]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/extractor/dreisat.py
Annotate some highlights of the release.
[youtubedl] / youtube_dl / extractor / dreisat.py
index 8ac8587be6af564af3674c8ff7e7754364bc311e..848d387d10967eb99f33067d88f9b5a70156042f 100644 (file)
@@ -4,14 +4,19 @@ import re
 
 from .common import InfoExtractor
 from ..utils import (
-    ExtractorError,
+    int_or_none,
     unified_strdate,
+    xpath_text,
+    determine_ext,
+    float_or_none,
+    ExtractorError,
 )
 
 
 class DreiSatIE(InfoExtractor):
     IE_NAME = '3sat'
-    _VALID_URL = r'(?:http://)?(?:www\.)?3sat\.de/mediathek/(?:index\.php|mediathek\.php)?\?(?:(?:mode|display)=[^&]+&)*obj=(?P<id>[0-9]+)$'
+    _GEO_COUNTRIES = ['DE']
+    _VALID_URL = r'https?://(?:www\.)?3sat\.de/mediathek/(?:(?:index|mediathek)\.php)?\?(?:(?:mode|display)=[^&]+&)*obj=(?P<id>[0-9]+)'
     _TESTS = [
         {
             'url': 'http://www.3sat.de/mediathek/index.php?mode=play&obj=45918',
@@ -21,8 +26,12 @@ class DreiSatIE(InfoExtractor):
                 'ext': 'mp4',
                 'title': 'Waidmannsheil',
                 'description': 'md5:cce00ca1d70e21425e72c86a98a56817',
-                'uploader': '3sat',
+                'uploader': 'SCHWEIZWEIT',
+                'uploader_id': '100000210',
                 'upload_date': '20140913'
+            },
+            'params': {
+                'skip_download': True,  # m3u8 downloads
             }
         },
         {
@@ -31,57 +40,154 @@ class DreiSatIE(InfoExtractor):
         },
     ]
 
-    def _real_extract(self, url):
-        mobj = re.match(self._VALID_URL, url)
-        video_id = mobj.group('id')
-        details_url = 'http://www.3sat.de/mediathek/xmlservice/web/beitragsDetails?ak=web&id=%s' % video_id
-        details_doc = self._download_xml(details_url, video_id, 'Downloading video details')
-
-        status_code = details_doc.find('./status/statuscode')
-        if status_code is not None and status_code.text != 'ok':
-            code = status_code.text
-            if code == 'notVisibleAnymore':
+    def _parse_smil_formats(self, smil, smil_url, video_id, namespace=None, f4m_params=None, transform_rtmp_url=None):
+        param_groups = {}
+        for param_group in smil.findall(self._xpath_ns('./head/paramGroup', namespace)):
+            group_id = param_group.get(self._xpath_ns(
+                'id', 'http://www.w3.org/XML/1998/namespace'))
+            params = {}
+            for param in param_group:
+                params[param.get('name')] = param.get('value')
+            param_groups[group_id] = params
+
+        formats = []
+        for video in smil.findall(self._xpath_ns('.//video', namespace)):
+            src = video.get('src')
+            if not src:
+                continue
+            bitrate = int_or_none(self._search_regex(r'_(\d+)k', src, 'bitrate', None)) or float_or_none(video.get('system-bitrate') or video.get('systemBitrate'), 1000)
+            group_id = video.get('paramGroup')
+            param_group = param_groups[group_id]
+            for proto in param_group['protocols'].split(','):
+                formats.append({
+                    'url': '%s://%s' % (proto, param_group['host']),
+                    'app': param_group['app'],
+                    'play_path': src,
+                    'ext': 'flv',
+                    'format_id': '%s-%d' % (proto, bitrate),
+                    'tbr': bitrate,
+                })
+        self._sort_formats(formats)
+        return formats
+
+    def extract_from_xml_url(self, video_id, xml_url):
+        doc = self._download_xml(
+            xml_url, video_id,
+            note='Downloading video info',
+            errnote='Failed to download video info')
+
+        status_code = xpath_text(doc, './status/statuscode')
+        if status_code and status_code != 'ok':
+            if status_code == 'notVisibleAnymore':
                 message = 'Video %s is not available' % video_id
             else:
-                message = '%s returned error: %s' % (self.IE_NAME, code)
+                message = '%s returned error: %s' % (self.IE_NAME, status_code)
             raise ExtractorError(message, expected=True)
 
-        thumbnail_els = details_doc.findall('.//teaserimage')
-        thumbnails = [{
-            'width': int(te.attrib['key'].partition('x')[0]),
-            'height': int(te.attrib['key'].partition('x')[2]),
-            'url': te.text,
-        } for te in thumbnail_els]
-
-        information_el = details_doc.find('.//information')
-        video_title = information_el.find('./title').text
-        video_description = information_el.find('./detail').text
-
-        details_el = details_doc.find('.//details')
-        video_uploader = details_el.find('./channel').text
-        upload_date = unified_strdate(details_el.find('./airtime').text)
-
-        format_els = details_doc.findall('.//formitaet')
-        formats = [{
-            'format_id': fe.attrib['basetype'],
-            'width': int(fe.find('./width').text),
-            'height': int(fe.find('./height').text),
-            'url': fe.find('./url').text,
-            'filesize': int(fe.find('./filesize').text),
-            'video_bitrate': int(fe.find('./videoBitrate').text),
-        } for fe in format_els
-            if not fe.find('./url').text.startswith('http://www.metafilegenerator.de/')]
+        title = xpath_text(doc, './/information/title', 'title', True)
+
+        urls = []
+        formats = []
+        for fnode in doc.findall('.//formitaeten/formitaet'):
+            video_url = xpath_text(fnode, 'url')
+            if not video_url or video_url in urls:
+                continue
+            urls.append(video_url)
+
+            is_available = 'http://www.metafilegenerator' not in video_url
+            geoloced = 'static_geoloced_online' in video_url
+            if not is_available or geoloced:
+                continue
+
+            format_id = fnode.attrib['basetype']
+            format_m = re.match(r'''(?x)
+                (?P<vcodec>[^_]+)_(?P<acodec>[^_]+)_(?P<container>[^_]+)_
+                (?P<proto>[^_]+)_(?P<index>[^_]+)_(?P<indexproto>[^_]+)
+            ''', format_id)
+
+            ext = determine_ext(video_url, None) or format_m.group('container')
+
+            if ext == 'meta':
+                continue
+            elif ext == 'smil':
+                formats.extend(self._extract_smil_formats(
+                    video_url, video_id, fatal=False))
+            elif ext == 'm3u8':
+                # the certificates are misconfigured (see
+                # https://github.com/ytdl-org/youtube-dl/issues/8665)
+                if video_url.startswith('https://'):
+                    continue
+                formats.extend(self._extract_m3u8_formats(
+                    video_url, video_id, 'mp4', 'm3u8_native',
+                    m3u8_id=format_id, fatal=False))
+            elif ext == 'f4m':
+                formats.extend(self._extract_f4m_formats(
+                    video_url, video_id, f4m_id=format_id, fatal=False))
+            else:
+                quality = xpath_text(fnode, './quality')
+                if quality:
+                    format_id += '-' + quality
+
+                abr = int_or_none(xpath_text(fnode, './audioBitrate'), 1000)
+                vbr = int_or_none(xpath_text(fnode, './videoBitrate'), 1000)
+
+                tbr = int_or_none(self._search_regex(
+                    r'_(\d+)k', video_url, 'bitrate', None))
+                if tbr and vbr and not abr:
+                    abr = tbr - vbr
+
+                formats.append({
+                    'format_id': format_id,
+                    'url': video_url,
+                    'ext': ext,
+                    'acodec': format_m.group('acodec'),
+                    'vcodec': format_m.group('vcodec'),
+                    'abr': abr,
+                    'vbr': vbr,
+                    'tbr': tbr,
+                    'width': int_or_none(xpath_text(fnode, './width')),
+                    'height': int_or_none(xpath_text(fnode, './height')),
+                    'filesize': int_or_none(xpath_text(fnode, './filesize')),
+                    'protocol': format_m.group('proto').lower(),
+                })
+
+        geolocation = xpath_text(doc, './/details/geolocation')
+        if not formats and geolocation and geolocation != 'none':
+            self.raise_geo_restricted(countries=self._GEO_COUNTRIES)
 
         self._sort_formats(formats)
 
+        thumbnails = []
+        for node in doc.findall('.//teaserimages/teaserimage'):
+            thumbnail_url = node.text
+            if not thumbnail_url:
+                continue
+            thumbnail = {
+                'url': thumbnail_url,
+            }
+            thumbnail_key = node.get('key')
+            if thumbnail_key:
+                m = re.match('^([0-9]+)x([0-9]+)$', thumbnail_key)
+                if m:
+                    thumbnail['width'] = int(m.group(1))
+                    thumbnail['height'] = int(m.group(2))
+            thumbnails.append(thumbnail)
+
+        upload_date = unified_strdate(xpath_text(doc, './/details/airtime'))
+
         return {
-            '_type': 'video',
             'id': video_id,
-            'title': video_title,
-            'formats': formats,
-            'description': video_description,
+            'title': title,
+            'description': xpath_text(doc, './/information/detail'),
+            'duration': int_or_none(xpath_text(doc, './/details/lengthSec')),
             'thumbnails': thumbnails,
-            'thumbnail': thumbnails[-1]['url'],
-            'uploader': video_uploader,
+            'uploader': xpath_text(doc, './/details/originChannelTitle'),
+            'uploader_id': xpath_text(doc, './/details/originChannelId'),
             'upload_date': upload_date,
+            'formats': formats,
         }
+
+    def _real_extract(self, url):
+        video_id = self._match_id(url)
+        details_url = 'http://www.3sat.de/mediathek/xmlservice/web/beitragsDetails?id=%s' % video_id
+        return self.extract_from_xml_url(video_id, details_url)