extra_tags for Discogs-plugin, and extended autotagger candidate information

[sindreruud]

Hi!

So this is a two-parter.

Musicbrainz got a new configuration option fairly recently, where you can specify extra_tags that it will use to match your music. For me this is really helpful because I often have to manually find the realease with the correct catalognum and media. So after adding:

extra_tags: [media, year, label, catalognum, country] 

I've seen a huge improvement in accuracy on the few albums I have imported so far.

Unfortunately for me, well over 50% of my albums are not on musicbrainz, but 99% of the time they are on Discogs. It would be amazing to have this option in the Discogs-plugin aswell. Took a quick look in discogs.py and found this

def candidates(self, items, artist, album, va_likely, extra_tags=None):

Would this be a good place to start if I want to look into this myself?

The output from the autotagger is sometimes a bit ambiguous when several discogs release look more or less the same (white label vs test pressing vs actual release for instance). Here is in example that actually illustrates the thing I want to solve in both 1. and 2. The autotagger gives equal match to two bandcamp results, aswell as two discogs results.

/incoming/music-beets/[12ROB49] Red Seal - Spinning Top [FLAC-VINYL] [1996] (3 items)
originquery: Using origin file /incoming/music-beets/[12ROB49] Red Seal - Spinning Top [FLAC-VINYL] [1996]/origin.yaml
originquery: ╔════════════════╤═════════════╤══════════════╗
originquery: ║ Field          │ Tagged Data │ Origin Data  ║
originquery: ╟────────────────┼─────────────┼──────────────╢
originquery: ║ Media          │             │ Vinyl        ║
originquery: ║ Edition year   │ 1996        │ 1996         ║
originquery: ║ Record label   │             │ Robs Records ║
originquery: ║ Catalog number │             │ 12ROB49      ║
originquery: ╚════════════════╧═════════════╧══════════════╝
Correcting tags from:
    Red Seal - Spinning Top
To:
    Red Seal - Spinning Top EP
URL:
    https://mintconditionrecordslondon.bandcamp.com/album/spinning-top-ep
(Similarity: 89.2%) (year, label) (bandcamp, Digital Media, 2023, GB, Mint Condition)
 * Spinning Top (#0) -> Spinning Top (#1)
 * Twister (#0)      -> Twister (#2)
 * Hula Loop (#0)    -> Hula Loop (#3)
[A]pply, More candidates, Skip, Use as-is, as Tracks, Group albums,
Enter search, enter Id, aBort, eDit, edit Candidates? M
Finding tags for album "Red Seal - Spinning Top".
Candidates:
1. Red Seal - Spinning Top EP (89.2%) (year, label) (bandcamp, Digital Media, 2023, GB, Mint Condition)
2. Red Seal - Spinning Top EP (89.2%) (year, label) (bandcamp, Vinyl, 2023, GB, Mint Condition)
3. Red Seal - Spinning Top (86.2%) (tracks, source) (Discogs, Vinyl, 1996, UK, Robs Records, 12ROB49)
4. Red Seal - Spinning Top (86.2%) (tracks, source) (Discogs, Vinyl, 1996, UK, Robs Records, 12ROB49)
5. Red Seal - Spinning Top  (79.3%) (tracks, source, catalognum, ...) (Discogs, Vinyl, 2023, UK, Mint Condition (2), MC061)
6. Spinning Top - Spinning Top (37.5%) (tracks, artist, missing tracks, ...) (bandcamp, Digital Media, 2017, US, Light Gray Records)
7. Boston Symphony Orchestra ... Charles Munch - The French Touch (23.8%) (tracks, missing tracks, artist, ...) (Discogs, Vinyl, 1959, US, RCA Victor Red Seal, LSC-2292)
8. Michala Petri , The London Philharmonic Orchestra - Scandinavian Moods (18.1%) (missing tracks, tracks, artist, ...) (Discogs, CD, 1999, Europe, RCA Red Seal, 743211679502)
# selection (default 1), Skip, Use as-is, as Tracks, Group albums,
Enter search, enter Id, aBort, eDit, edit Candidates?

In this scenario I know that I want to pick one of the Discogs candidates because they in my opinion have better metadata. But candidate 4 and 5 does not appear different. But the difference is:
Candidate 3: [r74882] (Vinyl, 12", 45 RPM)
Canditate 4: [r4383728] (Vinyl, 12", 45 RPM, White Label)

So i wonder how hard it would be to customize the output of this string
(Discogs, Vinyl, 1996, UK, Robs Records, 12ROB49)
so it uses the entire Format and looks something like this:
(Discogs, Vinyl, 12", 45 RPM, White Label, 1996, UK, Robs Records, 12ROB49)

Happy to look into this myself, but would be very grateful if someone could nudge me in the right direction as this is a bit outside my comfort-zone.

[arsaboo]

Yes, you would start there and modify the get_albums function to use the extra tags. You may want to check whether the underlying python3-discogs-client library can use extra tags. Looks like it can accept a few additional tags

https://www.discogs.com/developers/#page:database,header:database-search

[JOJ0]

Yeah go for it @sindreruud I'll review your PR. Will be a great addition to the Discogs plugin😍

[JOJ0]

And yes python3-discogs-client supports searching for everything the Discogs API allows: https://python3-discogs-client.readthedocs.io/en/latest/quickstart.html#searching

[sindreruud]

1,5 years and I'm back taking a stab at this 😁 Might not be the best place to ask, but I'll go for it anyway.

I have tested the discogs_client via cli and done queries like this

>>> import discogs_client
>>> d = discogs_client.Client('ExampleApplication/0.1', user_token="snip")
>>> results = d.search('Red Seal - Spinning Top', type='release', catno='12ROB49', year='1996')
>>> print(results.page(1))
[<Release 74882 'Red Seal - Spinning Top'>, <Release 4383728 'Red Seal - Spinning Top'>]

Looks quite good. But I am sure there are corner cases that needs to be tackled, like spaces, upper/lower case, etc.

Anyway. To get started, much like the autotagger does, i made a dict to map the fields

FIELDS_TO_DISCOGS_KEYS = {
    "catalognum": "catno",
    "country": "country",
    "label": "label",
    "barcode": "barcode",
    "media": "format",
    "year": "year",
}

And then I took a stab at changing the candidates()

def candidates(self, items, artist, album, va_likely, extra_tags: dict[str, Any] | None = None):
        """Returns a list of AlbumInfo objects for discogs search results
        matching an album and artist (if not various).
        """
        if not self.discogs_client:
            return

        if not album and not artist:
            self._log.debug(
                "Skipping Discogs query. Files missing album and "
                "artist tags."
            )
            return []

        if va_likely:
            query = album
        else:
            query = f"{artist} {album}"

        # Additional search cues from existing metadata.
        if extra_tags:
            tags = []
            for tag, value in extra_tags.items():
                key = FIELDS_TO_DISCOGS_KEYS[tag]
                if key and value:
                    tags.append(f"{key}='{value}'")

        # Append extra filters to the query
        if tags:
            query += ", " + ", ".join(tags)
        
        try:
            return self.get_albums(query)
        except DiscogsAPIError as e:
            self._log.debug("API Error: {0} (query: {1})", e, query)
            if e.status_code == 401:
                self.reset_auth()
                return self.candidates(items, artist, album, va_likely)
            else:
                return []
        except CONNECTION_ERRORS:
            self._log.debug("Connection error in album search", exc_info=True)
            return []

And now comes the stupid question, so please bear with me. But how do I go about testing this? Tried following the developer docs and the CONTRIBUTING.rst. So I have a virtual python environment with poetry, poethepoet and I did poetry install. But how do i go about actually building/installing (right terminology?) beets with my changes so I can start testing it?

If someone could nudge me in the right direction I would be very happy :)

[arsaboo]

I just install beets from the specific branch and test it out. If your branch is fix, you can use something like (please update the git URL):

pip install --upgrade --force-reinstall --no-deps git+https://github.com/arsaboo/beets.git@fix

[sindreruud]

Thanks, real simple solution! :)

[JOJ0]

Thanks, real simple solution! :)

There's an even simpler one. Go to your local repo clone directory and install like this:

pip install -e .

-e is editable mode. you can immediately run a beet something command that includes your uncommited code changes.