Discussion/Proposal: Pitches always have accidentals

[mscuthbert]

Similar in vein to #2009 (octave always exists):

Music21 started out with the goal of being able to represent many different ways of thinking about music, globally and historically. One of the things that I wanted to be able to represent was pitch ambiguity.

In a lot of the world's music, especially what I'm most familiar with, Medieval European music, there's a notion of a neutral pitch, like some B, which might be expressed as B-natural or B-flat, but which one is not specified, and tradition and context will determine which to sing. (Tradition and norms is, for instance, why the default B in England became B-natural, but the default B in Germany became B-flat...)

This is why music21 has the concept of Pitch.accidental is None -- it does not mean "don't show a natural" (that would be Pitch.accidental = Accidental('natural', displayType='never') ). It was intended to means that no accidental (sharp/natural/flat) had been applied to this not at all.

What a laudable goal, representing all this way of thinking about music.

Has anyone ever actually done this with music21 (outside some demos I gave ca. 2008)? No. Probably not. Has everyone who ever worked with music21 ever since had to deal with if p.accidental is None or p.accidental.name = 'natural':... -- yes. And has everyone who works with this just always treated None as meaning natural? Yes, surely.

---

Proposal

Every Pitch object gets an accidental object attached to it -- natural if not specified -- and .accidental will always refer to an Accidental object.

Is there a cost to this?

You bet. Pitches with naturals will take 50% longer to create than they currently do:

In [42]: %timeit pitch.Pitch('C4')
512 ns ± 10.5 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [43]: %timeit pitch.Pitch('C#4')
750 ns ± 1.92 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

this cascades down to Note, as a 17% slowdown:

In [48]: %timeit note.Note('C4')
1.56 μs ± 23.4 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [49]: %timeit note.Note('C#4')
1.83 μs ± 6.06 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

If Accidental had been made immutable like Microtone, then we might be able to get the cost down to near nothing, but instead we can try to make the speed up in other ways.

Like we might stop allowing 'SHARP' instead of 'sharp' or 'FLAT' instead of 'flat' -- save a few ns -- I think that .set() can also be sped up immediately using a one-time created lookup table instead of a bunch of if statements.

Proposal Part 2:

Remove non-standard terminology for half-sharp ~ and half-flat ` -- no one knows what they are and I think better to be explicit ("-half-flat") than obscure.