remove missing-length distance penalty

Previously, the thinking went that if the MB database didn't have a length, we
should penalize the track maximally in order to avoid prioritizing a bad match
just because its information was incomplete. As it turns out, though, missing
track lengths are pretty common and this was causing more problems than it was
solving; this way, mysteriously high distances won't appear.

beets/autotag/match.py

@@ -211,10 +211,8 @@ def track_distance(item, track_info, track_index=None, incl_artist=False):
     dist, dist_max = 0.0, 0.0
 
     # Check track length.
-    if not track_info.length:
-        # If there's no length to check, assume the worst.
-        dist += TRACK_LENGTH_WEIGHT
-    else:
+    # If there's no length to check, apply no penalty.
+    if track_info.length:
         diff = abs(item.length - track_info.length)
         diff = max(diff - TRACK_LENGTH_GRACE, 0.0)
         diff = min(diff, TRACK_LENGTH_MAX)

docs/changelog.rst

@@ -25,6 +25,11 @@ This release focuses on making beets' path formatting vastly more powerful.
   directory and music file names.  See :doc:`/reference/config`.
 * Beets now ensures that files have **unique filenames** by appending a number
   to any filename that would otherwise conflict with an existing file.
+* The autotagging heuristics have been tweaked in situations where the
+  MusicBrainz database did not contain track lengths. Previously, beets
+  penalized matches where this was the case, leading to situations where
+  seemingly good matches would have poor similarity. This penalty has been
+  removed.
 * Fix an incompatibility in BPD with libmpc (the library that powers mpc and
   ncmpc).
 * Fix a crash when importing a partial match whose first track was missing.