[fourtube] Fix extraction
This commit is contained in:
parent
4f4f642822
commit
9d22a7dfb0
|
@ -7,10 +7,9 @@ from ..compat import (
|
||||||
compat_urllib_request,
|
compat_urllib_request,
|
||||||
)
|
)
|
||||||
from ..utils import (
|
from ..utils import (
|
||||||
clean_html,
|
|
||||||
parse_duration,
|
parse_duration,
|
||||||
|
parse_iso8601,
|
||||||
str_to_int,
|
str_to_int,
|
||||||
unified_strdate,
|
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
|
@ -28,68 +27,81 @@ class FourTubeIE(InfoExtractor):
|
||||||
'uploader': 'WCP Club',
|
'uploader': 'WCP Club',
|
||||||
'uploader_id': 'wcp-club',
|
'uploader_id': 'wcp-club',
|
||||||
'upload_date': '20131031',
|
'upload_date': '20131031',
|
||||||
|
'timestamp': 1383263892,
|
||||||
'duration': 583,
|
'duration': 583,
|
||||||
|
'view_count': int,
|
||||||
|
'like_count': int,
|
||||||
|
'categories': list,
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
webpage_url = 'http://www.4tube.com/videos/' + video_id
|
webpage = self._download_webpage(url, video_id)
|
||||||
webpage = self._download_webpage(webpage_url, video_id)
|
|
||||||
|
|
||||||
self.report_extraction(video_id)
|
title = self._html_search_meta('name', webpage)
|
||||||
|
timestamp = parse_iso8601(self._html_search_meta(
|
||||||
|
'uploadDate', webpage))
|
||||||
|
thumbnail = self._html_search_meta('thumbnailUrl', webpage)
|
||||||
|
uploader_id = self._html_search_regex(
|
||||||
|
r'<a class="img-avatar" href="[^"]+/channels/([^/"]+)" title="Go to [^"]+ page">',
|
||||||
|
webpage, 'uploader id')
|
||||||
|
uploader = self._html_search_regex(
|
||||||
|
r'<a class="img-avatar" href="[^"]+/channels/[^/"]+" title="Go to ([^"]+) page">',
|
||||||
|
webpage, 'uploader')
|
||||||
|
|
||||||
playlist_json = self._html_search_regex(r'var playerConfigPlaylist\s+=\s+([^;]+)', webpage, 'Playlist')
|
categories_html = self._search_regex(
|
||||||
media_id = self._search_regex(r'idMedia:\s*(\d+)', playlist_json, 'Media Id')
|
r'(?s)><i class="icon icon-tag"></i>\s*Categories / Tags\s*.*?<ul class="list">(.*?)</ul>',
|
||||||
sources = self._search_regex(r'sources:\s*\[([^\]]*)\]', playlist_json, 'Sources').split(',')
|
webpage, 'categories', fatal=False)
|
||||||
title = self._search_regex(r'title:\s*"([^"]*)', playlist_json, 'Title')
|
categories = None
|
||||||
thumbnail_url = self._search_regex(r'image:\s*"([^"]*)', playlist_json, 'Thumbnail', fatal=False)
|
if categories_html:
|
||||||
|
categories = [
|
||||||
|
c.strip() for c in re.findall(
|
||||||
|
r'(?s)<li><a.*?>(.*?)</a>', categories_html)]
|
||||||
|
|
||||||
uploader_str = self._search_regex(r'<span>Uploaded by</span>(.*?)<span>', webpage, 'uploader', fatal=False)
|
view_count = str_to_int(self._search_regex(
|
||||||
mobj = re.search(r'<a href="/sites/(?P<id>[^"]+)"><strong>(?P<name>[^<]+)</strong></a>', uploader_str)
|
r'<meta itemprop="interactionCount" content="UserPlays:([0-9,]+)">',
|
||||||
(uploader, uploader_id) = (mobj.group('name'), mobj.group('id')) if mobj else (clean_html(uploader_str), None)
|
webpage, 'view count', fatal=False))
|
||||||
|
like_count = str_to_int(self._search_regex(
|
||||||
|
r'<meta itemprop="interactionCount" content="UserLikes:([0-9,]+)">',
|
||||||
|
webpage, 'like count', fatal=False))
|
||||||
|
duration = parse_duration(self._html_search_meta('duration', webpage))
|
||||||
|
|
||||||
upload_date = None
|
params_js = self._search_regex(
|
||||||
view_count = None
|
r'\$\.ajax\(url,\ opts\);\s*\}\s*\}\)\(([0-9,\[\] ]+)\)',
|
||||||
duration = None
|
webpage, 'initialization parameters'
|
||||||
description = self._html_search_meta('description', webpage, 'description')
|
)
|
||||||
if description:
|
params = self._parse_json('[%s]' % params_js, video_id)
|
||||||
upload_date = self._search_regex(r'Published Date: (\d{2} [a-zA-Z]{3} \d{4})', description, 'upload date',
|
media_id = params[0]
|
||||||
fatal=False)
|
sources = ['%s' % p for p in params[2]]
|
||||||
if upload_date:
|
|
||||||
upload_date = unified_strdate(upload_date)
|
|
||||||
view_count = self._search_regex(r'Views: ([\d,\.]+)', description, 'view count', fatal=False)
|
|
||||||
if view_count:
|
|
||||||
view_count = str_to_int(view_count)
|
|
||||||
duration = parse_duration(self._search_regex(r'Length: (\d+m\d+s)', description, 'duration', fatal=False))
|
|
||||||
|
|
||||||
token_url = "http://tkn.4tube.com/{0}/desktop/{1}".format(media_id, "+".join(sources))
|
token_url = 'http://tkn.4tube.com/{0}/desktop/{1}'.format(
|
||||||
|
media_id, '+'.join(sources))
|
||||||
headers = {
|
headers = {
|
||||||
b'Content-Type': b'application/x-www-form-urlencoded',
|
b'Content-Type': b'application/x-www-form-urlencoded',
|
||||||
b'Origin': b'http://www.4tube.com',
|
b'Origin': b'http://www.4tube.com',
|
||||||
}
|
}
|
||||||
token_req = compat_urllib_request.Request(token_url, b'{}', headers)
|
token_req = compat_urllib_request.Request(token_url, b'{}', headers)
|
||||||
tokens = self._download_json(token_req, video_id)
|
tokens = self._download_json(token_req, video_id)
|
||||||
|
|
||||||
formats = [{
|
formats = [{
|
||||||
'url': tokens[format]['token'],
|
'url': tokens[format]['token'],
|
||||||
'format_id': format + 'p',
|
'format_id': format + 'p',
|
||||||
'resolution': format + 'p',
|
'resolution': format + 'p',
|
||||||
'quality': int(format),
|
'quality': int(format),
|
||||||
} for format in sources]
|
} for format in sources]
|
||||||
|
|
||||||
self._sort_formats(formats)
|
self._sort_formats(formats)
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'title': title,
|
'title': title,
|
||||||
'formats': formats,
|
'formats': formats,
|
||||||
'thumbnail': thumbnail_url,
|
'categories': categories,
|
||||||
|
'thumbnail': thumbnail,
|
||||||
'uploader': uploader,
|
'uploader': uploader,
|
||||||
'uploader_id': uploader_id,
|
'uploader_id': uploader_id,
|
||||||
'upload_date': upload_date,
|
'timestamp': timestamp,
|
||||||
|
'like_count': like_count,
|
||||||
'view_count': view_count,
|
'view_count': view_count,
|
||||||
'duration': duration,
|
'duration': duration,
|
||||||
'age_limit': 18,
|
'age_limit': 18,
|
||||||
'webpage_url': webpage_url,
|
|
||||||
}
|
}
|
||||||
|
|
|
@ -1277,7 +1277,7 @@ def parse_duration(s):
|
||||||
s = s.strip()
|
s = s.strip()
|
||||||
|
|
||||||
m = re.match(
|
m = re.match(
|
||||||
r'''(?ix)T?
|
r'''(?ix)(?:P?T)?
|
||||||
(?:
|
(?:
|
||||||
(?P<only_mins>[0-9.]+)\s*(?:mins?|minutes?)\s*|
|
(?P<only_mins>[0-9.]+)\s*(?:mins?|minutes?)\s*|
|
||||||
(?P<only_hours>[0-9.]+)\s*(?:hours?)|
|
(?P<only_hours>[0-9.]+)\s*(?:hours?)|
|
||||||
|
|
Loading…
Reference in New Issue