Maintaining Different Music Formats

[ohitsdylan]

Hello,

I have a music library in FLAC that I am converting to OGG. No issues there with the convert plugin, things end up mostly how I want them.

What issue I am running into is maintaining a separate OGG library that has all non-ASCII characters removed from the paths and metadata. A device I am playing music on does not support Unicode characters.

I don't want to use the asciify function, as the unidecode Python package does not have the support I need for the Japanese titles. I understand there is some discussion about unihandecode and I was able to use that standalone to transliterate my metadata, but I've also located a tagging issue present in a FLAC file that I now need to propagate through my OGG files.

Is anyone else doing something similar to this? How are you accomplishing it without creating a second non-beets managed library? Or if it is not managed by beets, what are you using?

Thank you!

[aribradshaw]

I would keep the FLAC library as the only managed library and treat the OGG tree as a reproducible export. The convert plugin already has the two pieces needed for that model:

1. convert.paths can use a custom inline field for transliterated destination paths.
2. refresh: yes or beet convert -r rebuilds a destination file when its source is newer.

For example:

plugins: convert inline portabletags

item_fields:
  portable_artist: |
    from unihandecode import Unihandecoder
    return Unihandecoder().decode(albumartist or artist)
  portable_album: |
    from unihandecode import Unihandecoder
    return Unihandecoder().decode(album)
  portable_title: |
    from unihandecode import Unihandecoder
    return Unihandecoder().decode(title)

convert:
  dest: /path/to/ogg-library
  format: ogg
  refresh: yes
  paths:
    default: $portable_artist/$portable_album/$track $portable_title

That handles the filenames. To transliterate only the metadata in the converted copy, a small local plugin can listen for the convert plugin's after_convert event:

from beets.plugins import BeetsPlugin
from unihandecode import Unihandecoder


class PortableTagsPlugin(BeetsPlugin):
    def __init__(self):
        super().__init__()
        self.decoder = Unihandecoder()
        self.register_listener("after_convert", self.after_convert)

    def after_convert(self, item, dest, keepnew):
        fields = ("artist", "albumartist", "album", "title")
        tags = {
            field: self.decoder.decode(item.get(field) or "")
            for field in fields
        }
        item.write(path=dest, tags=tags)

Item.write(path=..., tags=...) writes the overrides to the destination file without changing the canonical item in the beets database. Add whichever other fields your player needs to fields.

For the FLAC tagging fix you mentioned, first make sure the corrected tags are written back to the source file, for example with beet write <query>. That updates the source mtime. The next beet convert -r <query> should then remove and regenerate the older OGG copy, including the portable path and metadata transformations.

Relevant implementation points are the convert plugin's --refresh behavior and after_convert event and Item.write(path, tags).

[ohitsdylan]

Wow, I was not expecting such a thorough answer, thank you!
I'll have to take some time to get this implemented, but thank you for your time with that.

[aribradshaw]

Glad to help! Lmk how it goes.

[ohitsdylan]

I think I may have to modify/update the Unihandecode package. It is complaining about missing the "pkg_resources" module and I remember having to do a bit of a hack to get it working standalone in a virtual environment.

But so far, so good! I was able to get that local plugin packaged.