mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-01-07 17:16:08 +00:00
[malltv] fix extraction(closes #27035)
This commit is contained in:
parent
cb2b9a22a5
commit
aa613ef7e1
|
@ -1,10 +1,16 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import merge_dicts
|
||||
from ..utils import (
|
||||
clean_html,
|
||||
dict_get,
|
||||
float_or_none,
|
||||
int_or_none,
|
||||
merge_dicts,
|
||||
parse_duration,
|
||||
try_get,
|
||||
)
|
||||
|
||||
|
||||
class MallTVIE(InfoExtractor):
|
||||
|
@ -17,7 +23,7 @@ class MallTVIE(InfoExtractor):
|
|||
'display_id': '18-miliard-pro-neziskovky-opravdu-jsou-sportovci-nebo-clovek-v-tisni-pijavice',
|
||||
'ext': 'mp4',
|
||||
'title': '18 miliard pro neziskovky. Opravdu jsou sportovci nebo Člověk v tísni pijavice?',
|
||||
'description': 'md5:25fc0ec42a72ba602b602c683fa29deb',
|
||||
'description': 'md5:db7d5744a4bd4043d9d98324aa72ab35',
|
||||
'duration': 216,
|
||||
'timestamp': 1538870400,
|
||||
'upload_date': '20181007',
|
||||
|
@ -37,20 +43,46 @@ class MallTVIE(InfoExtractor):
|
|||
webpage = self._download_webpage(
|
||||
url, display_id, headers=self.geo_verification_headers())
|
||||
|
||||
SOURCE_RE = r'(<source[^>]+\bsrc=(?:(["\'])(?:(?!\2).)+|[^\s]+)/(?P<id>[\da-z]+)/index)\b'
|
||||
video = self._parse_json(self._search_regex(
|
||||
r'videoObject\s*=\s*JSON\.parse\(JSON\.stringify\(({.+?})\)\);',
|
||||
webpage, 'video object'), display_id)
|
||||
video_source = video['VideoSource']
|
||||
video_id = self._search_regex(
|
||||
SOURCE_RE, webpage, 'video id', group='id')
|
||||
r'/([\da-z]+)/index\b', video_source, 'video id')
|
||||
|
||||
media = self._parse_html5_media_entries(
|
||||
url, re.sub(SOURCE_RE, r'\1.m3u8', webpage), video_id,
|
||||
m3u8_id='hls', m3u8_entry_protocol='m3u8_native')[0]
|
||||
formats = self._extract_m3u8_formats(
|
||||
video_source + '.m3u8', video_id, 'mp4', 'm3u8_native')
|
||||
self._sort_formats(formats)
|
||||
|
||||
subtitles = {}
|
||||
for s in (video.get('Subtitles') or {}):
|
||||
s_url = s.get('Url')
|
||||
if not s_url:
|
||||
continue
|
||||
subtitles.setdefault(s.get('Language') or 'cz', []).append({
|
||||
'url': s_url,
|
||||
})
|
||||
|
||||
entity_counts = video.get('EntityCounts') or {}
|
||||
|
||||
def get_count(k):
|
||||
v = entity_counts.get(k + 's') or {}
|
||||
return int_or_none(dict_get(v, ('Count', 'StrCount')))
|
||||
|
||||
info = self._search_json_ld(webpage, video_id, default={})
|
||||
|
||||
return merge_dicts(media, info, {
|
||||
return merge_dicts({
|
||||
'id': video_id,
|
||||
'display_id': display_id,
|
||||
'title': self._og_search_title(webpage, default=None) or display_id,
|
||||
'description': self._og_search_description(webpage, default=None),
|
||||
'thumbnail': self._og_search_thumbnail(webpage, default=None),
|
||||
})
|
||||
'title': video.get('Title'),
|
||||
'description': clean_html(video.get('Description')),
|
||||
'thumbnail': video.get('ThumbnailUrl'),
|
||||
'formats': formats,
|
||||
'subtitles': subtitles,
|
||||
'duration': int_or_none(video.get('DurationSeconds')) or parse_duration(video.get('Duration')),
|
||||
'view_count': get_count('View'),
|
||||
'like_count': get_count('Like'),
|
||||
'dislike_count': get_count('Dislike'),
|
||||
'average_rating': float_or_none(try_get(video, lambda x: x['EntityRating']['AvarageRate'])),
|
||||
'comment_count': get_count('Comment'),
|
||||
}, info)
|
||||
|
|
Loading…
Reference in a new issue