## Description
- removes all mailing list references in doc.
- removes outdated installation instructions (python3.8 + slackware)
- modernizes the getting started guide
closes#5462
This PR refactors the metadata source plugin architecture to fix
incorrect data source penalty calculations during import matching.
**Key Changes:**
- **Fixed distance calculation**: Removed `track_distance()` and
`album_distance()` methods from `MetadataSourcePlugin`. Data source
mismatch penalty is now calculated directly in
`beets/autotag/distance.py`, same as most of the rest of field-based
penalties.
- **Plugin registration**: Metadata source plugins now register
themselves in `_instance_by_data_source` dict on instantiation, enabling
efficient penalty lookups.
- **Renamed configuration options**:
- `source_weight` → `data_source_mismatch_penalty` (plugin-level)
- `match.distance_weights.source` → `match.distance_weights.data_source`
(global matching)
- **Documentation**: added a `yaml` block with default configuration
values to each metadata source documentation page. Included the renamed
penalty field in each to underline that it's available for each data
source.
The data_source penalty was not being calculated correctly because
`_get_distance` was being called for **all** enabled metadata plugins
which eventually meant that matches were being penalised needlessly.
This commit refactors the distance calculation to:
- Remove the plugin-based track_distance() and album_distance() methods
that were applying penalties incorrectly
- Calculate data_source penalties directly in track_distance() and
distance() functions when sources don't match
- Use a centralized get_penalty() function to retrieve plugin-specific
penalty values via a registry with O(1) lookup
- Change default data_source_penalty from 0.0 to 0.5 to ensure
mismatches are penalized by default
- Add data_source to get_most_common_tags() to determine the likely
original source for comparison
This ensures that tracks and albums from different data sources are
properly penalized during matching, improving match quality and
preventing cross-source matches.
### b5216a0 is a proposed fix for #5218
We don't assume anymore that if the pattern lacks the "artist" group it
also has the "title" group (this was causing the "title" KeyError, when
the pattern containg only the `?P<track>` group is used).
Instead, we check for existence of "title" before assigning
"title_field", otherwise we let the last pattern (which will always
match) to assign it. The only slight drawback is that files like
"01.mp3" will also get "01" as title (besides the "1" track number),
instead of having no title, but I guess that is not going to cause any
major problem in the rest of the import procedure (e.g. track matching
after a search). On the other hand, I guess allowing a file to have no
title would require a substantial rewriting of the plugin, since the
empty string is in BAD_TITLE_PATTERNS.
### 0966042 adds a log message about the pattern being tried
This is useful while debugging changes to the regexps in PATTERNS.
I used loglevel "debug" for this message, should we use "debug" also for
other messages (currently using level "info")?
### e6b7735 refactors the regexps in PATTERNS
I reviewed the regexps (especially after the changes I made some time
ago in 84cf336) and tried to make them cleaner.
- Allow " - " and "-" as separator between track/artist/title fields:
that should cover two common filename conventions: using spaces (e.g.
`01 - Artist Name - Title of the song`) or not using spaces (e.g.
`01-artist_name-title_of_the_song`). Some regexp groups are non-greedy,
to avoid capturing unwanted leading/trailing spaces in artist/titles; we
could think about some other enhancements, e.g. automatically converting
underscores to spaces in artist/title names, but I guess the main use of
the plugin is not to produce a perfectly formatted title, but a
sufficient meaningful one to be used as reference in the following
stages of the update procedure.
- Allow `?P<tag>` groups by making them optional in the first two
patterns (instead of having two versions of the pattern, with or without
tag). BTW I'm not sure what these groups are about: we optionally check
for them in the plugin, but we never use them to assign any metadata.
- Allow optional "." after the track number
- In the third pattern: also allow "_" as separator (e.g. for file names
like `01_some_tack.mp3`)
Implements #3354, updates #6040. This PR also improves typehints for the discogs plugin.
The name variations provided by Discogs can now be used by the
auto-tagger through 3 config options.
By default, the plugin will write the variation to the tag
`artist_credit`, but through three config options can also write it
to the album artist tag, the track's artist tag, or any combination of
the three.
This PR contains a small addition on #6040, letting the string used to
join featured artists be customized.
The new configuration options available, and their defaults, are as
follows:
```yaml
discogs:
featured_string: "Feat."
anv:
artist_credit: True
album_artist: False
artist: False
```
