From 0254f93b08ef5f8af8ea0f1b9f7558892626900b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sergey=20M=E2=80=A4?= <dstftw@gmail.com>
Date: Sat, 15 Apr 2017 20:30:51 +0700
Subject: [PATCH] [brightcove] Validate ambiguous embeds' URLs (closes #9163,
 closes #12005, closes #12178, closes #12480)

---
 youtube_dl/extractor/brightcove.py | 55 +++++++++++++++++-------------
 youtube_dl/extractor/generic.py    |  2 +-
 2 files changed, 33 insertions(+), 24 deletions(-)

diff --git a/youtube_dl/extractor/brightcove.py b/youtube_dl/extractor/brightcove.py
index c7059a12a..124497e95 100644
--- a/youtube_dl/extractor/brightcove.py
+++ b/youtube_dl/extractor/brightcove.py
@@ -489,7 +489,7 @@ def _extract_url(webpage):
         return urls[0] if urls else None
 
     @staticmethod
-    def _extract_urls(webpage):
+    def _extract_urls(ie, 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#tag
@@ -506,37 +506,46 @@ def _extract_urls(webpage):
 
         # 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
-        ):
+                r'''(?isx)
+                    (<video\s+[^>]+>)
+                    (?:.*?
+                        (<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 video_id:
+                continue
 
+            account_id = account_id or attrs.get('data-account')
             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')
+                continue
 
-            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))
+            player_id = player_id or attrs.get('data-player') or 'default'
+            embed = embed or attrs.get('data-embed') or 'default'
+
+            bc_url = 'http://players.brightcove.net/%s/%s_%s/index.html?videoId=%s' % (
+                account_id, player_id, embed, video_id)
+
+            # Some brightcove videos may be embedded with video tag only and
+            # without script tag or any mentioning of brightcove at all. Such
+            # embeds are considered ambiguous since they are matched based only
+            # on data-video-id and data-account attributes and in the wild may
+            # not be brightcove embeds at all. Let's check reconstructed
+            # brightcove URLs in case of such embeds and only process valid
+            # ones. By this we ensure there is indeed a brightcove embed.
+            if not script_tag and not ie._is_valid_url(
+                    bc_url, video_id, 'possible brightcove video'):
+                continue
+
+            entries.append(bc_url)
 
         return entries
 
diff --git a/youtube_dl/extractor/generic.py b/youtube_dl/extractor/generic.py
index dc60d77f5..bc7c21f7a 100644
--- a/youtube_dl/extractor/generic.py
+++ b/youtube_dl/extractor/generic.py
@@ -1967,7 +1967,7 @@ def _real_extract(self, url):
             }
 
         # Look for Brightcove New Studio embeds
-        bc_urls = BrightcoveNewIE._extract_urls(webpage)
+        bc_urls = BrightcoveNewIE._extract_urls(self, webpage)
         if bc_urls:
             return self.playlist_from_matches(bc_urls, video_id, video_title, ie='BrightcoveNew')