Music library scanning bugs

[a515]

Server Version#: 1.40.0.7998
Player Version#: 4.123.2

Music library agent / scanning issue - inconsistent results but with consistent folder structure and tags.

When scanning compilation albums, the Plex Scanner Agent (which I wish was open sourced because I could at least help diagnose the problem and propose a fix) is randomly not reading the track number for tracks it correctly lists under the compilation album, but for the others where it does read the track number correctly, it places it under separate album entries or the track artist (not the album artist which it should) and the album text is truncated to 30 characters.

This is not constrained to one compilation - it happens for many.

I have also spotted a potential “title_sort” issue where the artist “30 Seconds To Mars” is truncated to “Seconds Of Summer” which will lead to incorrect sorting (see ID 38818 below). Plus the scanner is not always populating the “title” field in the DB - see DB output example below.

So for a compilation containing 20 tracks for example, I might have 4 in within a single album entry under the album artist “Various” with no track number and the other 16 tracks are under their own individual album entries for each track artist (so another 16 albums / icons) but with their correct track number.

Files are structured as recommended:

Album Artist\Album\Track.mp3
For compilations, Album Artist is “Various”.
Each file is tagged with ID3v2.3 with Album Artist (being “Various” for compilations) , Artist, Title, Track (as well as others such as Comments, Publisher, Year, etc…) but the software used has also written out ID3v1 tags as well.

I have attached two images that show:

1. the correct album entry with only 2 tracks in it and the incorrect track numbering. Looking at the SQLite DB (table “metadata_items”), the track number has a value of -1 for these which is not correct as the ID3v2 tags have the track number in them.

2. an example of a single track by track artist for the compilation with the information truncated at 30 characters (this should be under the compilation album in screenshot 1) but with the correct track number.

Is the scanner randomly picking ID3v1 tags over ID3v2.3 tags when it should be only using the highest version? Why for the correctly aligned album tracks are they not picking up their track numbers from the ID3 tags?

My horrifically time consuming and manual solution to this has been to up date the metadata_items table to do the following, but this cannot scale / continue and needs a fix:

1. update the “parent_id” of tracks in incorrect individual track artist + albums to the correct compilation album.
2. delete the offending record for the track artist + album with truncated 30 character info
3. update “index” of tracks without track number information aligned to the correct compilation album “parent_id”.

Output from “metadata_items”:

• id 38817 is the correct compilation album.
• ids 38818 and 38819 are the two tracks aligned to the correct instance of the compilation album (see screenshot 1) but with missing track number info (see -1 in the output - that is for the column “index”)
• id 38816 is an example of one track per track artist + album title entries with the correct track number (10) but incorrectly aligned to The View + BBC Radio 1’s Live Lounge Volu album (note the truncated album name).
• id 38815 is the incorrect track artist + album title entry.

38815|1|38814|9|local://38815|0|BBC Radio 1's Live Lounge Volu|BBC Radio 1's Live Lounge Volu|||||||||||1|||metadata://posters/09abf20b5cda9d1323132a0758aa9c8448278730|||||Acoustic|||||1167609600|||1711197454|2007|1711194685|1711194685|1711194689|||{"ld:loudnessAnalysisVersion":"2","pv:blurHashesChangedAt":"756438","pv:lyricFindSearch":"1711194689","pv:thumbBlurHash":"LGE{-N^-9aK3,^v#$wtk0oxUx]R-","url":"ld%3AloudnessAnalysisVersion=2&pv%3AblurHashesChangedAt=756438&pv%3AlyricFindSearch=1711194689&pv%3AthumbBlurHash=LGE%7B-N%5E-9aK3%2C%5Ev%23%24wtk0oxUx%5DR-"}|ecdc9206a7329ace62c8b2c039e264a5aa06c5c1||756466|756438|||

38816|1|38815|10|local://38816|1|Same Jeans (BBC Radio 1 Live L|Same Jeans (BBC Radio 1 Live L|||||||||||10|1|||||||Acoustic||||||||1711197454||1711194685|1711194685|1711194689|||{"url":""}|ea28e469b01536d8bc4cb86ea46c9a33142a4892||756467|756441|||

38817|1|18|9|local://38817|0|BBC Radio 1's Live Lounge Volume 2 (CD2)|BBC Radio 1's Live Lounge Volume 2 (CD2)|||||||||||1|||metadata://posters/09abf20b5cda9d1323132a0758aa9c8448278730|||||||||||||1711285763||1711194686|1711194686|1711194689|||{"ld:loudnessAnalysisVersion":"2","pv:blurHashesChangedAt":"756446","pv:lyricFindSearch":"1711285763","pv:thumbBlurHash":"LGE{-N^-9aK3,^v#$wtk0oxUx]R-","url":"ld%3AloudnessAnalysisVersion=2&pv%3AblurHashesChangedAt=756446&pv%3AlyricFindSearch=1711285763&pv%3AthumbBlurHash=LGE%7B-N%5E-9aK3%2C%5Ev%23%24wtk0oxUx%5DR-"}|b6f99d84cf9d94434019bfda8247f27990901dc1||756463|756446|||

38818|1|38817|10|local://38818|1||Seconds To Mars - Stronger (Originally By Kanye West)|||||||||||-1|1|||||||||||||||1711285763||1711194686|1711194686|1711194689|||{"url":""}|2b8b991267e1fe6404cb24b6839092ced9b51a73||756464|756448|||

38819|1|38817|10|local://38819|1||Streets - Never Went To Church (BBC Radio 1 Live Lounge version)|||||||||||-1|1|||||||||||||||1711285764||1711194686|1711194686|1711194689|||{"url":""}|a9c27836cab1631f315f4e4163d72126a8e5e677||756465|756450|||

[a515]

So due to a lack of response, I deleted the library and recreated it… with varying results.

It’s better than before but around 10% of the 4,200 albums I have are still screwed up similar to the problem above and re-creating the library also introduced a new problem of not leveraging the album art in about 20% of albums, requiring me to go into each album, edit the Posters to activate the album art that was clearly scraped from the tags but for some unknown reason, not activated (or selected in the UI).

Looking at the number of complaints on this forum about music meta data, I am hereby willing to donate my time to fix or create a scanner that will read the music tags and output to the Plex DB CORRECTLY, along with any code rights - I just need this to work. Will need the official docs on the data mode, how you generate the GUIDs and how to include your own scanner into Plex - can whip this up quite easily. I have done this myself for a personal media browser project so it isn’t rocket science.

Plex is really good at the streaming piece, has a performant UI and is better at Movies / TV meta data, but the Music meta data side utterly sucks and I would be personally ashamed if I wrote code that worked as well as it does.

P.S. I don’t expect to be taken up on my offer.

P.P.S. I have given Jellyfin a whirl… the interface is clean but nowhere near as performant as Plex and the SSL setup is something Plex does very well with its wildcard cert and setup process compared to the rather manual steps in Jellyfin… BUT it is getting to a point where I am going to put up with it as that actually indexes Music correctly and works.