aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor')
-rw-r--r--youtube_dl/extractor/nfb.py108
1 files changed, 63 insertions, 45 deletions
diff --git a/youtube_dl/extractor/nfb.py b/youtube_dl/extractor/nfb.py
index 234e49047..adcc636bc 100644
--- a/youtube_dl/extractor/nfb.py
+++ b/youtube_dl/extractor/nfb.py
@@ -2,8 +2,12 @@ from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import (
- sanitized_Request,
+ clean_html,
+ determine_ext,
+ int_or_none,
+ qualities,
urlencode_postdata,
+ xpath_text,
)
@@ -16,12 +20,12 @@ class NFBIE(InfoExtractor):
'url': 'https://www.nfb.ca/film/qallunaat_why_white_people_are_funny',
'info_dict': {
'id': 'qallunaat_why_white_people_are_funny',
- 'ext': 'mp4',
+ 'ext': 'flv',
'title': 'Qallunaat! Why White People Are Funny ',
- 'description': 'md5:836d8aff55e087d04d9f6df554d4e038',
+ 'description': 'md5:6b8e32dde3abf91e58857b174916620c',
'duration': 3128,
+ 'creator': 'Mark Sandiford',
'uploader': 'Mark Sandiford',
- 'uploader_id': 'mark-sandiford',
},
'params': {
# rtmp download
@@ -31,64 +35,78 @@ class NFBIE(InfoExtractor):
def _real_extract(self, url):
video_id = self._match_id(url)
- page = self._download_webpage(
- 'https://www.nfb.ca/film/%s' % video_id, video_id,
- 'Downloading film page')
- uploader_id = self._html_search_regex(r'<a class="director-link" href="/explore-all-directors/([^/]+)/"',
- page, 'director id', fatal=False)
- uploader = self._og_search_property('video:director', page, 'director name')
-
- request = sanitized_Request(
+ config = self._download_xml(
'https://www.nfb.ca/film/%s/player_config' % video_id,
- urlencode_postdata({'getConfig': 'true'}))
- request.add_header('Content-Type', 'application/x-www-form-urlencoded')
- request.add_header('X-NFB-Referer', 'http://www.nfb.ca/medias/flash/NFBVideoPlayer.swf')
-
- config = self._download_xml(request, video_id, 'Downloading player config XML')
+ video_id, 'Downloading player config XML',
+ data=urlencode_postdata({'getConfig': 'true'}),
+ headers={
+ 'Content-Type': 'application/x-www-form-urlencoded',
+ 'X-NFB-Referer': 'http://www.nfb.ca/medias/flash/NFBVideoPlayer.swf'
+ })
- title = None
- description = None
- thumbnail = None
- duration = None
- formats = []
-
- def extract_thumbnail(media):
- thumbnails = {}
- for asset in media.findall('assets/asset'):
- thumbnails[asset.get('quality')] = asset.find('default/url').text
- if not thumbnails:
- return None
- if 'high' in thumbnails:
- return thumbnails['high']
- return list(thumbnails.values())[0]
+ title, description, thumbnail, duration, uploader, author = [None] * 6
+ thumbnails, formats = [[]] * 2
+ subtitles = {}
for media in config.findall('./player/stream/media'):
if media.get('type') == 'posterImage':
- thumbnail = extract_thumbnail(media)
+ quality_key = qualities(('low', 'high'))
+ thumbnails = []
+ for asset in media.findall('assets/asset'):
+ asset_url = xpath_text(asset, 'default/url', default=None)
+ if not asset_url:
+ continue
+ quality = asset.get('quality')
+ thumbnails.append({
+ 'url': asset_url,
+ 'id': quality,
+ 'preference': quality_key(quality),
+ })
elif media.get('type') == 'video':
- duration = int(media.get('duration'))
- title = media.find('title').text
- description = media.find('description').text
- # It seems assets always go from lower to better quality, so no need to sort
+ title = xpath_text(media, 'title', fatal=True)
for asset in media.findall('assets/asset'):
- for x in asset:
+ quality = asset.get('quality')
+ height = int_or_none(self._search_regex(
+ r'^(\d+)[pP]$', quality or '', 'height', default=None))
+ for node in asset:
+ streamer = xpath_text(node, 'streamerURI', default=None)
+ if not streamer:
+ continue
+ play_path = xpath_text(node, 'url', default=None)
+ if not play_path:
+ continue
formats.append({
- 'url': x.find('streamerURI').text,
- 'app': x.find('streamerURI').text.split('/', 3)[3],
- 'play_path': x.find('url').text,
+ 'url': streamer,
+ 'app': streamer.split('/', 3)[3],
+ 'play_path': play_path,
'rtmp_live': False,
- 'ext': 'mp4',
- 'format_id': '%s-%s' % (x.tag, asset.get('quality')),
+ 'ext': 'flv',
+ 'format_id': '%s-%s' % (node.tag, quality) if quality else node.tag,
+ 'height': height,
})
+ self._sort_formats(formats)
+ description = clean_html(xpath_text(media, 'description'))
+ uploader = xpath_text(media, 'author')
+ duration = int_or_none(media.get('duration'))
+ for subtitle in media.findall('./subtitles/subtitle'):
+ subtitle_url = xpath_text(subtitle, 'url', default=None)
+ if not subtitle_url:
+ continue
+ lang = xpath_text(subtitle, 'lang', default='en')
+ subtitles.setdefault(lang, []).append({
+ 'url': subtitle_url,
+ 'ext': (subtitle.get('format') or determine_ext(subtitle_url)).lower(),
+ })
return {
'id': video_id,
'title': title,
'description': description,
- 'thumbnail': thumbnail,
+ 'thumbnails': thumbnails,
'duration': duration,
+ 'creator': uploader,
'uploader': uploader,
- 'uploader_id': uploader_id,
'formats': formats,
+ 'subtitles': subtitles,
}