Harmony doesn't split the artists sometimes

[arsinclair]

I noticed that sometimes when importing the releases of the combination of artists with Harmony, instead of splitting the artists it puts the whole combination into the artist alias field.

Let me illustrate this, I've been importing the beatport release with 853567159090 GTIN (https://harmony.pulsewidth.org.uk/release?url=&gtin=+853567159090&region=&musicbrainz=&deezer=&itunes=&spotify=&tidal=&beatport=) and this is what it had created on MB:

Unless manually checked, this could lead to incorrect data imported.

[kellnerd]

This is bad data from the iTunes API which only ever returns one artist. While the artistId seems to be from the primary artist, the artistName is the whole artist credit with join phrases.

Harmony uses the data from iTunes because it's the first source which returned data (MB and Deezer did not) and there is no default preference for the artist property.

Since automatic splitting of artist credits is error-prone, having a new feature "multiple artists" that can be used to mark iTunes artist data as low-quality would be my first choice. An artist credit splitter (with the option to disable it) can be added later, once a general decision has been made how all these dynamic, optional features should be implemented.

[arsinclair]

Yes, default splitting is error prone, and maybe we shouldn't even attempt to do that. But perhaps we can think about adding a warning of some kind in Harmony if it detects the following:

• an artist credit has any split phrases in it (e.g. &, with, and, feat, ft, vs, /, |, etc.) AND
• the artist credit from the providers doesn't match the linked artist name on MB

And obviously, if there's no artist linked to this provider, there's not much we can do.