[yandexvideo] Add extractor
authorSergey M․ <dstftw@gmail.com>
Sun, 17 Mar 2019 10:33:07 +0000 (17:33 +0700)
committerSergey M․ <dstftw@gmail.com>
Sun, 17 Mar 2019 10:33:07 +0000 (17:33 +0700)
youtube_dl/extractor/extractors.py
youtube_dl/extractor/yandexvideo.py [new file with mode: 0644]

index f32f22a5af7832fdedbbda54d10aa5abd41c4263..8e7a5bf4163c0dc494a4be3ed3b0de8fd294617f 100644 (file)
@@ -1444,12 +1444,13 @@ from .yahoo import (
     YahooIE,
     YahooSearchIE,
 )
+from .yandexdisk import YandexDiskIE
 from .yandexmusic import (
     YandexMusicTrackIE,
     YandexMusicAlbumIE,
     YandexMusicPlaylistIE,
 )
-from .yandexdisk import YandexDiskIE
+from .yandexvideo import YandexVideoIE
 from .yapfiles import YapFilesIE
 from .yesjapan import YesJapanIE
 from .yinyuetai import YinYueTaiIE
diff --git a/youtube_dl/extractor/yandexvideo.py b/youtube_dl/extractor/yandexvideo.py
new file mode 100644 (file)
index 0000000..940c24a
--- /dev/null
@@ -0,0 +1,90 @@
+# coding: utf-8
+from __future__ import unicode_literals
+
+from .common import InfoExtractor
+from ..utils import (
+    int_or_none,
+    url_or_none,
+)
+
+
+class YandexVideoIE(InfoExtractor):
+    _VALID_URL = r'''(?x)
+                    https?://
+                        (?:
+                            yandex\.ru(?:/portal/(?:video|efir))?/?\?.*?stream_id=|
+                            frontend\.vh\.yandex\.ru/player/
+                        )
+                        (?P<id>[\da-f]+)
+                    '''
+    _TESTS = [{
+        'url': 'https://yandex.ru/portal/video?stream_id=4dbb262b4fe5cf15a215de4f34eee34d',
+        'md5': '33955d7ae052f15853dc41f35f17581c',
+        'info_dict': {
+            'id': '4dbb262b4fe5cf15a215de4f34eee34d',
+            'ext': 'mp4',
+            'title': 'В Нью-Йорке баржи и теплоход оторвались от причала и расплылись по Гудзону',
+            'description': '',
+            'thumbnail': r're:^https?://.*\.jpg$',
+            'timestamp': 0,
+            'duration': 30,
+            'age_limit': 18,
+        },
+    }, {
+        'url': 'https://yandex.ru/portal/efir?stream_id=4dbb36ec4e0526d58f9f2dc8f0ecf374&from=morda',
+        'only_matching': True,
+    }, {
+        'url': 'https://yandex.ru/?stream_id=4dbb262b4fe5cf15a215de4f34eee34d',
+        'only_matching': True,
+    }, {
+        'url': 'https://frontend.vh.yandex.ru/player/4dbb262b4fe5cf15a215de4f34eee34d?from=morda',
+        'only_matching': True,
+    }, {
+        # vod-episode, series episode
+        'url': 'https://yandex.ru/portal/video?stream_id=45b11db6e4b68797919c93751a938cee',
+        'only_matching': True,
+    }, {
+        # episode, sports
+        'url': 'https://yandex.ru/?stream_channel=1538487871&stream_id=4132a07f71fb0396be93d74b3477131d',
+        'only_matching': True,
+    }]
+
+    def _real_extract(self, url):
+        video_id = self._match_id(url)
+
+        content = self._download_json(
+            'https://frontend.vh.yandex.ru/v22/player/%s.json' % video_id,
+            video_id, query={
+                'stream_options': 'hires',
+                'disable_trackings': 1,
+            })['content']
+
+        m3u8_url = url_or_none(content.get('content_url')) or url_or_none(
+            content['streams'][0]['url'])
+        title = content.get('title') or content.get('computed_title')
+
+        formats = self._extract_m3u8_formats(
+            m3u8_url, video_id, 'mp4', entry_protocol='m3u8_native',
+            m3u8_id='hls')
+        self._sort_formats(formats)
+
+        description = content.get('description')
+        thumbnail = content.get('thumbnail')
+        timestamp = (int_or_none(content.get('release_date')) or
+                     int_or_none(content.get('release_date_ut')) or
+                     int_or_none(content.get('start_time')))
+        duration = int_or_none(content.get('duration'))
+        series = content.get('program_title')
+        age_limit = int_or_none(content.get('restriction_age'))
+
+        return {
+            'id': video_id,
+            'title': title,
+            'description': description,
+            'thumbnail': thumbnail,
+            'timestamp': timestamp,
+            'duration': duration,
+            'series': series,
+            'age_limit': age_limit,
+            'formats': formats,
+        }