Improve AniDB ID searching and matching

jellyfin / jellyfin-plugin-anidb

This plugin adds the metadata provider for aniDB to Jellyfin

GNU General Public License v2.0

31 stars 13 forks source link

Improve AniDB ID searching and matching #27

Closed nalsai closed 2 years ago

nalsai commented 2 years ago

I present to you ... almost an entire rewrite of the id searching and matching logic ... but not quite, just a lot of improvements!

Better searching:

the search logic is much improved to find more and more accurate results
all search results get returned, when using identify by name #16 (the year now also gets show there)

Better matching:

the difference between the title of the file and the tile of the result gets measured and the best one chosen, this makes sure that the closest title gets matched #20
added a title similarity threshold setting, which can be set to a low value, so, when in doubt, nothing gets matched, rather than matching the wrong thing #18

I fixed the formatting and removed old and unused code, in the files I worked on.

I plan to fix the episode metadata (and improve the series metadata) next, and clean up and format the rest of the files.

nalsai commented 2 years ago

I just found an issue: the series "Charlotte" (10997) gets picked up as "Wakakusa no Charlotte" (3486) because that show is also called "Charlotte" in French and Italian and it's earlier in the database. There's not much I can do about colliding titles. You can either name the Charlotte folder "Charlotte (2015)" or "Charlotte [anidbid-10997]", or you can use the identify function in Jellyfin, thanks to search now returning multiple results.

crobibero commented 2 years ago

I likely won't have time this week to review the PRs, but I have them on my todo list

nalsai commented 2 years ago

Sorry for that accidental merge.

h1dden-da3m0n commented 2 years ago

I just found an issue: the series "Charlotte" (10997) gets picked up as "Wakakusa no Charlotte" (3486) because that show is also called "Charlotte" in French and Italian and it's earlier in the database. There's not much I can do about colliding titles. You can either name the Charlotte folder "Charlotte (2015)" or "Charlotte [anidbid-10997]", or you can use the identify function in Jellyfin, thanks to search now returning multiple results.

If I am not totally mistaken, then this behaviour is to be expected and you can find the same with the other Metadata providers. I know for a fact that I had to specify the year on a series or two to get the proper metadata matched. But thank you for noting that :slightly_smiling_face: