Mensural Notation: **mens

kernpy reads and writes **mens spines, the Humdrum representation of white mensural notation (roughly 1450-1600), next to **kern spines in the same file.

import kernpy as kp

document, errors = kp.loads(
    "**mens\n*clefC1\n*met(C)\nSc\nsd\nMe\nmf\n*-\n")
print(kp.dumps(document))

One grammar, shared with mOOsicae

**mens cells are parsed with the combined **kern/**mens grammar of mOOsicae, the TypeScript library of the same ecosystem, vendored verbatim in kern/kernMensSpine*.g4. Both libraries therefore accept and reject the same **mens tokens. The grammar is never edited in kernpy: it is changed in mOOsicae and copied back (kern/README-mens.md). **kern spines keep their own grammar, kern/kernSpine*.g4, because in the combined one the mensural figures (M, m, S...) collide with **kern articulations and ornaments.

What a mensural note looks like

A note is its duration, then its pitch, then its marks:

Part	Signifiers	Example
Figure	X maxima, L longa, S breve, s semibreve, M minima, m semiminima, U fusa, u semifusa	Sc
Perfection	p perfect, i imperfect, I imperfect by alteration, right after the figure	Sic
Coloration	~, after the perfection	si~d
Dot	. augmentation, : division, last	sp:a
Ligature	< starts, > ends	<Sa, Sg>
Rest	the figure and r	sr

kernpy keeps the whole mensural duration (figure, perfection, coloration and dot) as one duration subtoken, as written, so the export never reorders it.

Tandem interpretations include mensurations (*met(O), *met(C|), with the optional *met(O)_2232 specification of maximodus, modus, tempus and prolatio), clefs, key signatures, *custosG (a custos with its pitch) and layout marks.

Export

The export is kernpy's normalised form, the same as for **kern:

• global comments (!!!) and hidden tokens such as invisible barlines (=1-) are not written; they still count measures;
• the marks of a note follow its pitch: <S~a is written S~a<, which the grammar reads as the same ligature start.

Exporting what kernpy imported and importing it again gives back the same document: the export is a fixed point. The test suite checks this, and that every note and rest survives, over the Willi Apel examples of mOOsicae and over two files of the SEILS dataset; set KERNPY_MENS_CORPUS to a directory to run the same checks over a whole corpus (all 180 **mens files of SEILS pass).

Dialects and errors

Like any token the grammar does not read, a non-standard **mens cell is reported in the errors list returned by kp.load/kp.loads (or raised with raise_on_errors=True) and kept verbatim in the document and in the export. SEILS, for instance, writes:

• a bare *custos: the grammar wants its pitch, *custosG;
• the perfection mark after the coloration, s~id: the grammar wants si~d. Read as it stands, the i would become a note decoration and the note would lose its imperfection, so kernpy reports it instead;
• the coloration after the dot, Mp:~.

Readers of such sources normalise them before parsing if they want those cells as notes.