|
|
# -*- coding: utf-8 -*- from __future__ import unicode_literals
import re
from .common import InfoExtractor from ..compat import ( compat_urllib_parse, compat_urllib_parse_unquote, compat_urllib_parse_urlparse, ) from ..utils import ( ExtractorError, float_or_none, sanitized_Request, )
class CeskaTelevizeIE(InfoExtractor): _VALID_URL = r'https?://www\.ceskatelevize\.cz/(porady|ivysilani)/(?:[^/]+/)*(?P<id>[^/#?]+)/*(?:[#?].*)?$' _TESTS = [{ 'url': 'http://www.ceskatelevize.cz/ivysilani/ivysilani/10441294653-hyde-park-civilizace/214411058091220', 'info_dict': { 'id': '61924494876951776', 'ext': 'mp4', 'title': 'Hyde Park Civilizace', 'description': 'md5:fe93f6eda372d150759d11644ebbfb4a', 'thumbnail': 're:^https?://.*\.jpg', 'duration': 3350, }, 'params': { # m3u8 download 'skip_download': True, }, }, { 'url': 'http://www.ceskatelevize.cz/ivysilani/10532695142-prvni-republika/bonus/14716-zpevacka-z-duparny-bobina', 'info_dict': { 'id': '61924494876844374', 'ext': 'mp4', 'title': 'První republika: Zpěvačka z Dupárny Bobina', 'description': 'Sága mapující atmosféru první republiky od r. 1918 do r. 1945.', 'thumbnail': 're:^https?://.*\.jpg', 'duration': 88.4, }, 'params': { # m3u8 download 'skip_download': True, }, }, { # video with 18+ caution trailer 'url': 'http://www.ceskatelevize.cz/porady/10520528904-queer/215562210900007-bogotart/', 'info_dict': { 'id': '215562210900007-bogotart', 'title': 'Queer: Bogotart', 'description': 'Alternativní průvodce současným queer světem', }, 'playlist': [{ 'info_dict': { 'id': '61924494876844842', 'ext': 'mp4', 'title': 'Queer: Bogotart (Varování 18+)', 'duration': 10.2, }, }, { 'info_dict': { 'id': '61924494877068022', 'ext': 'mp4', 'title': 'Queer: Bogotart (Queer)', 'thumbnail': 're:^https?://.*\.jpg', 'duration': 1558.3, }, }], 'params': { # m3u8 download 'skip_download': True, }, }]
def _real_extract(self, url): url = url.replace('/porady/', '/ivysilani/').replace('/video/', '')
mobj = re.match(self._VALID_URL, url) playlist_id = mobj.group('id')
webpage = self._download_webpage(url, playlist_id)
NOT_AVAILABLE_STRING = 'This content is not available at your territory due to limited copyright.' if '%s</p>' % NOT_AVAILABLE_STRING in webpage: raise ExtractorError(NOT_AVAILABLE_STRING, expected=True)
typ = self._html_search_regex( r'getPlaylistUrl\(\[\{"type":"(.+?)","id":".+?"\}\],', webpage, 'type') episode_id = self._html_search_regex( r'getPlaylistUrl\(\[\{"type":".+?","id":"(.+?)"\}\],', webpage, 'episode_id')
data = { 'playlist[0][type]': typ, 'playlist[0][id]': episode_id, 'requestUrl': compat_urllib_parse_urlparse(url).path, 'requestSource': 'iVysilani', }
req = sanitized_Request( 'http://www.ceskatelevize.cz/ivysilani/ajax/get-client-playlist', data=compat_urllib_parse.urlencode(data))
req.add_header('Content-type', 'application/x-www-form-urlencoded') req.add_header('x-addr', '127.0.0.1') req.add_header('X-Requested-With', 'XMLHttpRequest') req.add_header('Referer', url)
playlistpage = self._download_json(req, playlist_id)
playlist_url = playlistpage['url'] if playlist_url == 'error_region': raise ExtractorError(NOT_AVAILABLE_STRING, expected=True)
req = sanitized_Request(compat_urllib_parse_unquote(playlist_url)) req.add_header('Referer', url)
playlist_title = self._og_search_title(webpage) playlist_description = self._og_search_description(webpage)
playlist = self._download_json(req, playlist_id)['playlist'] playlist_len = len(playlist)
entries = [] for item in playlist: formats = [] for format_id, stream_url in item['streamUrls'].items(): formats.extend(self._extract_m3u8_formats( stream_url, playlist_id, 'mp4', entry_protocol='m3u8_native')) self._sort_formats(formats)
item_id = item.get('id') or item['assetId'] title = item['title']
duration = float_or_none(item.get('duration')) thumbnail = item.get('previewImageUrl')
subtitles = {} if item.get('type') == 'VOD': subs = item.get('subtitles') if subs: subtitles = self.extract_subtitles(episode_id, subs)
entries.append({ 'id': item_id, 'title': playlist_title if playlist_len == 1 else '%s (%s)' % (playlist_title, title), 'description': playlist_description if playlist_len == 1 else None, 'thumbnail': thumbnail, 'duration': duration, 'formats': formats, 'subtitles': subtitles, })
return self.playlist_result(entries, playlist_id, playlist_title, playlist_description)
def _get_subtitles(self, episode_id, subs): original_subtitles = self._download_webpage( subs[0]['url'], episode_id, 'Downloading subtitles') srt_subs = self._fix_subtitles(original_subtitles) return { 'cs': [{ 'ext': 'srt', 'data': srt_subs, }] }
@staticmethod def _fix_subtitles(subtitles): """ Convert millisecond-based subtitles to SRT """
def _msectotimecode(msec): """ Helper utility to convert milliseconds to timecode """ components = [] for divider in [1000, 60, 60, 100]: components.append(msec % divider) msec //= divider return "{3:02}:{2:02}:{1:02},{0:03}".format(*components)
def _fix_subtitle(subtitle): for line in subtitle.splitlines(): m = re.match(r"^\s*([0-9]+);\s*([0-9]+)\s+([0-9]+)\s*$", line) if m: yield m.group(1) start, stop = (_msectotimecode(int(t)) for t in m.groups()[1:]) yield "{0} --> {1}".format(start, stop) else: yield line
return "\r\n".join(_fix_subtitle(subtitles))
|