mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-24 03:41:00 +01:00
[escapist] Fix extractor (fixes #5090)
This commit is contained in:
parent
c86b61428b
commit
e2dc351d25
|
@ -1,128 +1,102 @@
|
||||||
from __future__ import unicode_literals
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
import json
|
||||||
|
|
||||||
from .common import InfoExtractor
|
from .common import InfoExtractor
|
||||||
from ..compat import (
|
from ..compat import compat_urllib_request
|
||||||
compat_urllib_parse,
|
|
||||||
compat_urllib_request,
|
|
||||||
)
|
|
||||||
from ..utils import (
|
from ..utils import (
|
||||||
ExtractorError,
|
determine_ext,
|
||||||
js_to_json,
|
clean_html,
|
||||||
parse_duration,
|
qualities,
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
|
def _decrypt_config(key, string):
|
||||||
|
a = ''
|
||||||
|
i = ''
|
||||||
|
r = ''
|
||||||
|
|
||||||
|
while len(a) < (len(string) / 2):
|
||||||
|
a += key
|
||||||
|
|
||||||
|
a = a[0:int(len(string) / 2)]
|
||||||
|
|
||||||
|
t = 0
|
||||||
|
while t < len(string):
|
||||||
|
i += chr(int(string[t] + string[t + 1], 16))
|
||||||
|
t += 2
|
||||||
|
|
||||||
|
icko = [s for s in i]
|
||||||
|
|
||||||
|
for t, c in enumerate(a):
|
||||||
|
r += chr(ord(c) ^ ord(icko[t]))
|
||||||
|
|
||||||
|
return r
|
||||||
|
|
||||||
|
|
||||||
class EscapistIE(InfoExtractor):
|
class EscapistIE(InfoExtractor):
|
||||||
_VALID_URL = r'https?://?(www\.)?escapistmagazine\.com/videos/view/[^/?#]+/(?P<id>[0-9]+)-[^/?#]*(?:$|[?#])'
|
_VALID_URL = r'https?://?(www\.)?escapistmagazine\.com/videos/view/[^/?#]+/(?P<id>[0-9]+)-[^/?#]*(?:$|[?#])'
|
||||||
_USER_AGENT = 'Mozilla/5.0 (Windows NT 6.1; WOW64; Trident/7.0; rv:11.0) like Gecko'
|
_TESTS = [{
|
||||||
_TEST = {
|
|
||||||
'url': 'http://www.escapistmagazine.com/videos/view/the-escapist-presents/6618-Breaking-Down-Baldurs-Gate',
|
'url': 'http://www.escapistmagazine.com/videos/view/the-escapist-presents/6618-Breaking-Down-Baldurs-Gate',
|
||||||
'md5': 'ab3a706c681efca53f0a35f1415cf0d1',
|
'md5': 'c6793dbda81388f4264c1ba18684a74d',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '6618',
|
'id': '6618',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'description': "Baldur's Gate: Original, Modded or Enhanced Edition? I'll break down what you can expect from the new Baldur's Gate: Enhanced Edition.",
|
'description': "Baldur's Gate: Original, Modded or Enhanced Edition? I'll break down what you can expect from the new Baldur's Gate: Enhanced Edition.",
|
||||||
'uploader_id': 'the-escapist-presents',
|
|
||||||
'uploader': 'The Escapist Presents',
|
|
||||||
'title': "Breaking Down Baldur's Gate",
|
'title': "Breaking Down Baldur's Gate",
|
||||||
'thumbnail': 're:^https?://.*\.jpg$',
|
'thumbnail': 're:^https?://.*\.jpg$',
|
||||||
'duration': 264,
|
'duration': 264,
|
||||||
}
|
}
|
||||||
}
|
}, {
|
||||||
|
'url': 'http://www.escapistmagazine.com/videos/view/zero-punctuation/10044-Evolve-One-vs-Multiplayer',
|
||||||
|
'md5': 'cf8842a8a46444d241f9a9980d7874f2',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '10044',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'description': 'This week, Zero Punctuation reviews Evolve.',
|
||||||
|
'title': 'Evolve - One vs Multiplayer',
|
||||||
|
'thumbnail': 're:^https?://.*\.jpg$',
|
||||||
|
'duration': 304,
|
||||||
|
}
|
||||||
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
webpage_req = compat_urllib_request.Request(url)
|
webpage = self._download_webpage(url, video_id)
|
||||||
webpage_req.add_header('User-Agent', self._USER_AGENT)
|
|
||||||
webpage = self._download_webpage(webpage_req, video_id)
|
|
||||||
|
|
||||||
uploader_id = self._html_search_regex(
|
imsVideo = json.loads(self._search_regex(r'imsVideo\.play\(([^\)]+)\);', webpage, 'imsVideo'))
|
||||||
r"<h1\s+class='headline'>\s*<a\s+href='/videos/view/(.*?)'",
|
video_id = imsVideo['videoID']
|
||||||
webpage, 'uploader ID', fatal=False)
|
key = imsVideo['hash']
|
||||||
uploader = self._html_search_regex(
|
|
||||||
r"<h1\s+class='headline'>(.*?)</a>",
|
|
||||||
webpage, 'uploader', fatal=False)
|
|
||||||
description = self._html_search_meta('description', webpage)
|
|
||||||
duration = parse_duration(self._html_search_meta('duration', webpage))
|
|
||||||
|
|
||||||
raw_title = self._html_search_meta('title', webpage, fatal=True)
|
quality = qualities(['lq', 'hq', 'hd'])
|
||||||
title = raw_title.partition(' : ')[2]
|
|
||||||
|
|
||||||
config_url = compat_urllib_parse.unquote(self._html_search_regex(
|
|
||||||
r'''(?x)
|
|
||||||
(?:
|
|
||||||
<param\s+name="flashvars".*?\s+value="config=|
|
|
||||||
flashvars="config=
|
|
||||||
)
|
|
||||||
(https?://[^"&]+)
|
|
||||||
''',
|
|
||||||
webpage, 'config URL'))
|
|
||||||
|
|
||||||
formats = []
|
formats = []
|
||||||
ad_formats = []
|
for q in ['lq', 'hq', 'hd']:
|
||||||
|
config_req = compat_urllib_request.Request('http://www.escapistmagazine.com/videos/'
|
||||||
|
'vidconfig.php?videoID=%s&hash=%s&quality=%s' % (video_id, key, 'mp4_' + q))
|
||||||
|
config_req.add_header('Referer', url)
|
||||||
|
config = self._download_webpage(config_req, video_id, 'Downloading video config ' + q.upper())
|
||||||
|
|
||||||
def _add_format(name, cfg_url, quality):
|
data = json.loads(_decrypt_config(key, config))
|
||||||
cfg_req = compat_urllib_request.Request(cfg_url)
|
|
||||||
cfg_req.add_header('User-Agent', self._USER_AGENT)
|
|
||||||
config = self._download_json(
|
|
||||||
cfg_req, video_id,
|
|
||||||
'Downloading ' + name + ' configuration',
|
|
||||||
'Unable to download ' + name + ' configuration',
|
|
||||||
transform_source=js_to_json)
|
|
||||||
|
|
||||||
playlist = config['playlist']
|
title = clean_html(data['videoData']['title'])
|
||||||
for p in playlist:
|
duration = data['videoData']['duration'] / 1000
|
||||||
if p.get('eventCategory') == 'Video':
|
|
||||||
ar = formats
|
|
||||||
elif p.get('eventCategory') == 'Video Postroll':
|
|
||||||
ar = ad_formats
|
|
||||||
else:
|
|
||||||
continue
|
|
||||||
|
|
||||||
ar.append({
|
for i, v in enumerate(data['files']['videos']):
|
||||||
'url': p['url'],
|
|
||||||
'format_id': name,
|
|
||||||
'quality': quality,
|
|
||||||
'http_headers': {
|
|
||||||
'User-Agent': self._USER_AGENT,
|
|
||||||
},
|
|
||||||
})
|
|
||||||
|
|
||||||
_add_format('normal', config_url, quality=0)
|
formats.append({
|
||||||
hq_url = (config_url +
|
'url': v,
|
||||||
('&hq=1' if '?' in config_url else config_url + '?hq=1'))
|
'format_id': determine_ext(v) + '_' + q + str(i),
|
||||||
try:
|
'quality': quality(q),
|
||||||
_add_format('hq', hq_url, quality=1)
|
})
|
||||||
except ExtractorError:
|
|
||||||
pass # That's fine, we'll just use normal quality
|
|
||||||
self._sort_formats(formats)
|
|
||||||
|
|
||||||
if '/escapist/sales-marketing/' in formats[-1]['url']:
|
|
||||||
raise ExtractorError('This IP address has been blocked by The Escapist', expected=True)
|
|
||||||
|
|
||||||
res = {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'formats': formats,
|
'formats': formats,
|
||||||
'uploader': uploader,
|
|
||||||
'uploader_id': uploader_id,
|
|
||||||
'title': title,
|
'title': title,
|
||||||
'thumbnail': self._og_search_thumbnail(webpage),
|
'thumbnail': self._og_search_thumbnail(webpage),
|
||||||
'description': description,
|
'description': self._og_search_description(webpage),
|
||||||
'duration': duration,
|
'duration': duration,
|
||||||
}
|
}
|
||||||
|
|
||||||
if self._downloader.params.get('include_ads') and ad_formats:
|
|
||||||
self._sort_formats(ad_formats)
|
|
||||||
ad_res = {
|
|
||||||
'id': '%s-ad' % video_id,
|
|
||||||
'title': '%s (Postroll)' % title,
|
|
||||||
'formats': ad_formats,
|
|
||||||
}
|
|
||||||
return {
|
|
||||||
'_type': 'playlist',
|
|
||||||
'entries': [res, ad_res],
|
|
||||||
'title': title,
|
|
||||||
'id': video_id,
|
|
||||||
}
|
|
||||||
|
|
||||||
return res
|
|
||||||
|
|
Loading…
Reference in New Issue
Block a user