Run a similarity report against all my music to a file (prior to any importing) ?

[nodecentral]

Hi

I’ve finished importing a large chunk of my music which scored above 95% on the similarity rating, but before importing the rest I’d like to run a similarity report against everything left to see what they scored and where I need to focus my attention on increasing their % (before running any import again ?)

[RollingStar]

You might get somewhere with

--pretend

--timid

and a few plugins: acoustid, fromfilename, discogs, and I think there's a barcode plugin as well.

You could import "asis" in quiet mode and then , I think, see what similarity score the imports got. Double check that before you go importing everything.

[nodecentral]

Many thanks for replying @RollingStar.

To perform a “dry run”, just use the -p (for “pretend”) flag. This will show you a list of files that would be moved but won’t actually change anything on disk. The -t option sets the timid mode which will ask again before really moving or copying the files.

I had briefly tried —pretend, but it didn’t seem to do anything different ?

What I’d love to see if a report run, that shows my album and then its closest match musicbrainz match (1st Candidate if multiple) with it’s % sent to a log file .. Something like this (mock up below) would be amazing !!

downloads/Bob Dylan - Random Collection/Bob Dylan's Greatest Hits V2 (10 items)
https://musicbrainz.org/release/7ffd002d-3969-4835-a101-3e4d6d4237b5
(Similarity 44.0%) (missing tracks, tracks, year, ...) (CD, 1999, US, Columbia, CK 65975)

/downloads/Jack Johnson - On and On (15 items)
https://musicbrainz.org/release/7ffd002d-3969-4835-a101-3e4d6d4237b5
(Similarity: 91.3%) (tracks, missing tracks) (CD, 2003, US, Brushfire Records, 440 075 012 2)

Any ideas of how to get close to this sort of thing?

[RollingStar]

I looked through the code and beet fields and couldn't find anywhere where beets saves the match %. Maybe one of the more advanced devs can elaborate. Here are some starting points https://github.com/beetbox/beets/search?q=strong_rec_thresh&type=code

Maybe if you did pretend, and piped the log to a file >mylog.txt in verbose mode -vv? If you can edit the python source code you might be able to write whatever info you need to the console, even in 'quiet' (no user interaction) mode.

[sampsyo]

It's true that we don't currently save the match distance anywhere. That's an interesting idea!

One thing you could try, @nodecentral, is combining the importer's quiet mode with adjusting strong_rec_thresh such that no matches get automatically applied. In quiet mode, by default, anything that's not an "obviously" correct match just gets skipped. So if you turn that latter setting up high enough, nothing will be considered "obvious" and beets will just show the match and skip everything.

[nodecentral]

Thanks @sampsyo - what’s the best way to run that and have the results outputted to a text file for me to work on separately ?

[sampsyo]

You could consider using shell redirection, like > mylog.txt 2>&1 or similar?