Only remove Twitter image URLs, not 3rd party media URLs

This is important because Flickr and Instagram images (amongst
others) may have comments and other details on the page and the URL
may be integral to the content of the tweet, whereas Twitter images
are reliably at the end of the tweet.

We also now won't load 3rd party images when Twitter images exist
because this used to lead to double-images, and we now keep the URLs

src/MediaDownloader.c

@@ -416,6 +416,21 @@ is_media_candidate (const char *url)
 
 }
 
+gboolean
+is_twitter_media_candidate (const char *url)
+{
+  url = canonicalize_url (url);
+
+  return
+#ifdef VIDEO
+         g_str_has_prefix (url, "/photo/1/") ||
+         g_str_has_prefix (url, "video.twimg.com/ext_tw_video") ||
+#endif
+         g_str_has_prefix (url, "pbs.twimg.com/media/")
+   ;
+
+}
+
 static void
 cb_media_downloader_init (CbMediaDownloader *downloader)
 {

src/MediaDownloader.h

@@ -57,6 +57,7 @@ gboolean cb_media_downloader_load_finish (CbMediaDownloader  *downloader,
 
 gboolean is_media_candidate (const char *url);
 
+gboolean is_twitter_media_candidate (const char *url);
 
 G_END_DECLS
 

src/TextTransform.c

@@ -54,7 +54,7 @@ is_media_url (const char *url,
               const char *display_text,
               gsize       media_count)
 {
-  return (is_media_candidate (url != NULL ? url : display_text) && media_count == 1) ||
+  return (is_twitter_media_candidate (url != NULL ? url : display_text) && media_count == 1) ||
          g_str_has_prefix (display_text, "pic.twitter.com/");
 }
 

src/Types.c

@@ -342,14 +342,6 @@ cb_mini_tweet_parse_entities (CbMiniTweet *t,
       const char *expanded_url = json_object_get_string_member (url, "expanded_url");
       JsonArray *indices;
 
-      if (is_media_candidate (expanded_url))
-        {
-          t->medias[t->n_medias] = cb_media_new ();
-          t->medias[t->n_medias]->url = g_strdup (expanded_url);
-          t->medias[t->n_medias]->type = cb_media_type_from_url (expanded_url);
-          t->n_medias ++;
-        }
-
       indices = json_object_get_array_member (url, "indices");
       t->entities[url_index].from = json_array_get_int_element (indices, 0);
       t->entities[url_index].to   = json_array_get_int_element (indices, 1);
@@ -537,6 +529,23 @@ cb_mini_tweet_parse_entities (CbMiniTweet *t,
         }
     }
 
+  if (t->n_medias == 0)
+    {
+      for (i  = 0, p = json_array_get_length (urls); i < p; i ++)
+        {
+          JsonObject *url = json_node_get_object (json_array_get_element (urls, i));
+          const char *expanded_url = json_object_get_string_member (url, "expanded_url");
+
+          if (is_media_candidate (expanded_url))
+           {
+             t->medias[t->n_medias] = cb_media_new ();
+             t->medias[t->n_medias]->url = g_strdup (expanded_url);
+             t->medias[t->n_medias]->type = cb_media_type_from_url (expanded_url);
+             t->n_medias ++;
+           }
+        }
+    }
+
   t->n_entities = url_index;
 #if 0
   g_debug ("Wasted entities: %d", max_entities - t->n_entities);