Work language problem


#1

Hi, need some help to investigate a problem.

Recently I found that Picard tags some of recording as having mul language. My original thought it was a bug in Picard, but it seems it is not — the problem is in data returned by MB. Let us consider two songs:

Песня о Ялте, and Проснись и пой from the same album Джентельмены удачи.

Songs data looks very similar:

Песня о Ялте:

<metadata>
  <work type="Song" id="ab13db9c-9503-4100-854a-97b8123142fc" type-id="f061270a-2fd6-32f1-a641-f0f8676d14e6">
    <title>Песня о Ялте</title>
    <language>rus</language>
    <language-list>
      <language>rus</language>
    </language-list>
  </work>
</metadata>

and

Проснись и пой:

<metadata>
  <work type="Song" type-id="f061270a-2fd6-32f1-a641-f0f8676d14e6" id="6ce9159a-530c-4d90-be8f-3890671b21da">
    <title>Проснись и пой</title>
    <language>rus</language>
    <language-list>
      <language>rus</language>
    </language-list>
  </work>
</metadata>

Pay attention to language and language-list elements: they are just identical in both works. In MB web interface I also do not see difference: both works are of song type and lyrics languages are Russian in both works too.

However, release request returns (complete output is too long, I copied just interesting parts):

...
<track id="a4b48047-b15f-4f77-8e77-c40db9ded9a7">
  <position>5</position>
  <number>5</number>
  <length>77000</length>
  <recording id="5b77aacb-1a1d-421a-a24a-45698718dbbc">
    <title>Песня о Ялте</title>
    <length>77000</length>
    <relation-list target-type="work">
      <relation type="performance" type-id="a3005666-a872-32c3-ad06-98af558e99b0">
        <target>ab13db9c-9503-4100-854a-97b8123142fc</target>
        <work id="ab13db9c-9503-4100-854a-97b8123142fc">
          <title>Песня о Ялте</title>
          <language>rus</language>
          <language-list>
            <language>rus</language>
          </language-list>
...
<track id="62b4f6e7-3d7f-482a-bd1f-d822bf6133c5">
  <position>18</position>
  <number>18</number>
  <length>152000</length>
  <recording id="6ed760d2-84f8-4985-9cbe-051b7bc1cf85">
    <title>Проснись и пой</title>
    <length>152000</length>
    <relation-list target-type="work">
      <relation type="performance" type-id="a3005666-a872-32c3-ad06-98af558e99b0">
        <target>6ce9159a-530c-4d90-be8f-3890671b21da</target>
        <work id="6ce9159a-530c-4d90-be8f-3890671b21da">
          <title>Проснись и пой</title>
          <language>mul</language>
          <language-list>
            <language>rus</language>
            <language>rus</language>
          </language-list>
...

Work of track 5 looks good (and track is tagged by Picard correctly), but work of track 18 has language-list of two rus languages, and so, mul language. I do not understand why…


#2

I’ve noticed this in some of my own (English) tracks as well. Don’t have them with me at the moment, but I can find them later if it will help. It definitely seems like that depends on the work, as the false mul seems both limited to certain tracks and consistently returned on those tracks.


#3

Looking at Work 6ce9159a-530c-4d90-be8f-3890671b21da, it only has Russian listed as language once (as expected). This seems like a bug, so I filed one:
https://tickets.metabrainz.org/browse/MBS-9447