Failed degradations keep copies of old notes
apmcleod opened this issue · 6 comments
Presumably this is a problem with the input data? For example, from PPDD?
For example: 00113f0b-d51e-4f14-9e89-eff29c257425.csv
00063cae-eeb4-4923-92e4-ee8b24a35968.csv is supposed to be time_shifted, but 2 notes have changed.
010a1298-f841-46c0-9dd8-204003a1d8ab.csv has both of the above problems.
003e3c87-79ab-4532-b200-291fbdf33d8e.csv has 2 onset_shifts...
0000f386-94de-4620-8172-577b8f9f83a8.csv has 4 of each note and 7 are degraded (offset_shift).
I'm guessing the multi-degradation excerpts have to do with the overlap check post-degradation somehow...
The multi-degraded outputs seem to be from forgetting to copy the excerpt before degrading it (since we may recurse up to 10 times if the resulting degradation is invalid).
I will split this into 2 issues. This one will remain as the multi-degradations. #122 will cover the duplicated notes.