Rootless slash chords, 2017 edition

[Dmitry]

Hi,

There's been a long standing feature request:

https://sourceforge.net/p/testlilyissues/issues/3909/

#3909 a feature to disable the chord root name printing - just print
the slash and inversion

Over the years, it has been addressed more or less. Time to give it
another try?

In a few words, we need a mechanism to suppress printing repeated root
chord name over a progression of slash chords. That means, we'd like to
see this:

F#min9 /F /E /D#

instead of this:

F#min9 F#min9/F F#min9/E F#min9/D#

The latter is much less readable, occupies more space and bloats the
score.

To say this feature is demanded in jazz means say nothing! I've
recently started working on a bunch of jazz scores, and the very first
piece features three (!) different progressions with changing bass
notes, like the above. OK, there is a well-known hack that uses
\whiteout: http://lsr.di.unimi.it/LSR/Item?id=776

Unfortunately, for the real-life scores it is simply a no-go. It
requires a chord exception for each and every rootless slash chord, it
messes up MIDI output (as slash chords are not recognized in chord
exceptions), not to mention that it produces unclean output.

There is another workaround from 2011: http://www.mail-archive.com/lily
pond-user%40gnu.org/msg67087.html
Needless to say, it doesn't work in 2017 :) but I've managed to fix it.
Here we go,

#(define (rootless-chord-names in-pitches bass inversion context) 
   (ignatzek-chord-names `(,(ly:make-pitch 0 0 0) ,(ly:make-pitch 0 0
0)) bass inversion context))

#(define (empty-namer pitch lower?) (make-simple-markup ""))

retainChordNoteNamer =
\applyContext
  #(lambda (context)
     (let ((rn (ly:context-property context 'chordRootNamer)))
       (ly:context-set-property! context 'chordNoteNamer rn)))

rootless = {
  \retainChordNoteNamer
  \once \set chordNameFunction = #rootless-chord-names
  \once \set chordRootNamer = #empty-namer
}

After that, one can use \rootless as follows:

fis2:m7.9 \rootless fis4:m7.9/f \rootless fis4:m7.9/e \rootless
fis1:m7.9/dis

This produces perfect output, but the code is ugly as hell. What I'm
doing here is basically the following:
- kill off chordRootNamer, but retain chordNoteNamer;
- after that, the root note note wouldn't be printed, but the suffix
(like m9, etc.) would be. To suppress it, we supply a
proxy chordNameFunction that would guarantee a suffix-less chord.

It should be easy to convert this one-shot semantics to something like
\rootlessOn and \rootlessOff. In fact, it would be nice if mainstream
Lilypond had something similar to "\set chordChanges = ..." but for
slash chords; but of course nobody would like to see ugly hacks like
the above in the mainstream :)

Off the top of my head, I could propose the following solution:

- inside Scheme code for chord name functions ({ignatzek,banter,jazz}-
chord-names), allow for NIL "pitches" argument. That should mean
"produce output for bass note only, omitting everything before the
slash";
- in Chord_name_engraver::process_music, track repeating root parts of
slash chords and pass NIL pitches to a chordNameFunction in case of
repetition. Everything should be similar to handling chordChanges,
however this time we should remember and compare whole chord
structures, not markup.

What do you think?

Cheers,
Dmitry