cookie.name == 'api_token':
headers['Twitch-Api-Token'] = cookie.value
request = sanitized_Request(url, headers=headers)
response = super(TwitchBaseIE, self)._download_json(request, video_id, note)
self._handle_error(response)
return response
def _real_initialize(self):
self._login()
def _login(self):
(username, password) = self._get_login_info()
if username is None:
return
login_page, handle = self._download_webpage_handle(
self._LOGIN_URL, None, 'Downloading login page')
login_form = self._hidden_inputs(login_page)
login_form.update({
'username': username,
'password': password,
})
redirect_url = handle.geturl()
post_url = self._search_regex(
r'<form[^>]+action=(["\'])(?P<url>.+?)\1', login_page,
'post url', default=redirect_url, group='url')
if not post_url.startswith('http'):
post_url = compat_urlparse.urljoin(redirect_url, post_url)
request = sanitized_Request(
post_url, compat_urllib_parse.urlencode(encode_dict(login_form)).encode('utf-8'))
request.add_header('Referer', redirect_url)
response = self._download_webpage(
request, None, 'Logging in as %s' % username)
error_message = self._search_regex(
r'<div[^>]+class="subwindow_notice"[^>]*>([^<]+)</div>',
response, 'error message', default=None)
if error_message:
raise ExtractorError(
'Unable to login. Twitch said: %s' % error_message, expected=True)
if '>Reset your password<' in response:
self.report_warning('Twitch asks you to reset your password, go to https://secure.twitch.tv/reset/submit')
def _prefer_source(self, formats):
try:
source = next(f for f in formats if f['format_id'] == 'Source')
source['preference'] = 10
except StopIteration:
pass # No Source stream present
self._sort_formats(formats)
class TwitchItemBaseIE(TwitchBaseIE):
def _download_info(self, item, item_id):
return self._extract_info(self._download_json(
'%s/kraken/videos/%s%s' % (self._API_BASE, item, item_id), item_id,
'Downloading %s info JSON' % self._ITEM_TYPE))
def _extract_media(self, item_id):
info = self._download_info(self._ITEM_SHORTCUT, item_id)
response = self._download_json(
'%s/api/videos/%s%s' % (self._API_BASE, self._ITEM_SHORTCUT, item_id), item_id,
'Downloading %s playlist JSON' % self._ITEM_TYPE)
entries = []
chunks = response['chunks']
qualities = list(chunks.keys())
for num, fragment in enumerate(zip(*chunks.values()), start=1):
formats = []
for fmt_num, fragment_fmt in enumerate(fragment):
format_id = qualities[fmt_num]
fmt = {
'url': fragment_fmt['url'],
'format_id': format_id,
'quality': 1 if format_id == 'live' else 0,
}
m = re.search(r'^(?P<height>\d+)[Pp]', format_id)
if m:
fmt['height'] = int(m.group('height'))
formats.append(fmt)
self._sort_formats(formats)
entry = dict(info)
entry['id'] = '%s_%d' % (entry['id'], num)
entry['title'] = '%s part %d' % (entry['title'], num)
entry['formats'] = formats
entries.append(entry)
return self.playlist_result(entries, info['id'], info['title'])
def _extract_info(self, info):
return {
'id': info['_id'],
'title': info.get('title') or 'Untitled Broadcast',
'description': info.get('description'),
'duration': int_or_none(info.get('length')),
'thumbnail': info.get('preview'),
'uploader': info.get('channel', {}).get('display_name'),
'uploader_id': info.get('channel', {}).get('name'),
'timestamp': parse_iso8601(info.get('recorded_at')),
'view_count': int_or_none(info.get('views')),
}
def _real_extract(self, url):
return self._extract_media(self._match_id(url))
class TwitchVideoIE(TwitchItemBaseIE):
IE_NAME = 'twitch:video'
_VALID_URL = r'%s/[^/]+/b/(?P<id>\d+)' % TwitchBaseIE._VALID_URL_BASE
_ITEM_TYPE = 'video'
_ITEM_SHORTCUT = 'a'
_TEST = {
'url': 'http://www.twitch.tv/riotgames/b/577357806',
'info_dict': {
'id': 'a577357806',
'title': 'Worlds Semifinals - Star Horn Royal Club vs. OMG',
},
'playlist_mincount': 12,
}
class TwitchChapterIE(TwitchItemBaseIE):
IE_NAME = 'twitch:chapter'
_VALID_URL = r'%s/[^/]+/c/(?P<id>\d+)' % TwitchBaseIE._VALID_URL_BASE
_ITEM_TYPE = 'chapter'
_ITEM_SHORTCUT = 'c'
_TESTS = [{
'url': 'http://www.twitch.tv/acracingleague/c/5285812',
'info_dict': {
'id': 'c5285812',
'title': 'ACRL Off Season - Sports Cars @ Nordschleife',
},
'playlist_mincount': 3,
}, {
'url': 'http://www.twitch.tv/tsm_theoddone/c/2349361',
'only_matching': True,
}]
class TwitchVodIE(TwitchItemBaseIE):
IE_NAME = 'twitch:vod'
_VALID_URL = r'%s/[^/]+/v/(?P<id>\d+)' % TwitchBaseIE._VALID_URL_BASE
_ITEM_TYPE = 'vod'
_ITEM_SHORTCUT = 'v'
_TESTS = [{
'url': 'http://www.twitch.tv/riotgames/v/6528877?t=5m10s',
'info_dict': {
'id': 'v6528877',
'ext': 'mp4',
'title': 'LCK Summer Split - Week 6 Day 1',
'thumbnail': 're:^https?://.*\.jpg$',
'duration': 17208,
'timestamp': 1435131709,
'upload_date': '20150624',
'uploader': 'Riot Games',
'uploader_id': 'riotgames',
'view_count': int,
'start_time': 310,
},
'params': {
# m3u8 download
'skip_download': True,
},
}, {
# Untitled broadcast (title is None)
'url': 'http://www.twitch.tv/belkao_o/v/11230755',
'info_dict': {
'id': 'v11230755',
'ext': 'mp4',
'title': 'Untitled Broadcast',
'thumbnail': 're:^https?://.*\.jpg$',
'duration': 1638,
'timestamp': 1439746708,
'upload_date': '20150816',
'uploader': 'BelkAO_o',
'uploader_id': 'belkao_o',
'view_count': int,
},
'params': {
# m3u8 download
'skip_download': True,
},
}]
def _real_extract(self, url):
item_id = self._match_id(url)
info = self._download_info(self._ITEM_SHORTCUT, item_id)
access_token = self._download_json(
'%s/api/vods/%s/access_token' % (self._API_BASE, item_id), item_id,
'Downloading %s access token' % self._ITEM_TYPE)
formats = self._extract_m3u8_formats(
'%s/vod/%s?%s' % (
self._USHER_BASE, item_id,
compat_urllib_parse.urlencode({
'allow_source': 'true',
'allow_spectre': 'true',
'player': 'twitchweb',
'nauth': access_token['token'],
'nauthsig': access_token['sig'],
})),
item_id, 'mp4')
self._prefer_source(formats)
info['formats'] = formats
parsed_url = compat_urllib_parse_urlparse(url)
query = compat_parse_qs(parsed_url.query)
if 't' in query:
info['start_time'] = parse_duration(query['t'][0])
return info
class TwitchPlaylistBaseIE(TwitchBaseIE):
_PLAYLIST_URL = '%s/kraken/channels/%%s/videos/?offset=%%d&limit=%%d' % TwitchBaseIE._API_BASE
_PAGE_LIMIT = 100
def _extract_playlist(self, channel_id):
info = self._download_json(
'%s/kraken/channels/%s' % (self._API_BASE, channel_id),
channel_id, 'Downloading channel info JSON')
channel_name = info.get('display_name') or info.get('name')
entries = []
offset = 0
limit = self._PAGE_LIMIT
for counter in itertools.count(1):
response = self._download_json(
self._PLAYLIST_URL % (channel_id, offset, limit),
channel_id, 'Downloading %s videos JSON page %d' % (self._PLAYLIST_TYPE, counter))
page_entries = self._extract_playlist_page(response)
if not page_entries:
break
entries.extend(page_entries)
offset += limit
return self.playlist_result(
[self.url_result(entry) for entry in set(entries)],
channel_id, channel_name)
def _extract_playlist_page(self, response):
videos = response.get('videos')
return [video['url'] for video in videos] if videos else []
def _real_extract(self, url):
return self._extract_playlist(self._match_id(url))
class TwitchProfileIE(TwitchPlaylistBaseIE):
IE_NAME = 'twitch:profile'
_VALID_URL = r'%s/(?P<id>[^/]+)/profile/?(?:\#.*)?$' % TwitchBaseIE._VALID_URL_BASE
_PLAYLIST_TYPE = 'profile'
_TEST = {
'url': 'http://www.twitch.tv/vanillatv/profile',
'info_dict': {
'id': 'vanillatv',
'title': 'VanillaTV',
},
'playlist_mincount': 412,
}
class TwitchPastBroadcastsIE(TwitchPlaylistBaseIE):
IE_NAME = 'twitch:past_broadcasts'
_VALID_URL = r'%s/(?P<id>[^/]+)/profile/past_broadcasts/?(?:\#.*)?$' % TwitchBaseIE._VALID_URL_BASE
_PLAYLIST_URL = TwitchPlaylistBaseIE._PLAYLIST_URL + '&broadcasts=true'
_PLAYLIST_TYPE = 'past broadcasts'
_TEST = {
'url': 'http://www.twitch.tv/spamfish/profile/past_broadcasts',
'info_dict': {
'id': 'spamfish',
'title': 'Spamfish',
},
'playlist_mincount': 54,
}
class TwitchBookmarksIE(TwitchPlaylistBaseIE):
IE_NAME = 'twitch:bookmarks'
_VALID_URL = r'%s/(?P<id>[^/]+)/profile/bookmarks/?(?:\#.*)?$' % TwitchBaseIE._VALID_URL_BASE
_PLAYLIST_URL = '%s/api/bookmark/?user=%%s&offset=%%d&limit=%%d' % TwitchBaseIE._API_BASE
_PLAYLIST_TYPE = 'bookmarks'
_TEST = {
'url': 'http://www.twitch.tv/ognos/profile/bookmarks',
'info_dict': {
'id': 'ognos',
'title': 'Ognos',
},
'playlist_mincount': 3,
}
def _extract_playlist_page(self, response):
entries = []
for bookmark in response.get('bookmarks', []):
video = bookmark.get('video')
if not video:
continue
entries.append(video['url'])
return entries
class TwitchStreamIE(TwitchBaseIE):
IE_NAME = 'twitch:stream'
_VALID_URL = r'%s/(?P<id>[^/#?]+)/?(?:\#.*)?$' % TwitchBaseIE._VALID_URL_BASE
_TESTS = [{
'url': 'http://www.twitch.tv/shroomztv',
'info_dict': {
'id': '12772022048',
'display_id': 'shroomztv',
'ext': 'mp4',
'title': 're:^ShroomzTV [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$',
'description': 'H1Z1 - lonewolfing with ShroomzTV | A3 Battle Royale later - @ShroomzTV',
'is_live': True,
'timestamp': 1421928037,
'upload_date': '20150122',
'uploader': 'ShroomzTV',
'uploader_id': 'shroomztv',
'view_count': int,
},
'params': {
# m3u8 download
'skip_download': True,
},
}, {
'url': 'http://www.twitch.tv/miracle_doto#profile-0',
'only_matching': True,
}]
def _real_extract(self, url):
channel_id = self._match_id(url)
stream = self._download_json(
'%s/kraken/streams/%s' % (self._API_BASE, channel_id), channel_id,
'Downloading stream JSON').get('stream')
# Fallback on profile extraction if stream is offline
if not stream:
return self.url_result(
'http://www.twitch.tv/%s/profile' % channel_id,
'TwitchProfile', channel_id)
# Channel name may be typed if different case than the original channel name
# (e.g. http://www.twitch.tv/TWITCHPLAYSPOKEMON) that will lead to constructing
# an invalid m3u8 URL. Working around by use of original channel name from stream
# JSON and fallback to lowercase if it's not available.
channel_id = stream.get('channel', {}).get('name') or channel_id.lower()
access_token = self._download_json(
'%s/api/channels/%s/access_token' % (self._API_BASE, channel_id), channel_id,
'Downloading channel access token')
query = {
'allow_source': 'true',
'p': random.randint(1000000, 10000000),
'player': 'twitchweb',
'segment_preference': '4',
'sig': access_token['sig'].encode('utf-8'),
'token': access_token['token'].encode('utf-8'),
}
formats = self._extract_m3u8_formats(
'%s/api/channel/hls/%s.m3u8?%s'
% (self._USHER_BASE, channel_id, compat_urllib_parse.urlencode(query)),
channel_id, 'mp4')
self._prefer_source(formats)
view_count = stream.get('viewers')
timestamp = parse_iso8601(stream.get('created_at'))
channel = stream['channel']
title = self._live_title(channel.get('display_name') or channel.get('name'))
description = channel.get('status')
thumbnails = []
for thumbnail_key, thumbnail_url in stream['preview'].items():
m = re.search(r'(?P<width>\d+)x(?P<height>\d+)\.jpg$', thumbnail_key)
if not m:
continue
thumbnails.append({
'url': thumbnail_url,
'width': int(m.group('width')),
'height': int(m.group('height')),
})
return {
'id': compat_str(stream['_id']),
'display_id': channel_id,
'title': title,
'description': description,
'thumbnails': thumbnails,
'uploader': channel.get('display_name'),
'uploader_id': channel.get('name'),
'timestamp': timestamp,
'view_count': view_count,
'formats': formats,
'is_live': True,
}