Providers assume all search results sane, fragile
hseg opened this issue · 3 comments
Trying to diagnose why #14 doesn't work, it appears that besides using recent
songs (more likely to be subject to licensing restrictions and thus not
available in certain providers), certain providers assume all tracks returned
have synced lyrics (or, more specifically, that always picking the first track
from a list of matches is correct). Offending lines are
https://github.com/txtsd/syncedlyrics/blob/9bd9ddcf92c01f7090927b42827aaa4c11710afe/syncedlyrics/providers/deezer.py#L49
https://github.com/txtsd/syncedlyrics/blob/9bd9ddcf92c01f7090927b42827aaa4c11710afe/syncedlyrics/providers/lrclib.py#L34
https://github.com/txtsd/syncedlyrics/blob/9bd9ddcf92c01f7090927b42827aaa4c11710afe/syncedlyrics/providers/musixmatch.py#L92
https://github.com/txtsd/syncedlyrics/blob/9bd9ddcf92c01f7090927b42827aaa4c11710afe/syncedlyrics/providers/megalobiz.py#L25
(noticed when testing against a query of ave maria -- it exists on lrclib, but the first match doesn't have synced lyrics)
Thanks for the notice. The CI problem is solved now since it was because of a bug in the NetEase provider which was failing.
About your concern, I suppose the provider's sorting system is enough, but in the case of wrong results could you give more examples? I could not find the mentioned song for evaluation (is it an artist or a track?)
https://en.wikipedia.org/wiki/Hail_Mary#Musical_settings
It's a prayer that's been covered by many classical composers, hence should usually have a good version in the database.
In this case, checked with curl https://lrclib.net/api/search?q='ave+maria' | jq '.[0] , which lists track 15003, however replacing the jq filter with map(select(.syncedLyrics)).[0] returns track 131735, which indeed does have synced lyrics.
Yes, you were right. I hope this commit solves the issue. Feel free to reopen in case of any problems.