[brightcove] Relax video tag embeds extraction

BrightcoveNewIE.extract_urls(): Handle player data when it is all in attributes of the <video> tag, and also the id is data-brightcove-video-id not data-video-id. Add test to generic extractor, note HEAD gives 404, no checksum computed(?).
7 years ago · 40158f55c9
parent 3dfceb286c
commit 40158f55c9
2 changed files with 92 additions and 20 deletions
--- a/youtube_dl/extractor/brightcove.py
+++ b/youtube_dl/extractor/brightcove.py
@ -17,6 +17,7 @@ from ..compat import (
 from ..utils import (
    determine_ext,
    ExtractorError,
+    extract_attributes,
    find_xpath_attr,
    fix_xml_ampersands,
    float_or_none,
@ -109,6 +110,7 @@ class BrightcoveLegacyIE(InfoExtractor):
                'upload_date': '20140827',
                'uploader_id': '710858724001',
            },
+            'skip': 'Video gone',
        },
        {
            # playlist with 'videoList'
@ -490,9 +492,10 @@ class BrightcoveNewIE(InfoExtractor):
    def _extract_urls(webpage):
        # Reference:
        # 1. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/publish-video.html#setvideoiniframe
-        # 2. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/publish-video.html#setvideousingjavascript
-        # 3. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/embed-in-page.html
-        # 4. https://support.brightcove.com/en/video-cloud/docs/dynamically-assigning-videos-player
+        # 2. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/publish-video.html#tag
+        # 3. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/publish-video.html#setvideousingjavascript
+        # 4. http://docs.brightcove.com/en/video-cloud/brightcove-player/guides/in-page-embed-player-implementation.html
+        # 5. https://support.brightcove.com/en/video-cloud/docs/dynamically-assigning-videos-player

        entries = []

@ -501,22 +504,39 @@ class BrightcoveNewIE(InfoExtractor):
                r'<iframe[^>]+src=(["\'])((?:https?:)?//players\.brightcove\.net/\d+/[^/]+/index\.html.+?)\1', webpage):
            entries.append(url if url.startswith('http') else 'http:' + url)

-        # Look for embed_in_page embeds [2]
-        for video_id, account_id, player_id, embed in re.findall(
-                # According to examples from [3] it's unclear whether video id
-                # may be optional and what to do when it is
-                # According to [4] data-video-id may be prefixed with ref:
-                r'''(?sx)
-                    <video[^>]+
-                        data-video-id=["\'](\d+|ref:[^"\']+)["\'][^>]*>.*?
-                    </video>.*?
-                    <script[^>]+
-                        src=["\'](?:https?:)?//players\.brightcove\.net/
-                        (\d+)/([^/]+)_([^/]+)/index(?:\.min)?\.js
-                ''', webpage):
-            entries.append(
-                'http://players.brightcove.net/%s/%s_%s/index.html?videoId=%s'
-                % (account_id, player_id, embed, video_id))
+        # Look for <video> tags [2] and embed_in_page embeds [3]
+        # [2] looks like:
+        # <video data-video-id="5320421710001" data-account="245991542" data-player="SJWAiyYWg" data-embed="default" class="video-js" controls itemscope itemtype="http://schema.org/VideoObject">
+
+        for video, script_tag, account_id, player_id, embed in re.findall(
+            r'''(?isx)
+                (<video[^>]+>)
+                (?:.*?
+                  (<script[^>]+
+                    src=["\'](?:https?:)?//players\.brightcove\.net/
+                    (\d+)/([^/]+)_([^/]+)/index(?:\.min)?\.js
+                  )
+                )?
+            ''', webpage
+        ):
+            attrs = extract_attributes(video)
+
+            # According to examples from [4] it's unclear whether video id
+            # may be optional and what to do when it is
+            video_id = attrs.get('data-video-id')
+            # See PR#12099/bostonglobe.py for 'data-brightcove-video-id' variant
+
+            if not account_id:
+                account_id = attrs.get('data-account')
+            if not player_id:
+                player_id = attrs.get('data-player')
+            if not embed:
+                embed = attrs.get('data-embed')
+
+            if video_id and account_id and player_id and embed:
+                entries.append(
+                    'http://players.brightcove.net/%s/%s_%s/index.html?videoId=%s'
+                    % (account_id, player_id, embed, video_id))

        return entries

--- a/youtube_dl/extractor/generic.py
+++ b/youtube_dl/extractor/generic.py
@ -449,6 +449,59 @@ class GenericIE(InfoExtractor):
                },
            }],
        },
+        {
+            # Brightcove with UUID in videoPlayer
+            'url': 'http://www8.hp.com/cn/zh/home.html',
+            'info_dict': {
+                'id': '5255815316001',
+                'ext': 'mp4',
+                'title': 'Sprocket Video - China',
+                'description': 'Sprocket Video - China',
+                'uploader': 'HP-Video Gallery',
+                'timestamp': 1482263210,
+                'upload_date': '20161220',
+                'uploader_id': '1107601872001',
+            },
+            'params': {
+                'skip_download': True,  # m3u8 download
+            },
+            'skip': 'video rotates...weekly?',
+        },
+        {
+            # Brightcove:new type [2].
+            'url': 'http://www.delawaresportszone.com/video-st-thomas-more-earns-first-trip-to-basketball-semis',
+            'md5': '2b35148fcf48da41c9fb4591650784f3',
+            'info_dict': {
+                'id': '5348741021001',
+                'ext': 'mp4',
+                'upload_date': '20170306',
+                'uploader_id': '4191638492001',
+                'timestamp': 1488769918,
+                'title': 'VIDEO:  St. Thomas More earns first trip to basketball semis',
+
+            },
+        },
+        {
+            # Alternative brightcove <video> attributes
+            'url': 'http://www.programme-tv.net/videos/extraits/81095-guillaume-canet-evoque-les-rumeurs-d-infidelite-de-marion-cotillard-avec-brad-pitt-dans-vivement-dimanche/',
+            'info_dict': {
+                'id': '81095-guillaume-canet-evoque-les-rumeurs-d-infidelite-de-marion-cotillard-avec-brad-pitt-dans-vivement-dimanche',
+                'title': "Guillaume Canet évoque les rumeurs d'infidélité de Marion Cotillard avec Brad Pitt dans Vivement Dimanche, Extraits : toutes les vidéos avec Télé-Loisirs",
+            },
+            'playlist': [{
+                'md5': '732d22ba3d33f2f3fc253c39f8f36523',
+                'info_dict': {
+                    'id': '5311302538001',
+                    'ext': 'mp4',
+                    'title': "Guillaume Canet évoque les rumeurs d'infidélité de Marion Cotillard avec Brad Pitt dans Vivement Dimanche",
+                    'description': "Guillaume Canet évoque les rumeurs d'infidélité de Marion Cotillard avec Brad Pitt dans Vivement Dimanche (France 2, 5 février 2017)",
+                    'timestamp': 1486321708,
+                    'upload_date': '20170205',
+                    'uploader_id': '800000640001',
+                },
+                'only_matching': True,
+            }],
+        },
        {
            # Brightcove with UUID in videoPlayer
            'url': 'http://www8.hp.com/cn/zh/home.html',
@ -1900,7 +1953,6 @@ class GenericIE(InfoExtractor):
        # Look for Brightcove Legacy Studio embeds
        bc_urls = BrightcoveLegacyIE._extract_brightcove_urls(webpage)
        if bc_urls:
-            self.to_screen('Brightcove video detected.')
            entries = [{
                '_type': 'url',
                'url': smuggle_url(bc_url, {'Referer': url}),