youtube-dl/youtube_dl/extractor/zype.py
Remita Amine a9866c0366 [zype] improve extraction
- extract subtitles(closes #21258)
- support URLs with alternative keys/tokens(#21258)
- extract more metadata
2020-01-15 14:18:04 +01:00

135 lines
5.1 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..compat import compat_HTTPError
from ..utils import (
dict_get,
ExtractorError,
int_or_none,
js_to_json,
parse_iso8601,
)
class ZypeIE(InfoExtractor):
_ID_RE = r'[\da-fA-F]+'
_COMMON_RE = r'//player\.zype\.com/embed/%s\.(?:js|json|html)\?.*?(?:access_token|(?:ap[ip]|player)_key)='
_VALID_URL = r'https?:%s[^&]+' % (_COMMON_RE % ('(?P<id>%s)' % _ID_RE))
_TEST = {
'url': 'https://player.zype.com/embed/5b400b834b32992a310622b9.js?api_key=jZ9GUhRmxcPvX7M3SlfejB6Hle9jyHTdk2jVxG7wOHPLODgncEKVdPYBhuz9iWXQ&autoplay=false&controls=true&da=false',
'md5': 'eaee31d474c76a955bdaba02a505c595',
'info_dict': {
'id': '5b400b834b32992a310622b9',
'ext': 'mp4',
'title': 'Smoky Barbecue Favorites',
'thumbnail': r're:^https?://.*\.jpe?g',
'description': 'md5:5ff01e76316bd8d46508af26dc86023b',
'timestamp': 1504915200,
'upload_date': '20170909',
},
}
@staticmethod
def _extract_urls(webpage):
return [
mobj.group('url')
for mobj in re.finditer(
r'<script[^>]+\bsrc=(["\'])(?P<url>(?:https?:)?%s.+?)\1' % (ZypeIE._COMMON_RE % ZypeIE._ID_RE),
webpage)]
def _real_extract(self, url):
video_id = self._match_id(url)
try:
response = self._download_json(re.sub(
r'\.(?:js|html)\?', '.json?', url), video_id)['response']
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code in (400, 401, 403):
raise ExtractorError(self._parse_json(
e.cause.read().decode(), video_id)['message'], expected=True)
raise
body = response['body']
video = response['video']
title = video['title']
if isinstance(body, dict):
formats = []
for output in body.get('outputs', []):
output_url = output.get('url')
if not output_url:
continue
name = output.get('name')
if name == 'm3u8':
formats = self._extract_m3u8_formats(
output_url, video_id, 'mp4',
'm3u8_native', m3u8_id='hls', fatal=False)
else:
f = {
'format_id': name,
'tbr': int_or_none(output.get('bitrate')),
'url': output_url,
}
if name in ('m4a', 'mp3'):
f['vcodec'] = 'none'
else:
f.update({
'height': int_or_none(output.get('height')),
'width': int_or_none(output.get('width')),
})
formats.append(f)
text_tracks = body.get('subtitles') or []
else:
m3u8_url = self._search_regex(
r'(["\'])(?P<url>(?:(?!\1).)+\.m3u8(?:(?!\1).)*)\1',
body, 'm3u8 url', group='url')
formats = self._extract_m3u8_formats(
m3u8_url, video_id, 'mp4', 'm3u8_native', m3u8_id='hls')
text_tracks = self._search_regex(
r'textTracks\s*:\s*(\[[^]]+\])',
body, 'text tracks', default=None)
if text_tracks:
text_tracks = self._parse_json(
text_tracks, video_id, js_to_json, False)
self._sort_formats(formats)
subtitles = {}
if text_tracks:
for text_track in text_tracks:
tt_url = dict_get(text_track, ('file', 'src'))
if not tt_url:
continue
subtitles.setdefault(text_track.get('label') or 'English', []).append({
'url': tt_url,
})
thumbnails = []
for thumbnail in video.get('thumbnails', []):
thumbnail_url = thumbnail.get('url')
if not thumbnail_url:
continue
thumbnails.append({
'url': thumbnail_url,
'width': int_or_none(thumbnail.get('width')),
'height': int_or_none(thumbnail.get('height')),
})
return {
'id': video_id,
'display_id': video.get('friendly_title'),
'title': title,
'thumbnails': thumbnails,
'description': dict_get(video, ('description', 'ott_description', 'short_description')),
'timestamp': parse_iso8601(video.get('published_at')),
'duration': int_or_none(video.get('duration')),
'view_count': int_or_none(video.get('request_count')),
'average_rating': int_or_none(video.get('rating')),
'season_number': int_or_none(video.get('season')),
'episode_number': int_or_none(video.get('episode')),
'formats': formats,
'subtitles': subtitles,
}