Re: problem with constants in scheme-text-spanner.ly

[David Nalesnik]

On Mon, Jul 22, 2013 at 12:06 PM, David Kastrup <address@hidden> wrote:

David Nalesnik <address@hidden> writes:

>  All spanners appear when I replace the variable definition
> (event-drul '(() . ()))
>
> with
>
> (cons (list) (list))
>
> and later (line 157):
>
> (set! event-drul '(() . ()))
>
> with
>
> (set! event-drul (cons (list) (list)))
>
> I'd like to fix the problem, but I have several questions:
>
> (1) Would (cons '() '()) be acceptable?  The problem is fixed, but I wonder
> about the
> usage of the literal _expression_ '() since there are multiple usages of
> set-car! and set-cdr! applied to the variable.

That's a misconception.  You can't apply set-car! or set-cdr! to '()
since '() has neither car nor cdr, not being a pair.

Sorry--I meant applying set-car! or set-cdr! to (cons '() '())

From your explanations below, I understand why this is OK.  It's really no different that applying set-car! to (cons 5 6)

 

> (2) Should '() in all variable definitions, set! expressions be changed to
> (list) ?

No, '() is just fine and safe.  It is a constant like 7, #t or #f and
can't be modified at all.  '(() ()) however is a pair.  Both its car and
cdr are the constant '() (which can't be changed), but you can replace
its car and cdr.  Which is undefined behavior: Scheme is free to throw
an error or actually change the code in every instance or do other
nasties.

So with (cons '() '()) you should be fine.

IIRC, completize-grob-entry running on all-grob-descriptions also
commits this kind of sacrilege.

Eek.  So that means that the variable all-grob-descriptions should be modified to some monstrosity using list and cons, or else the actions of completize-grob-entry and lines like

 (set! all-grob-descriptions (sort all-grob-descriptions alist<?))

at the bottom of define-grobs.scm

ought to be done manually or else there could be problems in guilev2?

Thanks,

David