|
|
@ -1,23 +1,23 @@ |
|
|
|
# coding: utf-8 |
|
|
|
from __future__ import unicode_literals |
|
|
|
|
|
|
|
import re |
|
|
|
|
|
|
|
from .common import InfoExtractor |
|
|
|
from ..utils import ( |
|
|
|
parse_duration, |
|
|
|
parse_iso8601, |
|
|
|
js_to_json, |
|
|
|
) |
|
|
|
from ..compat import compat_str |
|
|
|
|
|
|
|
|
|
|
|
class RDSIE(InfoExtractor): |
|
|
|
IE_DESC = 'RDS.ca' |
|
|
|
_VALID_URL = r'https?://(?:www\.)?rds\.ca/vid(?:[eé]|%C3%A9)os/(?:[^/]+/)*(?P<display_id>[^/]+)-(?P<id>\d+\.\d+)' |
|
|
|
_VALID_URL = r'https?://(?:www\.)?rds\.ca/vid(?:[eé]|%C3%A9)os/(?:[^/]+/)*(?P<id>[^/]+)-\d+\.\d+' |
|
|
|
|
|
|
|
_TESTS = [{ |
|
|
|
'url': 'http://www.rds.ca/videos/football/nfl/fowler-jr-prend-la-direction-de-jacksonville-3.1132799', |
|
|
|
'info_dict': { |
|
|
|
'id': '3.1132799', |
|
|
|
'id': '604333', |
|
|
|
'display_id': 'fowler-jr-prend-la-direction-de-jacksonville', |
|
|
|
'ext': 'mp4', |
|
|
|
'title': 'Fowler Jr. prend la direction de Jacksonville', |
|
|
@ -33,22 +33,17 @@ class RDSIE(InfoExtractor): |
|
|
|
}] |
|
|
|
|
|
|
|
def _real_extract(self, url): |
|
|
|
mobj = re.match(self._VALID_URL, url) |
|
|
|
video_id = mobj.group('id') |
|
|
|
display_id = mobj.group('display_id') |
|
|
|
display_id = self._match_id(url) |
|
|
|
|
|
|
|
webpage = self._download_webpage(url, display_id) |
|
|
|
|
|
|
|
# TODO: extract f4m from 9c9media.com |
|
|
|
video_url = self._search_regex( |
|
|
|
r'<span[^>]+itemprop="contentURL"[^>]+content="([^"]+)"', |
|
|
|
webpage, 'video url') |
|
|
|
|
|
|
|
title = self._og_search_title(webpage) or self._html_search_meta( |
|
|
|
item = self._parse_json(self._search_regex(r'(?s)itemToPush\s*=\s*({.+?});', webpage, 'item'), display_id, js_to_json) |
|
|
|
video_id = compat_str(item['id']) |
|
|
|
title = item.get('title') or self._og_search_title(webpage) or self._html_search_meta( |
|
|
|
'title', webpage, 'title', fatal=True) |
|
|
|
description = self._og_search_description(webpage) or self._html_search_meta( |
|
|
|
'description', webpage, 'description') |
|
|
|
thumbnail = self._og_search_thumbnail(webpage) or self._search_regex( |
|
|
|
thumbnail = item.get('urlImageBig') or self._og_search_thumbnail(webpage) or self._search_regex( |
|
|
|
[r'<link[^>]+itemprop="thumbnailUrl"[^>]+href="([^"]+)"', |
|
|
|
r'<span[^>]+itemprop="thumbnailUrl"[^>]+content="([^"]+)"'], |
|
|
|
webpage, 'thumbnail', fatal=False) |
|
|
@ -61,13 +56,15 @@ class RDSIE(InfoExtractor): |
|
|
|
age_limit = self._family_friendly_search(webpage) |
|
|
|
|
|
|
|
return { |
|
|
|
'_type': 'url_transparent', |
|
|
|
'id': video_id, |
|
|
|
'display_id': display_id, |
|
|
|
'url': video_url, |
|
|
|
'url': '9c9media:rds_web:%s' % video_id, |
|
|
|
'title': title, |
|
|
|
'description': description, |
|
|
|
'thumbnail': thumbnail, |
|
|
|
'timestamp': timestamp, |
|
|
|
'duration': duration, |
|
|
|
'age_limit': age_limit, |
|
|
|
'ie_key': 'NineCNineMedia', |
|
|
|
} |