1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-05 09:54:32 +01:00
youtube-dl/youtube_dl/extractor/ivi.py
2015-01-28 23:58:14 +06:00

183 lines
7.1 KiB
Python

# encoding: utf-8
from __future__ import unicode_literals
import re
import json
from .common import InfoExtractor
from ..compat import (
compat_urllib_request,
)
from ..utils import (
ExtractorError,
)
class IviIE(InfoExtractor):
IE_DESC = 'ivi.ru'
IE_NAME = 'ivi'
_VALID_URL = r'https?://(?:www\.)?ivi\.ru/(?:watch/(?:[^/]+/)?|video/player\?.*?videoId=)(?P<id>\d+)'
_TESTS = [
# Single movie
{
'url': 'http://www.ivi.ru/watch/53141',
'md5': '6ff5be2254e796ed346251d117196cf4',
'info_dict': {
'id': '53141',
'ext': 'mp4',
'title': 'Иван Васильевич меняет профессию',
'description': 'md5:b924063ea1677c8fe343d8a72ac2195f',
'duration': 5498,
'thumbnail': 'http://thumbs.ivi.ru/f20.vcp.digitalaccess.ru/contents/d/1/c3c885163a082c29bceeb7b5a267a6.jpg',
},
'skip': 'Only works from Russia',
},
# Serial's serie
{
'url': 'http://www.ivi.ru/watch/dvoe_iz_lartsa/9549',
'md5': '221f56b35e3ed815fde2df71032f4b3e',
'info_dict': {
'id': '9549',
'ext': 'mp4',
'title': 'Двое из ларца - Серия 1',
'duration': 2655,
'thumbnail': 'http://thumbs.ivi.ru/f15.vcp.digitalaccess.ru/contents/8/4/0068dc0677041f3336b7c2baad8fc0.jpg',
},
'skip': 'Only works from Russia',
}
]
# Sorted by quality
_known_formats = ['MP4-low-mobile', 'MP4-mobile', 'FLV-lo', 'MP4-lo', 'FLV-hi', 'MP4-hi', 'MP4-SHQ']
# Sorted by size
_known_thumbnails = ['Thumb-120x90', 'Thumb-160', 'Thumb-640x480']
def _extract_description(self, html):
m = re.search(r'<meta name="description" content="(?P<description>[^"]+)"/>', html)
return m.group('description') if m is not None else None
def _extract_comment_count(self, html):
m = re.search('(?s)<a href="#" id="view-comments" class="action-button dim gradient">\s*Комментарии:\s*(?P<commentcount>\d+)\s*</a>', html)
return int(m.group('commentcount')) if m is not None else 0
def _real_extract(self, url):
video_id = self._match_id(url)
api_url = 'http://api.digitalaccess.ru/api/json/'
data = {
'method': 'da.content.get',
'params': [
video_id, {
'site': 's183',
'referrer': 'http://www.ivi.ru/watch/%s' % video_id,
'contentid': video_id
}
]
}
request = compat_urllib_request.Request(api_url, json.dumps(data))
video_json_page = self._download_webpage(
request, video_id, 'Downloading video JSON')
video_json = json.loads(video_json_page)
if 'error' in video_json:
error = video_json['error']
if error['origin'] == 'NoRedisValidData':
raise ExtractorError('Video %s does not exist' % video_id, expected=True)
raise ExtractorError(
'Unable to download video %s: %s' % (video_id, error['message']),
expected=True)
result = video_json['result']
formats = [{
'url': x['url'],
'format_id': x['content_format'],
'preference': self._known_formats.index(x['content_format']),
} for x in result['files'] if x['content_format'] in self._known_formats]
self._sort_formats(formats)
if not formats:
raise ExtractorError('No media links available for %s' % video_id)
duration = result['duration']
compilation = result['compilation']
title = result['title']
title = '%s - %s' % (compilation, title) if compilation is not None else title
previews = result['preview']
previews.sort(key=lambda fmt: self._known_thumbnails.index(fmt['content_format']))
thumbnail = previews[-1]['url'] if len(previews) > 0 else None
video_page = self._download_webpage(url, video_id, 'Downloading video page')
description = self._extract_description(video_page)
comment_count = self._extract_comment_count(video_page)
return {
'id': video_id,
'title': title,
'thumbnail': thumbnail,
'description': description,
'duration': duration,
'comment_count': comment_count,
'formats': formats,
}
class IviCompilationIE(InfoExtractor):
IE_DESC = 'ivi.ru compilations'
IE_NAME = 'ivi:compilation'
_VALID_URL = r'https?://(?:www\.)?ivi\.ru/watch/(?!\d+)(?P<compilationid>[a-z\d_-]+)(?:/season(?P<seasonid>\d+))?$'
_TESTS = [{
'url': 'http://www.ivi.ru/watch/dvoe_iz_lartsa',
'info_dict': {
'id': 'dvoe_iz_lartsa',
'title': 'Двое из ларца (2006 - 2008)',
},
'playlist_mincount': 24,
}, {
'url': 'http://www.ivi.ru/watch/dvoe_iz_lartsa/season1',
'info_dict': {
'id': 'dvoe_iz_lartsa/season1',
'title': 'Двое из ларца (2006 - 2008) 1 сезон',
},
'playlist_mincount': 12,
}]
def _extract_entries(self, html, compilation_id):
return [self.url_result('http://www.ivi.ru/watch/%s/%s' % (compilation_id, serie), 'Ivi')
for serie in re.findall(r'<strong><a href="/watch/%s/(\d+)">(?:[^<]+)</a></strong>' % compilation_id, html)]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
compilation_id = mobj.group('compilationid')
season_id = mobj.group('seasonid')
if season_id is not None: # Season link
season_page = self._download_webpage(url, compilation_id, 'Downloading season %s web page' % season_id)
playlist_id = '%s/season%s' % (compilation_id, season_id)
playlist_title = self._html_search_meta('title', season_page, 'title')
entries = self._extract_entries(season_page, compilation_id)
else: # Compilation link
compilation_page = self._download_webpage(url, compilation_id, 'Downloading compilation web page')
playlist_id = compilation_id
playlist_title = self._html_search_meta('title', compilation_page, 'title')
seasons = re.findall(r'<a href="/watch/%s/season(\d+)">[^<]+</a>' % compilation_id, compilation_page)
if len(seasons) == 0: # No seasons in this compilation
entries = self._extract_entries(compilation_page, compilation_id)
else:
entries = []
for season_id in seasons:
season_page = self._download_webpage(
'http://www.ivi.ru/watch/%s/season%s' % (compilation_id, season_id),
compilation_id, 'Downloading season %s web page' % season_id)
entries.extend(self._extract_entries(season_page, compilation_id))
return self.playlist_result(entries, playlist_id, playlist_title)