Inconsistent use of slow and fast 3/8

[malcolmsailor]

Among the analyses in 3/8, some are notated in "slow" 3/8 (3 beats to the measure), while others are notated in "fast" 3/8 (1 beat to the measure).

Slow 3/8 causes the music21 RomanText parser to raise an exception for b2 or b3, as it assumes fast 3/8 (see this issue). But in any case, a parser has no way of knowing whether slow or fast 3/8 is intended, so probably this should either be specified explicitly, or else a consistent choice made across files.

I haven't manually inspected all the files below, but I can attest that at least the following two files are inconsistent:

slow 3/8: Corpus/Quartets/Beethoven,_Ludwig_van/Op132/3/analysis.txt fast 3/8: Corpus/Quartets/Beethoven,_Ludwig_van/Op018_No4/2/analysis.txt

Files that may use slow 3/8 include (output of the command rg -l '3/8' . | rg 'analysis.txt' | xargs rg -l 'b[2-9]\s'):

./Piano_Sonatas/Beethoven,_Ludwig_van/Op026/1/analysis.txt
./Etudes_and_Preludes/Bach,_Johann_Sebastian/The_Well-Tempered_Clavier_I/03/analysis.txt
./OpenScore-LiederCorpus/Paradis,_Maria_Theresia_von/12_Lieder,_1786/05_Die_Tanne_(Sieh_Doris,_wie_vom_Mond_bestrahlt_)/analysis.txt
./OpenScore-LiederCorpus/Chaminade,_Cécile/_/Berceuse/analysis.txt
./OpenScore-LiederCorpus/Schumann,_Robert/Frauenliebe_und_Leben,_Op.42/3_Ich_kann’s_nicht_fassen/analysis.txt
./OpenScore-LiederCorpus/Reichardt,_Louise/Zwölf_Gesänge,_Op.3/04_Wachtelwacht/analysis.txt
./OpenScore-LiederCorpus/Schubert,_Franz/Op.59/3_Du_bist_die_Ruh/analysis.txt
./OpenScore-LiederCorpus/Coleridge-Taylor,_Samuel/_/Oh,_the_Summer/analysis.txt
./OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/09_Irrlicht/analysis.txt
./Quartets/Haydn,_Franz_Joseph/Op20_No1/3/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op018_No6/4/analysis.txt
./OpenScore-LiederCorpus/Schumann,_Robert/Dichterliebe,_Op.48/09_Das_ist_ein_Flöten_und_Geigen/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op132/3/analysis.txt

Files that may use fast 3/8 include (output of the command rg -l '3/8' . | rg 'analysis.txt' | xargs rg -l 'b1\.\d+\s'):

./Quartets/Haydn,_Franz_Joseph/Op20_No1/3/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op018_No4/2/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op130/4/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op132/3/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op018_No6/4/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op074/2/analysis.txt
./Quartets/Beethoven,_Ludwig_van/Op059_No1/2/analysis.txt

[MarkGotham]

Hi! Thanks for flagging this up. I'm happy to implement and / or accept a PR on necessary changes, though it sounds like this may need attention higher up (in music21) before any such action here. I'll reply to your issue there to that effect. Thanks again!

[MarkGotham]

Cancel that – no higher power needed here.

Romantext accepts slow 3/8 as a time signature string, so on this (corpus) end, it's simply a matter of specifying what's been used where. Am I right in thinking that this a safe and comprehensive solution:

• replace all cases of Time Signature: 3/8 with Time Signature: fast 3/8 (to clarify how they're currently read and flag this as a potential issue);
• in all cases that raise the error you mention above, change Time Signature: fast 3/8 to Time Signature: slow 3/8

I think that would reliably catch every case except for the unlikely eventuality of a slow 3/8 that never changes chord within the measure.

Are you happy to implement that? I'll gladly then check the PR.

Finally, are there any related issues? E.g. use of slow 6/8 or 9/8?

[malcolmsailor]

I can submit a pull request, yes.

I am not aware of related issues concerning slow 6/8 or 9/8. I only became aware of this issue because of the music 21 parser raising an exception.

[jacobtylerwalls]

E.g. use of slow 6/8 or 9/8?

In music21 these also default to fast, which I think is assumed in the examples from the Roman Text spec, and music21 also supports setting fast or slow to be explicit.

https://github.com/cuthbertLab/music21/blob/6255321e15473791a8f3f6e7983605aeb39af658/music21/meter/base.py#L538-L543

[malcolmsailor]

I just ripgrep'd the files for beats 3+ in 6/8 and beats 4+ in 9/8 and manually inspected the results. From these, it appears that there is no slow 6/8 or 9/8 in the corpus.