Same figure in rntxt file differs from instantiated RomanNumeral object

[napulen]

Minimal reproducibility code:

In [1]: import music21

In [2]: music21.__version__
Out[2]: '7.1.0'

In [3]: rntxt = music21.converter.parse("text.rntxt")

In [4]: for rn in rntxt.flat.notes:
   ...:     print(rn, rn.pitchClasses)
   ...:
<music21.roman.RomanNumeral #viø4/3 in f minor> [8, 0, 2, 5]

In [5]: rn2 = music21.roman.RomanNumeral("#viø4/3", "f")

In [6]: rn2.pitchClasses
Out[6]: [9, 1, 3, 6]

In [7]: !cat text.rntxt
m1 f: #vi/o4/3

In [8]: rn2 = music21.roman.RomanNumeral("#vi/o4/3", "f")

In [9]: rn2.pitchClasses
Out[9]: [9, 1, 3, 6]

The last lines are just to confirm that this doesn't seem to be related to the /o notation for half-diminished.

The annotation shows up in several instances of the When-in-Rome dataset (@MarkGotham might be interested). That's how I detected it.

[napulen]

So I just realized that the value of scaleDegreeWithAlteration is different.

In [12]: rn.scaleDegreeWithAlteration
Out[12]: (6, <music21.pitch.Accidental sharp>)

In [13]: rn2.scaleDegreeWithAlteration
Out[13]: (6, <music21.pitch.Accidental double-sharp>)

So I guess whenever someone wants to parse an annotation in the same way as the RomanText parser would, the right line is?

rn2 = music21.roman.RomanNumeral("#viø4/3", "f", sixthMinor=music21.roman.Minor67Default.FLAT)

[napulen]

I can confirm that I've tried several combinations of sixthMinor and seventhMinor args, but nothing seems to mimic the same interpretation of a Roman numeral that the RomanText converter does.

The only thing that has worked for me, if one wanted to mimic the same behavior is something like this:

key = rn.key.tonicPitchNameWithCase
figure = rn.figure
s = music21.converter.parseData(f"m1 {key.replace('-', 'b')}: {figure}", format="romantext")
rn = [rn for rn in s.flat.notes][0]

That won't reproduce the offset location, of course.

[jacobtylerwalls]

Hi @napulen thanks for the report. I haven't looked in detail at your second and third messages, but from the first one, this looks like another manifestation of #1217, don't you think? Looking at my solution there it's looking like that function shouldn't be editing rootAlterationString at all.

[jacobtylerwalls]

Ah, a more backwards compatible solution would be to just set CAUTIONARY for the sixthMinor and seventhMinor attributes on the roman numeral objects created by romanNumeralFromChord. And then you would be able to rely on those values when parsing figures.

e.g.

rn2 = RomanNumeral(otherRn.figure, 'f', sixthMinor=otherRn.sixthMinor, seventhMinor=otherRn.seventhMinor)

I can confirm that I've tried several combinations of sixthMinor and seventhMinor args

Have you tried CAUTIONARY?

>>> music21.roman.RomanNumeral("#viø4/3", "f", sixthMinor=roman.Minor67Default.CAUTIONARY).pitchClasses
[8, 0, 2, 5]

[napulen]

I did not try the CAUTIONARY. I'll run the full corpus with this and report back (same in PR).