I have an instance where SongKong is making an uncharacteristically big mess of a 14 CD box set classical release
The Master Pianist: The Complete EMI Recordings
It is attempting to add all of the conductors and orchestras who appear on various tracks to every track in the entire release! Other fields seem OK - it’s just CONDUCTOR and ORCHESTRA handling that appears messed up.
Thanks for support files but there is issue sometimes that the logs are cached and not written to file so the logs are empty. But I have just fixed this as part of SongKong 12.6 so please upgrade to SongKong 12.6, retry and resend support files.
I found another, simpler instance of the same thing, on an album with just two works. Three have a conductor and orchestra, but three should not. SK adds there fields for all six. Support files uploaded.
I think it is because the MusicBrainz release is linked to this Discogs release and this Discogs release has the conductor and orchestra applied to all tracks.
The way SongKong works is first it matches from MusicBrainz, then if there is a link in MusicBrainz or it independently finds matching Discogs release it will not overwrite fields with data but it will add data if Discogs has it where it is missing. So for this release I assume after matching to MusicBrainz there was no value for the Composer or Orchestra fields for tracks 4-6 and therefore because Discogs has these values for all tracks they are added to tracks 4-6.
And that seems to be the issue with the other release as well - https://www.discogs.com/release/6305780-Sviatoslav-Richter-The-Master-Pianist-The-Complete-EMI-Recordings
I think you can avoid/test this by deleting these fields with Metagrater task, then when run Fix Songs uncheck Update from Discogs and Search for a Discogs match, please confirm.
Yes, I have tried this and can confirm that with “Update from Discogs” unchecked, only the first three tracks (correctly) have a conductor and orchestera added.
If that’s the case, this could occur for just about any credited artist/performer on a mixed-works album - not just classical. Clearly the absence of data does not signify that the MB data is incomplete - it could be intentionally null.
In Discogs credits can be applied at track or release level, in these cases the credits have simply been applied incorrectly
It could be either, the data could be incomplete or it could be intentional. Unfortunately there is no way of knowing but we took the view that credits are usually applied to Discogs correctly, so empty fields would only be added to by Discogs if Discogs had valid credits
Maybe I need to finesse this, Im not sure of the situation with Discogs credits, but one idea would be if MusicBrainz has added a role (such as Conductor) to some tracks then we can assume it is done for all tracks and therefore should not allow Discogs to add if has data for tracks where MusicBrainz has not added, whereas if MusicBrainz has not added a role to any tracks then if Discogs has data should allow it?
Thank you for looking into this. Of course the preferred solution to bad data is to fix the data.
I have seen plenty cases in MB of albums (especially compilations) with inconsistently complete data across tracks. But I don’t have a real sense (apart from the infrequency in my collection) of the prevalence of discogs album vs track attribution errors. It is hard to guess which is more common. However the “principle of least surprise” suggests that missing a credit would be more understandable and less puzzling than adding an erroneous one (or many).
