configurable match thresholds (GC-290)

beets/autotag/__init__.py

@@ -26,7 +26,6 @@ from .hooks import AlbumInfo, TrackInfo, AlbumMatch, TrackMatch
 from .match import AutotagError
 from .match import tag_item, tag_album
 from .match import RECOMMEND_STRONG, RECOMMEND_MEDIUM, RECOMMEND_NONE
-from .match import STRONG_REC_THRESH, MEDIUM_REC_THRESH, REC_GAP_THRESH
 
 # Global logger.
 log = logging.getLogger('beets')

beets/autotag/match.py

@@ -75,10 +75,6 @@ SD_REPLACE = [
 RECOMMEND_STRONG = 'RECOMMEND_STRONG'
 RECOMMEND_MEDIUM = 'RECOMMEND_MEDIUM'
 RECOMMEND_NONE = 'RECOMMEND_NONE'
-# Thresholds for recommendations.
-STRONG_REC_THRESH = 0.04
-MEDIUM_REC_THRESH = 0.25
-REC_GAP_THRESH = 0.25
 
 # Artist signals that indicate "various artists". These are used at the
 # album level to determine whether a given release is likely a VA
@@ -335,16 +331,17 @@ def recommendation(results):
         rec = RECOMMEND_NONE
     else:
         min_dist = results[0].distance
-        if min_dist < STRONG_REC_THRESH:
+        if min_dist < config['match']['strong_rec_thresh'].as_number():
             # Strong recommendation level.
             rec = RECOMMEND_STRONG
         elif len(results) == 1:
             # Only a single candidate. Medium recommendation.
             rec = RECOMMEND_MEDIUM
-        elif min_dist <= MEDIUM_REC_THRESH:
+        elif min_dist <= config['match']['medium_rec_thresh'].as_number():
             # Medium recommendation level.
             rec = RECOMMEND_MEDIUM
-        elif results[1].distance - min_dist >= REC_GAP_THRESH:
+        elif results[1].distance - min_dist >= \
+                    config['match']['rec_gap_thresh'].as_number():
             # Gap between first two candidates is large.
             rec = RECOMMEND_MEDIUM
         else:

beets/config_default.yaml

@@ -48,3 +48,8 @@ musicbrainz:
     host: musicbrainz.org
     ratelimit: 1
     ratelimit_interval: 1.0
+
+match:
+    strong_rec_thresh: 0.04
+    medium_rec_thresh: 0.25
+    rec_gap_thresh: 0.25

beets/ui/commands.py

@@ -106,13 +106,13 @@ PARTIAL_MATCH_MESSAGE = u'(partial match!)'
 # Importer utilities and support.
 
 def dist_string(dist):
-    """Formats a distance (a float) as a similarity percentage string.
+    """Formats a distance (a float) as a colorized similarity percentage
-    The string is colorized if color is True.
+    string.
     """
     out = '%.1f%%' % ((1 - dist) * 100)
-    if dist <= autotag.STRONG_REC_THRESH:
+    if dist <= config['match']['strong_rec_thresh'].as_number():
         out = ui.colorize('green', out)
-    elif dist <= autotag.MEDIUM_REC_THRESH:
+    elif dist <= config['match']['medium_rec_thresh'].as_number():
         out = ui.colorize('yellow', out)
     else:
         out = ui.colorize('red', out)

docs/changelog.rst

@@ -26,6 +26,8 @@ It also adds some new features:
   encoding used to print messages to standard output.
 * The MusicBrainz hostname (and rate limiting) are now configurable. See
   :ref:`musicbrainz-config`.
+* You can now configure the similarity thresholds used to determine when the
+  autotagger automatically accepts a metadata match. See :ref:`match-config`.
 * :doc:`/plugins/lyrics`: Fix an issue that failed to find lyrics when metadata
   contained "real" apostrophes.
 

docs/reference/config.rst

@@ -268,7 +268,7 @@ no log is written. This can be overridden with the ``-l`` flag to
 .. _musicbrainz-config:
 
 MusicBrainz Options
-~~~~~~~~~~~~~~~~~~~
+-------------------
 
 If you run your own `MusicBrainz`_ server, you can instruct beets to use it
 instead of the main server. Use the ``host`` and ``ratelimit`` options under a
@@ -287,6 +287,34 @@ server---on this public server, you're `limited`_ to one request per second.
 .. _limited: http://musicbrainz.org/doc/XML_Web_Service/Rate_Limiting
 .. _MusicBrainz: http://musicbrainz.org/
 
+.. _match-config:
+
+Autotagger Matching Options
+---------------------------
+
+You can configure some aspects of the logic beets uses when automatically
+matching MusicBrainz results under the ``match:`` section. To control how
+*tolerant* the autotagger is of differences, use the ``strong_rec_thresh``
+option, which reflects the distance threshold below which beets will make a
+"strong recommendation" that the metadata be used. Strong recommendations
+are accepted automatically (except in "timid" mode), so you can use this to
+make beets ask your opinion more or less often.
+
+The threshold is a *distance* value between 0.0 and 1.0, so you can think of it
+as the opposite of a *similarity* value. For example, if you want to
+automatically accept any matches above 90% similarity, use::
+
+    match:
+        strong_rec_thresh: 0.10
+
+The default strong recommendation threshold is 0.04.
+
+The ``medium_rec_thresh`` and ``rec_gap_thresh`` options work similarly. When a
+match is above the *medium* recommendation threshold or the distance between it
+and the next-best match is above the *gap* threshold, the importer will suggest
+that match but not automatically confirm it. Otherwise, you'll see a list of
+options to choose from.
+
 .. _path-format-config:
 
 Path Format Configuration