>Thanks, I'll take a closer look at this query later.  The reason it's done 
>per-track now is so the code can be the same for in-process and external 
>scanning.  In-process does not have phases, so there is no "merge" phase, 
>everything runs at track time to ensure a consistent state after each track.
>
Ah, I see.  Still, it could probably run the same query after every track 
in-process mode, but passing an album id clause in to avoid recalculating all 
non-compilation albums.

Is there an issue that if tags are updated such that an album isn't a 
compilation any more, it isn't refreshed?  i.e. it only considers albums where 
compilation IS NULL.  i.e. if someone fixes a contributor name because it 
caused the album to be a compilation, then it won't set COMPILATION back to 
NULL.  How does that case get handled?

>The merge VA code has always only used role=1, it does not care about the 
>other types of artists or any prefs.  Your Mozart case would be considered a 
>compilation, and you'd want to fix the tags in that case anyway.
>
Yeah, bad example ;-)

I was really wondering if there were a different composer, conductor or 
band/orchestra on each song, whether that qualified as a compilation.  I don't 
really have much classical music, or use varying Band values across songs on an 
album, so I guess it's not normal, and if necessary, people can tag the album 
manually as a compilation anyway.  So, role=1 is fine ;-)
_______________________________________________
beta mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/beta

Reply via email to