[webern]

AI Summary

A niente hairpin has a small circle at its point: a crescendo that grows out of nothing, or a diminuendo that fades away to nothing. MusicXML spells it <wedge niente="yes">, on the start of a crescendo but on the stop of a diminuendo, because that is where the point is. A read and write through mx::api dropped the attribute, so the circle was lost.

WedgeStart gains one field, niente (Bool, unspecified by default, part of WedgeStart equality). It is set on the start for either direction: yes draws the circle, no or unspecified leave a plain point. WedgeStop is unchanged, so a caller cannot put a circle on the wide end of a hairpin.

On write, a crescendo's niente goes on its start, and a diminuendo's goes on the stop that ends it, found with the same pairing the writer already uses to give an octave-shift stop the size of its start: within a part, by number, a stop ends the most recently opened start, across measures and staves. A start of unspecified type is written as a crescendo, as before, so its niente goes on the start. A diminuendo with niente and no stop has nowhere to put it; the value is left out and reported as dropped data (error severity), for no as well as yes.

On read, one pass over each part's wedges in document order pairs every stop with its start before the measures are read, so a diminuendo takes the niente of its stop even when the stop is in a later measure. Reader and writer pair the same way: the writer pairs in the order it writes, and a continue segment, which mx::api reads as a start of unspecified type and writes as a crescendo start, is paired as a start on both sides. niente on the wide end (a diminuendo's start, a crescendo's stop), on a continue segment, or on a stop that ends a continue is not read, and each is reported once as dropped data with the measure where it appears.

The pairing rule now lives in one small header shared by the reader, the octave-shift sizes and the wedge niente. The octave-shift code that moved onto it behaves exactly as before.

What existing callers see: scores that do not set niente write exactly what they wrote before. Reading a file now fills niente where the file states it, and a file with niente where MusicXML ignores it now produces a dropped-data diagnostic where it was silently dropped before. No file in the corpus outside the two pinned here uses niente. No public signature changes.

Out of scope: a continue segment is still read as a start of unspecified type and written back as a crescendo start, which this change does not fix. An absent number and number 1 still pair separately, as they already do for every spanner in mx, although MusicXML says an absent number means 1.

Pins two files in the api round-trip baseline:

• synthetic/wedge.3.0.xml
• synthetic/wedge.3.1.xml

Testing

• New tests in DirectionDataTest: niente on a crescendo with and without a stop; a start of unspecified type; a diminuendo whose stop is in a later measure; no on both shapes; unspecified writes nothing; overlapping numbered diminuendos; a reused number; a diminuendo with no stop reported once for yes and for no, with its location; niente read from a file on each end; a diminuendo, a continue and a stop; two unnumbered diminuendos open at once on different staves, built through the api and read from a file that lists the lower staff first; WedgeStart equality
• New tests in OttavaSizeApiTest: a nested unnumbered octave shift, and an unnumbered shift beside one numbered 1 on another staff
• make api-test: 7175 assertions in 861 test cases pass
• make api-roundtrip: 475 of 475 pinned files pass
• make api-roundtrip-discover: 476 PASS, 365 FAIL; the PASS list is the parent's 474 plus the two files pinned here
• make audit-test: 19 tests pass
• mx builds as a single-translation-unit unity build; make fmt-check passes

References

• Stacked on feat: keep the time a grace note steals or makes #502
• Builds on feat: derive octave-shift stop size from its start #415, which derived an octave-shift stop's size from its start, and feat: unified spanner numbering with writer-side assignment #320, which added writer-side spanner numbering

[webern]

I think this could imply an n instead of a circle, even if the spec calls it niente, so lets disambiguate here.

Suggested change

	Bool niente;
	Bool nienteCircle;

[webern]

🛑 I'm concerned about this PR. It's a lot of code churn in many files for something that doesn't seem to appear in the wild corpus at all. Holding on this one for further consideration.