add preference rules when comparing duplicate songs, favoring 1) bitrate 2)...

add preference rules when comparing duplicate songs, favoring 1) bitrate 2) duration 3) number of metadata tags

server/djrandom/fingerprint/dedup.py

@@ -136,22 +136,54 @@ class DeDuper(object):
         # Actually de-duplicate the songs we've found.
         self._resolve_dupes([x[0] for x in dupes])
 
+    def _get_song_score(self, mp3):
+        af = eyeD3.Mp3AudioFile(mp3.path)
+
+        # Get encoding parameters.
+        bitrate = af.getBitRate()[1]
+        duration = 30 * (int(af.getPlayTime()) / 30) # round to 30 secs
+
+        # Count metadata tags.
+        tag = af.getTag()
+        has_album = not (not tag.getAlbum())
+        has_artist = not (not tag.getArtist())
+        has_title = not (not tag.getTitle())
+        has_genre = not (not tag.getGenre())
+        has_year = not (not tag.getYear())
+        has_tracknum = (tag.getTrackNum()[0] is not None)
+        has_images = not (not tag.getImages())
+        num_meta = (4 * int(has_images) 
+                    + 2 * sum(map(int, (has_album, has_artist, has_title)))
+                    + sum(map(int, (has_genre, has_year, has_tracknum))))
+
+        return (bitrate, duration, num_meta)
+
     def _resolve_dupes(self, hashes):
         """Perform best duplicate selection and remove dupes from db."""
         log.debug('remove_dupes(%s)' % ','.join(hashes))
 
+        def _compare_score(a, b):
+            a_bitrate, a_duration, a_nmeta = a
+            b_bitrate, b_duration, b_nmeta = b
+            res = cmp(a_bitrate, b_bitrate)
+            if res == 0:
+                res = cmp(a_duration, b_duration)
+                if res == 0:
+                    res = cmp(a_nmeta, b_nmeta)
+            return res
+
         # Compute 'score' for each song and sort them.
         scores = []
         mp3s = MP3.query.filter(MP3.sha1 in hashes)
         for mp3 in mp3s:
             scores.append((self._get_song_score(mp3), mp3.sha1))
-        scores.sort(key=lambda x: x[0])
+        scores.sort(cmp=_compare_score, reverse=True)
         best_song = scores[0][1]
         log.debug('remove_dupes: best song is %s' % best_song)
 
         # Remove all the other songs.
         songs_to_remove = [x for x in hashes if x != best_song]
-        
+        log.info('remove_dupes: songs to remove: %s' % (','.join(songs_to_remove)))
 
 
 def run_deduper(db_url):