In _resolv_genres wo do not have to explicitly check if self.c14n_branches is
not empty. The tree will have been loaded when self.canonicalize is truthy. Also
if self.c14n_branches would be empty the canonicalization is a no-op anyway...
This improves lastgenre's behaviour when the configuration option
`prefer_specific` is set but `canonical` is not.
Previously it would not set any tags. Now it does apply tags, sorted using the
canonicalization tree, but not canonicalized.
For this the default tree is loaded even when `canonical` is not set.
An extra check is added to only use it for canonicalization when `canonical` is
set.
MBID of recording could become invalid after merging. The existing
code always copies metadata from first track after updating. But for
albums with invalid track MBID that happens to be the first track,
MusicBrainz changes won't be applied to whole album, only whose
tracks with valid MBID. This is particularly annoying since those
changes are actually displayed for every `beet mbsync` run, but never
get applied.
Fix this issue by finding any track that get MusicBrainz updates, and
apply it to whole album.
Master releases roughly correspond to MusicBrainz' release groups.
It will be usefull to store master IDs, for example to retrieve
original release dates (#1122).
Discogs does not provide track IDs. As a workaround, invent our own
IDs by combining release ID with original track position strings
returned by Discogs (#2336).
Fix incorrect split of a tracklist by medium for the case of
two-sided mediums (#2887).
Following the discussion in #2887, the 'medium_total' value should
contain the number of tracks on the medium to which each particular
track belongs, not the total number of different mediums present on
a release.
Fix unit tests accordingly.
Converts bytes to unicode using util.text_string, assuming that the
string is a UTF-8 string.
If that fails, it falls back to a hardcoded fallback filename.
