- # coding: utf-8
- from __future__ import unicode_literals
-
- from .common import InfoExtractor
- from ..utils import (
- float_or_none,
- int_or_none,
- parse_iso8601,
- sanitized_Request,
- )
-
-
- class EitbIE(InfoExtractor):
- IE_NAME = 'eitb.tv'
- _VALID_URL = r'https?://(?:www\.)?eitb\.tv/(?:eu/bideoa|es/video)/[^/]+/\d+/(?P<id>\d+)'
-
- _TEST = {
- 'url': 'http://www.eitb.tv/es/video/60-minutos-60-minutos-2013-2014/4104995148001/4090227752001/lasa-y-zabala-30-anos/',
- 'md5': 'edf4436247185adee3ea18ce64c47998',
- 'info_dict': {
- 'id': '4090227752001',
- 'ext': 'mp4',
- 'title': '60 minutos (Lasa y Zabala, 30 años)',
- 'description': 'Programa de reportajes de actualidad.',
- 'duration': 3996.76,
- 'timestamp': 1381789200,
- 'upload_date': '20131014',
- 'tags': list,
- },
- }
-
- def _real_extract(self, url):
- video_id = self._match_id(url)
-
- video = self._download_json(
- 'http://mam.eitb.eus/mam/REST/ServiceMultiweb/Video/MULTIWEBTV/%s/' % video_id,
- video_id, 'Downloading video JSON')
-
- media = video['web_media'][0]
-
- formats = []
- for rendition in media['RENDITIONS']:
- video_url = rendition.get('PMD_URL')
- if not video_url:
- continue
- tbr = float_or_none(rendition.get('ENCODING_RATE'), 1000)
- format_id = 'http'
- if tbr:
- format_id += '-%d' % int(tbr)
- formats.append({
- 'url': rendition['PMD_URL'],
- 'format_id': format_id,
- 'width': int_or_none(rendition.get('FRAME_WIDTH')),
- 'height': int_or_none(rendition.get('FRAME_HEIGHT')),
- 'tbr': tbr,
- })
-
- hls_url = media.get('HLS_SURL')
- if hls_url:
- request = sanitized_Request(
- 'http://mam.eitb.eus/mam/REST/ServiceMultiweb/DomainRestrictedSecurity/TokenAuth/',
- headers={'Referer': url})
- token_data = self._download_json(
- request, video_id, 'Downloading auth token', fatal=False)
- if token_data:
- token = token_data.get('token')
- if token:
- formats.extend(self._extract_m3u8_formats(
- '%s?hdnts=%s' % (hls_url, token), video_id, m3u8_id='hls', fatal=False))
-
- hds_url = media.get('HDS_SURL')
- if hds_url:
- formats.extend(self._extract_f4m_formats(
- '%s?hdcore=3.7.0' % hds_url.replace('euskalsvod', 'euskalvod'),
- video_id, f4m_id='hds', fatal=False))
-
- self._sort_formats(formats)
-
- return {
- 'id': video_id,
- 'title': media.get('NAME_ES') or media.get('name') or media['NAME_EU'],
- 'description': media.get('SHORT_DESC_ES') or video.get('desc_group') or media.get('SHORT_DESC_EU'),
- 'thumbnail': media.get('STILL_URL') or media.get('THUMBNAIL_URL'),
- 'duration': float_or_none(media.get('LENGTH'), 1000),
- 'timestamp': parse_iso8601(media.get('BROADCST_DATE'), ' '),
- 'tags': media.get('TAGS'),
- 'formats': formats,
- }
|