Remove poor format sorting

Default format sorting produces a significantly better result now that
metadata is available more consistently.

yt_dlp/extractor/dreisat.py

@@ -27,7 +27,7 @@ class DreiSatIE(ZDFBaseIE):
         },
     }, {
         'url': 'https://www.3sat.de/film/ab-18/ab-18---mein-fremdes-ich-100.html',
-        'md5': '66cb9013ce37f6e008dc99bfcf1356bc',
+        'md5': 'f92638413a11d759bdae95c9d8ec165c',
         'info_dict': {
             'id': '221128_mein_fremdes_ich2_ab18',
             'ext': 'mp4',
@@ -42,7 +42,7 @@ class DreiSatIE(ZDFBaseIE):
         },
     }, {
         'url': 'https://www.3sat.de/gesellschaft/37-grad-leben/aus-dem-leben-gerissen-102.html',
-        'md5': '69f276184c9a24147e4baae728fbc3c4',
+        'md5': 'a903eaf8d1fd635bd3317cd2ad87ec84',
         'info_dict': {
             'id': '250323_0903_sendung_sgl',
             'ext': 'mp4',
@@ -58,7 +58,7 @@ class DreiSatIE(ZDFBaseIE):
     }, {
         # Video with chapters
         'url': 'https://www.3sat.de/kultur/buchmesse/dein-buch-das-beste-von-der-leipziger-buchmesse-2025-teil-1-100.html',
-        'md5': '65e8f365799d2266c9110d83d748445e',
+        'md5': '6b95790ce52e75f0d050adcdd2711ee6',
         'info_dict': {
             'id': '250330_dein_buch1_bum',
             'ext': 'mp4',
@@ -97,7 +97,9 @@ class DreiSatIE(ZDFBaseIE):
             (('streams', 'default'), None),
             ('http://zdf.de/rels/streams/ptmd', 'http://zdf.de/rels/streams/ptmd-template'),
             {str}, any, {require('ptmd path')}))
-        info = self._extract_ptmd(player_url, ptmd_path, video_id, api_token)
+        ptmd_url = self._expand_ptmd_template(player_url, ptmd_path)
+        aspect_ratio = self._parse_aspect_ratio(video_target.get('aspectRatio'))
+        info = self._extract_ptmd_urls(ptmd_url, video_id, api_token, aspect_ratio)
 
         return merge_dicts(info, {
             **traverse_obj(content, {

yt_dlp/extractor/phoenix.py

@@ -13,7 +13,7 @@ class PhoenixIE(ZDFBaseIE):
     _VALID_URL = r'https?://(?:www\.)?phoenix\.de/(?:[^/?#]+/)*[^/?#&]*-a-(?P<id>\d+)\.html'
     _TESTS = [{
         'url': 'https://www.phoenix.de/sendungen/dokumentationen/spitzbergen-a-893349.html',
-        'md5': '768bbcd67a51fbfb22263ecdbc49146d',
+        'md5': 'a79e86d9774d0b3f2102aff988a0bd32',
         'info_dict': {
             'id': '221215_phx_spitzbergen',
             'ext': 'mp4',
@@ -70,9 +70,8 @@ class PhoenixIE(ZDFBaseIE):
         title = title or details['title']
         content_id = details['tracking']['nielsen']['content']['assetid']
 
-        info = self._extract_ptmd(
-            'https://tmd.phoenix.de',
-            f'tmd/2/{{playerId}}/vod/ptmd/phoenix/{content_id}',
+        info = self._extract_ptmd_urls(
+            f'https://tmd.phoenix.de/tmd/2/android_native_6/vod/ptmd/phoenix/{content_id}',
             content_id)
 
         duration = int_or_none(try_get(

yt_dlp/extractor/zdf.py

@@ -1,3 +1,4 @@
+import functools
 import json
 import re
 import time
@@ -14,8 +15,6 @@ from ..utils import (
     parse_codecs,
     parse_iso8601,
     parse_qs,
-    qualities,
-    try_get,
     unified_timestamp,
     url_or_none,
     urljoin,
@@ -26,7 +25,6 @@ from ..utils.traversal import traverse_obj
 
 class ZDFBaseIE(InfoExtractor):
     _GEO_COUNTRIES = ['DE']
-    _QUALITIES = ('auto', 'low', 'med', 'high', 'veryhigh', 'hd', 'fhd', 'uhd')
 
     def _call_api(self, url, video_id, item, api_token=None):
         headers = {'Api-Auth': api_token} if api_token else {}
@@ -34,6 +32,12 @@ class ZDFBaseIE(InfoExtractor):
             url, video_id, note=f'Downloading {item}',
             errnote=f'Failed to download {item}', headers=headers)
 
+    def _parse_aspect_ratio(self, aspect_ratio):
+        if not aspect_ratio or not isinstance(aspect_ratio, str):
+            return None
+        mobj = re.match(r'(?P<width>\d+):(?P<height>\d+)', aspect_ratio)
+        return int(mobj.group('width')) / int(mobj.group('height')) if mobj else None
+
     def _extract_chapters(self, data):
         return traverse_obj(data, (lambda _, v: 'anchorOffset' in v, {
             'start_time': ('anchorOffset', {float_or_none}),
@@ -55,75 +59,66 @@ class ZDFBaseIE(InfoExtractor):
             })
         return subtitles
 
-    def _extract_format(self, video_id, formats, format_urls, meta):
-        format_url = meta['url']
-        if format_url in format_urls:
-            return
-        format_urls.add(format_url)
-        ext = determine_ext(format_url)
-        if ext == 'm3u8':
-            fmts = self._extract_m3u8_formats(
-                format_url, video_id, 'mp4', m3u8_id='hls', fatal=False)
-        elif ext == 'mpd':
-            fmts = self._extract_mpd_formats(
-                format_url, video_id, mpd_id='dash', fatal=False)
-        else:
-            f = parse_codecs(meta.get('mimeCodec'))
-            if not f and meta.get('type'):
-                data = meta['type'].split('_')
-                if len(data) >= 3 and data[2] == ext:
-                    f = {'vcodec': data[0], 'acodec': data[1]}
-            f.update({
-                'url': format_url,
-                'height': int_or_none(meta.get('highestVerticalResolution')),
-                'format_id': join_nonempty('http', meta.get('type'), meta.get('quality')),
-                'tbr': int_or_none(self._search_regex(r'_(\d+)k_', format_url, 'tbr', default=None)),
+    def _expand_ptmd_template(self, api_base_url, template):
+        return urljoin(api_base_url, template.replace('{playerId}', 'android_native_6'))
+
+    def _extract_ptmd_urls(self, ptmd_urls, video_id, api_token=None, aspect_ratio=None):
+        ptmd_urls = variadic(ptmd_urls)
+        ptmd_info = []
+        for url in ptmd_urls:
+            ptmd_info.append({
+                'url': url,
+                'dgs': False,
             })
-            fmts = [f]
-        formats.extend(merge_dicts(f, {
-            'format_note': join_nonempty(meta.get('quality'), meta.get('class'), delim=', '),
-            'language': meta.get('language'),
-            'language_preference': 10 if meta.get('class') == 'main' else -10 if meta.get('class') == 'ad' else -1,
-            'quality': qualities(self._QUALITIES)(meta.get('quality')),
-        }) for f in fmts)
+        return self._extract_ptmd(ptmd_info, video_id, api_token, aspect_ratio)
 
-    def _extract_ptmd(self, api_base_url, templates, video_id, api_token=None):
-        # TODO: If there are multiple PTMD templates,
-        # usually one of them is a sign-language variant of the video.
-        # The format order works out fine as is and prefers the "regular" video,
-        # but this should probably be made more explicit.
-
-        # TODO: HTTPS formats are extracted without resolution information
-        # However, we know vertical resolution and the caller often knows apsect ratio.
-        # So we could calculate the correct resulution from those two data points.
-        templates = variadic(templates)
+    def _extract_ptmd(self, ptmd_info, video_id, api_token=None, aspect_ratio=None):
+        ptmd_info = variadic(ptmd_info)
         src_captions = []
 
         content_id = None
         duration = None
         formats = []
         track_uris = set()
-        for template in templates:
-            ptmd_url = urljoin(api_base_url, template.replace(
-                '{playerId}', 'android_native_6'))
-            ptmd = self._call_api(ptmd_url, video_id, 'PTMD data', api_token)
-            # As per above TODO on sign language videos variants,
-            # prefer content_id from the last entry to get the "regular" ID.
-            content_id = ptmd.get('basename') or ptmd_url.split('/')[-2]
+        for info in ptmd_info:
+            ptmd = self._call_api(info['url'], video_id, 'PTMD data', api_token)
+            basename = ptmd.get('basename') or info['url'].split('/')[-2]
+            if not content_id and not info['dgs']:
+                content_id = basename
             duration = (duration or traverse_obj(ptmd, ('attributes', 'duration', 'value', {float_or_none(scale=1000)})))
             src_captions += ptmd.get('captions') or []
             for stream in traverse_obj(ptmd, ('priorityList', ..., 'formitaeten', ..., {dict})):
                 for quality in traverse_obj(stream, ('qualities', ..., {dict})):
                     for variant in traverse_obj(quality, ('audio', 'tracks', lambda _, v: url_or_none(v['uri']))):
-                        self._extract_format(
-                            content_id, formats, track_uris, {
-                                'url': variant.get('uri'),
-                                'type': stream.get('type'),
-                                'mimeCodec': quality.get('mimeCodec'),
-                                'quality': quality.get('quality'),
-                                'class': variant.get('class'),
-                                'language': variant.get('language'),
-                            })
+                        format_url = variant['uri']
+                        if format_url in track_uris:
+                            continue
+                        track_uris.add(format_url)
+                        ext = determine_ext(format_url)
+                        if ext == 'm3u8':
+                            fmts = self._extract_m3u8_formats(
+                                format_url, basename, 'mp4', m3u8_id='hls', fatal=False)
+                        elif ext == 'mpd':
+                            fmts = self._extract_mpd_formats(
+                                format_url, basename, mpd_id='dash', fatal=False)
+                        else:
+                            height = int_or_none(quality.get('highestVerticalResolution'))
+                            width = round(aspect_ratio * height) if aspect_ratio and height else None
+                            fmts = [{
+                                'url': format_url,
+                                **parse_codecs(quality.get('mimeCodec')),
+                                'height': height,
+                                'width': width,
+                                'format_id': join_nonempty('http', stream.get('type')),
+                                'tbr': int_or_none(self._search_regex(r'_(\d+)k_', format_url, 'tbr', default=None)),
+                            }]
+                        formats.extend(merge_dicts(f, {
+                            'format_note': join_nonempty(
+                                variant.get('class'), 'dgs' if info['dgs'] else '', delim=', '),
+                            'language': variant.get('language'),
+                            'preference': -2 if info['dgs'] else -1,
+                            'language_preference': 10 if variant.get('class') == 'main' else -10 if variant.get('class') == 'ad' else -1,
+                        }) for f in fmts)
 
         return {
             'extractor_key': ZDFIE.ie_key(),
@@ -131,7 +126,6 @@ class ZDFBaseIE(InfoExtractor):
             'duration': duration,
             'formats': formats,
             'subtitles': self._extract_subtitles(src_captions),
-            '_format_sort_fields': ('tbr', 'res', 'quality', 'language_preference'),
         }
 
     def _get_api_token(self, video_id):
@@ -303,7 +297,7 @@ class ZDFIE(ZDFBaseIE):
         # Video with chapters
         # Also: video with sign-language variant
         'url': 'https://www.zdf.de/video/magazine/heute-journal-104/heute-journal-vom-19-12-2021-100',
-        'md5': '1175003f28507bd27b266181c4de9f56',
+        'md5': '6ada39465497a84fb98d48ffff69e7b7',
         'info_dict': {
             'id': 'heute-journal-vom-19-12-2021-100',
             'ext': 'mp4',
@@ -326,7 +320,7 @@ class ZDFIE(ZDFBaseIE):
     }, {
         # Video that requires fallback extraction
         'url': 'https://www.zdf.de/nachrichten/politik/deutschland/koalitionsverhandlungen-spd-cdu-csu-dobrindt-100.html',
-        'md5': '95903ecbd37f2881b4462d074b8f8c44',
+        'md5': 'c3a78514dd993a5781aa3afe50db51e2',
         'info_dict': {
             'id': 'koalitionsverhandlungen-spd-cdu-csu-dobrindt-100',
             'ext': 'mp4',
@@ -381,7 +375,7 @@ class ZDFIE(ZDFBaseIE):
         'url': 'https://www.zdf.de/serien/northern-lights/begegnung-auf-der-bruecke-100.html',
         'info_dict': {
             'id': 'begegnung-auf-der-bruecke-100',
-            'ext': 'mp4',
+            'ext': 'webm',
             'title': 'Begegnung auf der Brücke',
             'description': 'md5:e53a555da87447f7f1207f10353f8e45',
             'duration': 3083.0,
@@ -483,6 +477,16 @@ query VideoByCanonical($canonical: String!) {
 }
     '''
 
+    def _extract_ptmd(self, ptmd_info, video_id, api_token=None, aspect_ratio=None):
+        ptmd_data = super()._extract_ptmd(ptmd_info, video_id, api_token, aspect_ratio)
+        # We can't use the ID from PTMD extraction as the video ID
+        # because it is not available during playlist extraction.
+        # We fix it here manually instead of inside the base class
+        # because other extractors do rely on using it as their ID.
+        ptmd_data['_old_archive_ids'] = [make_archive_id(ZDFIE, ptmd_data['id'])]
+        del ptmd_data['id']
+        return ptmd_data
+
     # This fallback should generally only happen for pages under `zdf.de/nachrichten`.
     # They are on a separate website for which GraphQL often doesn't return results.
     # The API used here is no longer in use by official clients and likely deprecated.
@@ -495,11 +499,10 @@ query VideoByCanonical($canonical: String!) {
             errnote='Failed to download fallback metadata')
         document = video['document']
 
-        content_id = document.get('basename')
-        formats = []
-        format_urls = set()
-        for f in traverse_obj(document, ('formitaeten', ..., {dict})):
-            self._extract_format(document_id, formats, format_urls, f)
+        ptmd_url = traverse_obj(document, (
+            ('streamApiUrlAndroid', ('streams', 0, 'streamApiUrlAndroid')),
+            {url_or_none}, any))
+        ptmd_data = self._extract_ptmd_urls(ptmd_url, document_id, self._get_api_token(document_id))
 
         thumbnails = []
         for thumbnail_key, thumbnail in traverse_obj(document, ('teaserBild', {dict.items})):
@@ -515,15 +518,16 @@ query VideoByCanonical($canonical: String!) {
 
         return {
             'id': document_id,
-            'title': document.get('titel'),
-            'description': document.get('beschreibung'),
-            'duration': int_or_none(document.get('length')),
-            'timestamp': unified_timestamp(document.get('date')) or unified_timestamp(
-                try_get(video, lambda x: x['meta']['editorialDate'], str)),
             'thumbnails': thumbnails,
-            'subtitles': self._extract_subtitles(document.get('captions') or []),
-            'formats': formats,
-            '_old_archive_ids': [make_archive_id(ZDFIE, content_id)] if content_id else [],
+            **ptmd_data,
+            **traverse_obj(video, {
+                'title': ('document', 'titel', {str}),
+                'description': ('document', 'beschreibung', {str}),
+                'timestamp': (
+                    (('document', 'date'), ('meta', 'editorialDate')),
+                    {unified_timestamp}, any),
+                'subtitles': ('document', 'captions', {self._extract_subtitles}),
+            }),
         }
 
     def _real_extract(self, url):
@@ -536,22 +540,20 @@ query VideoByCanonical($canonical: String!) {
 
         if not video_data:
             return self._extract_fallback(video_id)
-
-        ptmd_templates = traverse_obj(
-            video_data, ('currentMedia', 'nodes', ..., 'ptmdTemplate'))
-        ptmd_data = self._extract_ptmd(
-            'https://api.zdf.de', ptmd_templates, video_id,
-            self._get_api_token(video_id))
-        # We can't use the ID from PTMD extraction as the video ID
-        # because it is not available during playlist extraction.
-        # We fix it here manually instead of inside the method
-        # because other extractors do rely on using it as their ID.
-        ptmd_data['_old_archive_ids'] = [make_archive_id(ZDFIE, ptmd_data['id'])]
-        del ptmd_data['id']
+        ptmd_nodes = traverse_obj(video_data, ('currentMedia', 'nodes'))
+        ptmd_info = traverse_obj(ptmd_nodes, (..., {
+            'url': ('ptmdTemplate', {functools.partial(self._expand_ptmd_template, 'https://api.zdf.de')}),
+            # Sign-language variant (DGS = *D*eutsche *G*ebärden*s*prache')
+            'dgs': ('vodMediaType', {lambda x: x == 'DGS'}),
+        }))
+        aspect_ratio = traverse_obj(ptmd_nodes, (
+            ..., 'aspectRatio', {self._parse_aspect_ratio}, any))
+        ptmd_result = self._extract_ptmd(
+            ptmd_info, video_id, self._get_api_token(video_id), aspect_ratio)
 
         return {
             'id': video_id,
-            **ptmd_data,
+            **ptmd_result,
             **traverse_obj(video_data, {
                 'title': ('title', {str}),
                 'description': (('leadParagraph', ('teaser', 'description')), any, {str}),