]> Raphaël G. Git Repositories - youtubedl/blobdiff - youtube_dl/extractor/francetv.py
Annotate changelog bug being closed.
[youtubedl] / youtube_dl / extractor / francetv.py
index 461dac8efba90ea09f4f1e585d93754c81bd0136..170d6807529ac9b121187786cf9329b3b3525dc3 100644 (file)
@@ -1,38 +1,99 @@
 # encoding: utf-8
+
+from __future__ import unicode_literals
+
 import re
-import xml.etree.ElementTree
 import json
 
 from .common import InfoExtractor
-from ..utils import (
+from ..compat import (
+    compat_urllib_parse_urlparse,
     compat_urlparse,
 )
+from ..utils import (
+    clean_html,
+    ExtractorError,
+    int_or_none,
+    parse_duration,
+)
 
 
 class FranceTVBaseInfoExtractor(InfoExtractor):
-    def _extract_video(self, video_id):
-        xml_desc = self._download_webpage(
-            'http://www.francetvinfo.fr/appftv/webservices/video/'
-            'getInfosOeuvre.php?id-diffusion='
-            + video_id, video_id, 'Downloading XML config')
-        info = xml.etree.ElementTree.fromstring(xml_desc.encode('utf-8'))
-
-        manifest_url = info.find('videos/video/url').text
-        video_url = manifest_url.replace('manifest.f4m', 'index_2_av.m3u8')
-        video_url = video_url.replace('/z/', '/i/')
-        thumbnail_path = info.find('image').text
-
-        return {'id': video_id,
-                'ext': 'mp4',
-                'url': video_url,
-                'title': info.find('titre').text,
-                'thumbnail': compat_urlparse.urljoin('http://pluzz.francetv.fr', thumbnail_path),
-                'description': info.find('synopsis').text,
-                }
+    def _extract_video(self, video_id, catalogue):
+        info = self._download_json(
+            'http://webservices.francetelevisions.fr/tools/getInfosOeuvre/v2/?idDiffusion=%s&catalogue=%s'
+            % (video_id, catalogue),
+            video_id, 'Downloading video JSON')
+
+        if info.get('status') == 'NOK':
+            raise ExtractorError(
+                '%s returned error: %s' % (self.IE_NAME, info['message']), expected=True)
+        allowed_countries = info['videos'][0].get('geoblocage')
+        if allowed_countries:
+            georestricted = True
+            geo_info = self._download_json(
+                'http://geo.francetv.fr/ws/edgescape.json', video_id,
+                'Downloading geo restriction info')
+            country = geo_info['reponse']['geo_info']['country_code']
+            if country not in allowed_countries:
+                raise ExtractorError(
+                    'The video is not available from your location',
+                    expected=True)
+        else:
+            georestricted = False
+
+        formats = []
+        for video in info['videos']:
+            if video['statut'] != 'ONLINE':
+                continue
+            video_url = video['url']
+            if not video_url:
+                continue
+            format_id = video['format']
+            if video_url.endswith('.f4m'):
+                if georestricted:
+                    # See https://github.com/rg3/youtube-dl/issues/3963
+                    # m3u8 urls work fine
+                    continue
+                video_url_parsed = compat_urllib_parse_urlparse(video_url)
+                f4m_url = self._download_webpage(
+                    'http://hdfauth.francetv.fr/esi/urltokengen2.html?url=%s' % video_url_parsed.path,
+                    video_id, 'Downloading f4m manifest token', fatal=False)
+                if f4m_url:
+                    f4m_formats = self._extract_f4m_formats(f4m_url, video_id)
+                    for f4m_format in f4m_formats:
+                        f4m_format['preference'] = 1
+                    formats.extend(f4m_formats)
+            elif video_url.endswith('.m3u8'):
+                formats.extend(self._extract_m3u8_formats(video_url, video_id, 'mp4'))
+            elif video_url.startswith('rtmp'):
+                formats.append({
+                    'url': video_url,
+                    'format_id': 'rtmp-%s' % format_id,
+                    'ext': 'flv',
+                    'preference': 1,
+                })
+            else:
+                formats.append({
+                    'url': video_url,
+                    'format_id': format_id,
+                    'preference': -1,
+                })
+        self._sort_formats(formats)
+
+        return {
+            'id': video_id,
+            'title': info['titre'],
+            'description': clean_html(info['synopsis']),
+            'thumbnail': compat_urlparse.urljoin('http://pluzz.francetv.fr', info['image']),
+            'duration': parse_duration(info['duree']),
+            'timestamp': int_or_none(info['diffusion']['timestamp']),
+            'formats': formats,
+        }
 
 
 class PluzzIE(FranceTVBaseInfoExtractor):
-    IE_NAME = u'pluzz.francetv.fr'
+    IE_NAME = 'pluzz.francetv.fr'
     _VALID_URL = r'https?://pluzz\.francetv\.fr/videos/(.*?)\.html'
 
     # Can't use tests, videos expire in 7 days
@@ -42,88 +103,180 @@ class PluzzIE(FranceTVBaseInfoExtractor):
         webpage = self._download_webpage(url, title)
         video_id = self._search_regex(
             r'data-diffusion="(\d+)"', webpage, 'ID')
-        return self._extract_video(video_id)
+        return self._extract_video(video_id, 'Pluzz')
 
 
 class FranceTvInfoIE(FranceTVBaseInfoExtractor):
-    IE_NAME = u'francetvinfo.fr'
-    _VALID_URL = r'https?://www\.francetvinfo\.fr/replay.*/(?P<title>.+).html'
+    IE_NAME = 'francetvinfo.fr'
+    _VALID_URL = r'https?://(?:www|mobile)\.francetvinfo\.fr/.*/(?P<title>.+)\.html'
 
-    _TEST = {
-        u'url': u'http://www.francetvinfo.fr/replay-jt/france-3/soir-3/jt-grand-soir-3-lundi-26-aout-2013_393427.html',
-        u'file': u'84981923.mp4',
-        u'info_dict': {
-            u'title': u'Soir 3',
+    _TESTS = [{
+        'url': 'http://www.francetvinfo.fr/replay-jt/france-3/soir-3/jt-grand-soir-3-lundi-26-aout-2013_393427.html',
+        'info_dict': {
+            'id': '84981923',
+            'ext': 'flv',
+            'title': 'Soir 3',
+            'upload_date': '20130826',
+            'timestamp': 1377548400,
         },
-        u'params': {
-            u'skip_download': True,
+    }, {
+        'url': 'http://www.francetvinfo.fr/elections/europeennes/direct-europeennes-regardez-le-debat-entre-les-candidats-a-la-presidence-de-la-commission_600639.html',
+        'info_dict': {
+            'id': 'EV_20019',
+            'ext': 'mp4',
+            'title': 'Débat des candidats à la Commission européenne',
+            'description': 'Débat des candidats à la Commission européenne',
         },
-    }
+        'params': {
+            'skip_download': 'HLS (reqires ffmpeg)'
+        },
+        'skip': 'Ce direct est terminé et sera disponible en rattrapage dans quelques minutes.',
+    }]
 
     def _real_extract(self, url):
         mobj = re.match(self._VALID_URL, url)
         page_title = mobj.group('title')
         webpage = self._download_webpage(url, page_title)
-        video_id = self._search_regex(r'id-video=(\d+?)"', webpage, u'video id')
-        return self._extract_video(video_id)
+        video_id, catalogue = self._search_regex(
+            r'id-video=([^@]+@[^"]+)', webpage, 'video id').split('@')
+        return self._extract_video(video_id, catalogue)
 
 
-class France2IE(FranceTVBaseInfoExtractor):
-    IE_NAME = u'france2.fr'
-    _VALID_URL = r'''(?x)https?://www\.france2\.fr/
+class FranceTVIE(FranceTVBaseInfoExtractor):
+    IE_NAME = 'francetv'
+    IE_DESC = 'France 2, 3, 4, 5 and Ô'
+    _VALID_URL = r'''(?x)https?://www\.france[2345o]\.fr/
         (?:
-            emissions/.*?/videos/(?P<id>\d+)
-        |   emission/(?P<key>[^/?]+)
+            emissions/.*?/(videos|emissions)/(?P<id>[^/?]+)
+        |   (emissions?|jt)/(?P<key>[^/?]+)
         )'''
 
-    _TEST = {
-        u'url': u'http://www.france2.fr/emissions/13h15-le-samedi-le-dimanche/videos/75540104',
-        u'file': u'75540104.mp4',
-        u'info_dict': {
-            u'title': u'13h15, le samedi...',
-            u'description': u'md5:2e5b58ba7a2d3692b35c792be081a03d',
+    _TESTS = [
+        # france2
+        {
+            'url': 'http://www.france2.fr/emissions/13h15-le-samedi-le-dimanche/videos/75540104',
+            'md5': 'c03fc87cb85429ffd55df32b9fc05523',
+            'info_dict': {
+                'id': '109169362',
+                'ext': 'flv',
+                'title': '13h15, le dimanche...',
+                'description': 'md5:9a0932bb465f22d377a449be9d1a0ff7',
+                'upload_date': '20140914',
+                'timestamp': 1410693600,
+            },
         },
-        u'params': {
-            u'skip_download': True,
+        # france3
+        {
+            'url': 'http://www.france3.fr/emissions/pieces-a-conviction/diffusions/13-11-2013_145575',
+            'md5': '679bb8f8921f8623bd658fa2f8364da0',
+            'info_dict': {
+                'id': '000702326_CAPP_PicesconvictionExtrait313022013_120220131722_Au',
+                'ext': 'mp4',
+                'title': 'Le scandale du prix des médicaments',
+                'description': 'md5:1384089fbee2f04fc6c9de025ee2e9ce',
+                'upload_date': '20131113',
+                'timestamp': 1384380000,
+            },
         },
-    }
+        # france4
+        {
+            'url': 'http://www.france4.fr/emissions/hero-corp/videos/rhozet_herocorp_bonus_1_20131106_1923_06112013172108_F4',
+            'md5': 'a182bf8d2c43d88d46ec48fbdd260c1c',
+            'info_dict': {
+                'id': 'rhozet_herocorp_bonus_1_20131106_1923_06112013172108_F4',
+                'ext': 'mp4',
+                'title': 'Hero Corp Making of - Extrait 1',
+                'description': 'md5:c87d54871b1790679aec1197e73d650a',
+                'upload_date': '20131106',
+                'timestamp': 1383766500,
+            },
+        },
+        # france5
+        {
+            'url': 'http://www.france5.fr/emissions/c-a-dire/videos/92837968',
+            'md5': '78f0f4064f9074438e660785bbf2c5d9',
+            'info_dict': {
+                'id': '108961659',
+                'ext': 'flv',
+                'title': 'C à dire ?!',
+                'description': 'md5:1a4aeab476eb657bf57c4ff122129f81',
+                'upload_date': '20140915',
+                'timestamp': 1410795000,
+            },
+        },
+        # franceo
+        {
+            'url': 'http://www.franceo.fr/jt/info-afrique/04-12-2013',
+            'md5': '52f0bfe202848b15915a2f39aaa8981b',
+            'info_dict': {
+                'id': '108634970',
+                'ext': 'flv',
+                'title': 'Infô Afrique',
+                'description': 'md5:ebf346da789428841bee0fd2a935ea55',
+                'upload_date': '20140915',
+                'timestamp': 1410822000,
+            },
+        },
+    ]
 
     def _real_extract(self, url):
         mobj = re.match(self._VALID_URL, url)
-        if mobj.group('key'):
-            webpage = self._download_webpage(url, mobj.group('key'))
-            video_id = self._html_search_regex(
-                r'''(?x)<div\s+class="video-player">\s*
-                    <a\s+href="http://videos.francetv.fr/video/([0-9]+)"\s+
-                    class="francetv-video-player">''',
-                webpage, u'video ID')
-        else:
-            video_id = mobj.group('id')
-        return self._extract_video(video_id)
+        webpage = self._download_webpage(url, mobj.group('key') or mobj.group('id'))
+        video_id, catalogue = self._html_search_regex(
+            r'href="http://videos\.francetv\.fr/video/([^@]+@[^"]+)"',
+            webpage, 'video ID').split('@')
+        return self._extract_video(video_id, catalogue)
 
 
 class GenerationQuoiIE(InfoExtractor):
-    IE_NAME = u'http://generation-quoi.france2.fr'
-    _VALID_URL = r'https?://generation-quoi\.france2\.fr/portrait/(?P<name>.*)(\?|$)'
+    IE_NAME = 'france2.fr:generation-quoi'
+    _VALID_URL = r'https?://generation-quoi\.france2\.fr/portrait/(?P<id>[^/?#]+)'
 
     _TEST = {
-        u'url': u'http://generation-quoi.france2.fr/portrait/garde-a-vous',
-        u'file': u'k7FJX8VBcvvLmX4wA5Q.mp4',
-        u'info_dict': {
-            u'title': u'Génération Quoi - Garde à Vous',
-            u'uploader': u'Génération Quoi',
+        'url': 'http://generation-quoi.france2.fr/portrait/garde-a-vous',
+        'info_dict': {
+            'id': 'k7FJX8VBcvvLmX4wA5Q',
+            'ext': 'mp4',
+            'title': 'Génération Quoi - Garde à Vous',
+            'uploader': 'Génération Quoi',
         },
-        u'params': {
+        'params': {
             # It uses Dailymotion
-            u'skip_download': True,
+            'skip_download': True,
         },
     }
 
     def _real_extract(self, url):
-        mobj = re.match(self._VALID_URL, url)
-        name = mobj.group('name')
-        info_url = compat_urlparse.urljoin(url, '/medias/video/%s.json' % name)
-        info_json = self._download_webpage(info_url, name)
+        display_id = self._match_id(url)
+        info_url = compat_urlparse.urljoin(url, '/medias/video/%s.json' % display_id)
+        info_json = self._download_webpage(info_url, display_id)
         info = json.loads(info_json)
         return self.url_result('http://www.dailymotion.com/video/%s' % info['id'],
-            ie='Dailymotion')
+                               ie='Dailymotion')
+
+
+class CultureboxIE(FranceTVBaseInfoExtractor):
+    IE_NAME = 'culturebox.francetvinfo.fr'
+    _VALID_URL = r'https?://(?:m\.)?culturebox\.francetvinfo\.fr/(?P<name>.*?)(\?|$)'
+
+    _TEST = {
+        'url': 'http://culturebox.francetvinfo.fr/festivals/dans-les-jardins-de-william-christie/dans-les-jardins-de-william-christie-le-camus-162553',
+        'md5': '5ad6dec1ffb2a3fbcb20cc4b744be8d6',
+        'info_dict': {
+            'id': 'EV_22853',
+            'ext': 'flv',
+            'title': 'Dans les jardins de William Christie - Le Camus',
+            'description': 'md5:4710c82315c40f0c865ca8b9a68b5299',
+            'upload_date': '20140829',
+            'timestamp': 1409317200,
+        },
+    }
+
+    def _real_extract(self, url):
+        mobj = re.match(self._VALID_URL, url)
+        name = mobj.group('name')
+        webpage = self._download_webpage(url, name)
+        video_id, catalogue = self._search_regex(
+            r'"http://videos\.francetv\.fr/video/([^@]+@[^"]+)"', webpage, 'video id').split('@')
+
+        return self._extract_video(video_id, catalogue)