Setting album_id during import results in corrupted beets db

[odoublewen]

Using the beet import --set option to set album_id results in corrupted beets db due to "orphaned" album record.

Cross references:

• I discovered this while trying to find a workaround to feature request Beet doesn't offer to merge albums when import run with --noautotag #5012.
• Before I suspected that this was a bug, I started discussion corrupt beets db: empy albums #5525 on this topic.

Problem

Here is a reproducible toy example:

Create two dummy mp3 files and add minimal id3 tags

$ ffmpeg -f lavfi -i anullsrc=r=44100:cl=mono -t 5 -q:a 9 -acodec libmp3lame foo.mp3
$ cp foo.mp3 bar.mp3
$ id3tag -s foo -a foobar -A foobar -t 1 foo.mp3 
$ id3tag -s bar -a foobar -A foobar -t 2 bar.mp3 

Start from a fresh, minimal beet 2.1.0 config:

$ rm ~/.config/beets/{state.pickle,library.db}
$ cat ~/.config/beets/config.yaml
directory: /home/odoublewen/mp3
import:
    copy: yes
    write: yes
    log: beetslog.txt
paths:
    default: $albumartist/$album/$track $title
$ beet --version
beets version 2.1.0
Python version 3.9.2
no plugins loaded

Import the dummy mp3 files

• Import the files one at a time.
• The second file is imported using --set to assign the album_id of the second mp3 file.

As shown in the beet ls command, it seemed to do the right thing:

$ beet import -AW foo.mp3
$ beet import -AW --set album_id=1 bar.mp3
$ beet ls -f 'album_id=$album_id $album $path'
album_id=1 foobar /home/odoublewen/mp3/foobar/foobar/01 foo.mp3
album_id=1 foobar /home/odoublewen/mp3/foobar/foobar/02 bar.mp3

But look at the weird results from beet ls -a:

$ beet ls -a -f '$album_id $album'
$album_id foobar
1 foobar

And beets raises a ValueError when I try beet ls -a -p:

$ beet ls -a -p
/home/odoublewen/mp3/foobar/foobar
Traceback (most recent call last):
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/util/functemplate.py", line 557, in substitute
    res = self.compiled(values, functions)
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/util/functemplate.py", line 581, in wrapper_func
    args[VARIABLE_PREFIX + varname] = values[varname]
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/dbcore/db.py", line 108, in __getitem__
    return self._get_formatted(self.model, key)
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/dbcore/db.py", line 131, in _get_formatted
    value = model._type(key).format(model.get(key))
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/dbcore/db.py", line 456, in _get
    return getters[key](self)
  File "/home/odoublewen/beets2/lib/python3.9/site-packages/beets/library.py", line 1409, in item_dir
    raise ValueError("empty album for album id %d" % self.id)
ValueError: empty album for album id 2

Examine beets database

Notice that beets created an album with id=2, but set the album_id attribute to 1. Maybe this is problematic because the album with album_id=2 is "orphaned" -- it has no items in it. I had hoped, or assumed, that using --set to set the album_id to an existing album would cause beet to not create a new album_id.

$ sqlite3 .config/beets/library.db
SQLite version 3.34.1 2021-01-20 14:10:07
Enter ".help" for usage hints.
sqlite> .headers on

sqlite> select id, album from albums;
id|album
1|foobar
2|foobar

sqlite> select * from album_attributes;
id|entity_id|key|value
1|2|album_id|1

But I also notice that beets did assign the bar.mp3 file to album_id=1 -- so that part seems good.

sqlite> select id, album_id, path from items;
id|album_id|path
1|1|/home/odoublewen/mp3/foobar/foobar/01 foo.mp3
2|1|/home/odoublewen/mp3/foobar/foobar/02 bar.mp3

Setup

• OS: debian linux
• Python version: 3.9.2
• beets version: 2.1.0
• Turning off plugins made problem go away (yes/no): no

My configuration (output of beet config) is:

directory: /home/odoublewen/mp3
import:
    copy: yes
    write: yes
    log: beetslog.txt
paths:
    default: $albumartist/$album/$track $title