Adrian Sampson

Huh; that's interesting! It would be super useful to have someone do the "detective work" to figure out how this ordering is determined. Checking for identical sets in the strings...

Aha, thanks for the extra info! I like the idea of a configuration option to sort alphabetically.

This is a great point; reusing the MBID fields is an abuse that can cause weird problems. We should instead set a flexible attribute. Alas, we currently don't have a...

Absolutely; the second approach is definitely the way to go! The AlbumInfo and TrackInfo objects should be able to carry along arbitrary field settings that are then applied in [apply_metadata](https://github.com/sampsyo/beets/blob/master/beets/autotag/__init__.py#L159)....

I would actually be in favor of abandoning all the hand-coded attributes and just moving over complete to an unstructured `dict`-like object. It could simplify changes in the future quite...

I'm still in favor but not thrilled about a new core dependency to support it. I like the idea of a plugin, though!

Good point—what's missing here, FWIW, is a dbcore type declaration for the `singleton` field. We don't yet have type declarations for the two built-in "computed" fields on `Item`: namely, `singleton`...

Sounds reasonable. FWIW, you can also try AcousticBrainz, which gets not only `bpm` but also many other acoustic measurements: https://beets.readthedocs.io/en/stable/plugins/acousticbrainz.html

Yes, but there is also `absubmit`, which actually does the analysis: https://beets.readthedocs.io/en/stable/plugins/absubmit.html

All this sounds totally reasonable, including the separate subsections in the config! I just also wanted to mention that it could even make sense to establish a second plugin, called...