Benign mismatch in MusicBrainz IDs causes poor similarity scores

[GoogleCodeExporter]

While running import -L few albums weren't processed while I added them with 
beets before. Here are some examples: http://pastebin.com/HMY9d1TW

Original issue reported on code.google.com by trash.xa...@gmail.com on 27 Jan 2013 at 3:47

[GoogleCodeExporter]

Can you clarify what's going wrong here? I don't see any "import -L" commands 
in this log. And you seem to be getting matches for everything.

Original comment by adrian.sampson on 27 Jan 2013 at 8:19

[GoogleCodeExporter]

I am trying to import stuff that wasn't automatically checked with "import -L" 
because the similarity rate was too low to be treated in quiet mode.

All these files are correctly tagged (I tagged them manually with beet), but 
their similarity rate is too low from my understanding. The similarity rate 
should be around 95% for these albums, but strangely it is not the case.

Original comment by trash.xa...@gmail.com on 27 Jan 2013 at 8:22

[GoogleCodeExporter]

Thanks for clarifying.

To investigate this any further, I'll need to run experiments on my machine. 
Can you upload any of these albums somewhere where I can get them and try 
reproducing the problem?

Also, which plugins do you have enabled? (Type "beet version" for a list.)

Original comment by adrian.sampson on 27 Jan 2013 at 8:24

[GoogleCodeExporter]

beets version 1.1b1-dev
plugins: embedart, lastgenre, fetchart, lyrics, replaygain

Original comment by trash.xa...@gmail.com on 27 Jan 2013 at 8:25

[GoogleCodeExporter]

Hey there -- I did some detective work on the files you sent me (thanks for 
supplying them!) and I've found the culprit.

The problem originates in the fact that you seem to have originally tagged 
these files before the MusicBrainz NGS transition, in which all recordings 
(previously tracks) had their IDs reassigned. So, for example, one of your 
files ("Picture This") has an ID3 tag for its MusicBrainz track ID that is 
86badeba-b06b-4347-ad4e-604d04cf4dc7. But this recording's official ID is 
14a01e05-5291-4d17-966d-41c2cd1954aa.

You can see at these two URLs that the two IDs now resolve, on the web service, 
to the same info:
http://musicbrainz.org/recording/86badeba-b06b-4347-ad4e-604d04cf4dc7
http://musicbrainz.org/recording/14a01e05-5291-4d17-966d-41c2cd1954aa

Since these IDs are, in fact, different, beets sees this as a mismatch. 
MusicBrainz ID discrepancies are weighted rather highly, so this impacts the 
score a lot.

Unfortunately, I can't see a way around this in beets without making lots and 
lots of calls to the MusicBrainz web service that would severely slow down the 
import process. I think the best solution at this point would be to remove 
those legacy MBID tags from your files before importing. Sorry that there isn't 
a cleaner solution here!

Original comment by adrian.sampson on 29 Jan 2013 at 7:54

• Changed title: Benign mismatch in MusicBrainz IDs causes poor similarity scores
• Changed state: WontFix