What are distance_weight id and source

[JOJ0]

In the importer: What is ment by source and id in an output like that:

(beets) jojo@gin ~/git/beets (skip_write) $ beet import ~/Music/Techno/015-spin_12-03/Oliver\ Deutschmann\ -\ Darkroom\ Tales\ \[Aim\ X01\]\ \(2012\)\ FLAC/

/Users/jojo/Music/Techno/015-spin_12-03/Oliver Deutschmann - Darkroom Tales [Aim X01] (2012) FLAC (2 items)
Tagging:
    Oliver Deutschmann - Darkroom Tales
URL:
    https://www.discogs.com/release/3424470-Oliver-Deutschmann-Darkroom-Tales
(Similarity: 72.4%) (id, source) (Discogs, Vinyl, 2012, Germany, Aim (3), Aim X01)
 * Darkroom Tales feat. Loganic FLAC - Bm - 62    -> Darkroom Tales
 * Darkroom Tales (Tristen Edit) FLAC - Ebm - 123 -> Darkroom Tales (Tristen Edit)
[A]pply, More candidates, Skip, Use as-is, as Tracks, Group albums,
Enter search, enter Id, aBort, plaY, eDit, edit Candidates?

• Is source the tagging source? Eg. MusicBrainz, Discogs, ...
• Is id the track id's not matching Eg. A1 vs 1, 1 vs. 2, etc ? Or is it the MusicBrainz Release ID or the Discogs Release ID?

[JOJ0]

To give some context what I am aiming for. I am playing around with distance_weights. I want to get higher scores, closer to strong_rec_thresh, I'd like to ignore things I don't consider important like country and media the entry is from. I prefer MusicBrainz over Discogs as a source (source_weights there are configured accordingly). This is my current match config:

match:
    strong_rec_thresh: 0.12
    medium_rec_thresh: 0.50
    rec_gap_thresh: 0.2
    max_rec:
        missing_tracks: strong
        unmatched_tracks: medium
     distance_weights:
        source: 0.0  # Default 2.0
        country: 0.0  # Default 0.5
        media: 0.0  # Default 1.0
        #mediums: 0.0  # Default 1.0
        track_title: 0.0  # Default 3.0
        album: 4.0  # Default 3.0
        catalognum: 4.0  # Default 0.5
        track_id: 2.0 # Default 5.0

Also there is one major problem in this collection: I used a third party tool that manipulates the track title, thus it is never 100% correct (adds - Key - BPM), also sometimes I wrote in stuff into the title field myself. Like FLAC in above's example.

Question on this regard: How exactly does the title matching work? Is it comparing 1:1 or does it any fuzzy regexing thing? A link to code where this happens would be interesting.

What is the mediums setting? I understand that media is wether the source provides the info of the same release from a vinyl or CD or Digital Media. What is mediums then?.........Ah now that I write it: It might be the count of the mediums the record was put out. E.g 2xCD, 3xLP, is that right?

[arsaboo]

So, I tested it out with one album and here's the output:

WITH track_id: 0.0

spotify: Searching Spotify for 'album:Ras artist:Snatam Kaur'
spotify: Found 1 result(s) from Spotify for 'album:Ras artist:Snatam Kaur'
spotify: Searching Spotify for album '4HH4W6sXqpxxOJAUarPuL6'
Sending event: albuminfo_received
Candidate: Snatam Kaur - Ras (4HH4W6sXqpxxOJAUarPuL6)
Computing track assignment...
...done.
Success. Distance: 0.18

WITH track_id: 5.0

spotify: Searching Spotify for 'album:Ras artist:Snatam Kaur'
spotify: Found 1 result(s) from Spotify for 'album:Ras artist:Snatam Kaur'
spotify: Searching Spotify for album '4HH4W6sXqpxxOJAUarPuL6'
Sending event: albuminfo_received
Candidate: Snatam Kaur - Ras (4HH4W6sXqpxxOJAUarPuL6)
Computing track assignment...
...done.
Success. Distance: 0.35

So, we should have lower weights when we want to de-emphasize the attribute. If you want to ignore certain attributes, make them 0.

[JOJ0]

I think one of my questions back then was: What is the track_id actually? Is it the title of a track ooor is it the position of the track on the record. Do tou know? And is it the same as id, as shown in the importer screenshot. That was confusing to me.

[arsaboo]

track_id is the track position (order in which tracks are ordered)

[sampsyo]

In fact, it's the penalty applied for mismatches in the MusicBrainz recording ID:

beets/beets/autotag/match.py

Lines 151 to 152 in 50bd693

	if item.mb_trackid:
	dist.add_expr('track_id', item.mb_trackid != track_info.track_id)

[JOJ0]

Since the answer is hidden in a thread above that cant be marked as answered, I'll post it again here and mark it.

#4297 (reply in thread)

source is the tagging source (Discogs, MusicBrainz, ...)

track_id is the recording id (eg mb_trackid) and is not the track position (eg 1, 2, A1, B2, ...)