Re: global alignment tweak for ChordName

[Jean Abou Samra]

Le 21/12/2021 à 22:29, David Kastrup a écrit :
> Jean Abou Samra <jean@abou-samra.fr> writes:
>
> > [Valentin]
> > > Hello Jean, hello David, hello Kieren,
> > >
> > > you should even be able to write (if sten (ly:stencil-extent sten)),
> > > as the stencil should always be a stencil or #f.
> >
> > The two universally accepted values for any property
> > regardless of the predicate are #f and '(), so it
> > may still be '(), which is why checking with specific
> > predicates like this is often used in LilyPond's
> > source.
> '() is a bit of an ugly cookie since it for some reason has been chosen
> to be the default for unset properties (seems like a Lisp rather than
> Scheme idea to me) but shouldn't be used by users for stuff not formally
> fitting the predicate.
>
> In my handwavy impression of trying to sort-of guess unwritten coding
> conventions.


Hm. Reminds me of

\version "2.22.1"

{
  \once \override NoteHead.X-extent = #'()
  a'1
  \once \override NoteHead.X-extent = ##f
  a'1
}


Which happens because of

Grob::Grob (SCM basicprops)
{
  // [...]
  if (scm_is_null (get_property_data (this, "X-extent")))
    set_property (this, "X-extent", Grob::stencil_width_proc);
  if (scm_is_null (get_property_data (this, "Y-extent")))
    set_property (this, "Y-extent",
                  Unpure_pure_container::make_smob 
(Grob::stencil_height_proc,
Grob::pure_stencil_height_proc));
  if (scm_is_null (get_property_data (this, "vertical-skylines")))
    set_property (this, "vertical-skylines",
                  Unpure_pure_container::make_smob 
(Grob::simple_vertical_skylines_from_extents_proc,
Grob::pure_simple_vertical_skylines_from_extents_proc));
  if (scm_is_null (get_property_data (this, "horizontal-skylines")))
    set_property (this, "horizontal-skylines",
                  Unpure_pure_container::make_smob 
(Grob::simple_horizontal_skylines_from_extents_proc,
Grob::pure_simple_horizontal_skylines_from_extents_proc));
}

So I admit that my practice of setting properties
to #'() is not in line with the expectations of
part of the code base and I should take the habit
of writing ##f instead. Thanks for the notice.
But I wonder if we should update the comment in

bool
type_check_assignment (SCM sym, SCM val, SCM type_symbol)
{
  // [...]
  /*
    [...]
    TODO: should remove #f from allowed vals?
  */
  if (scm_is_null (val) || scm_is_false (val))
    return true;
  // [...]
}


Best,
Jean