Re: Markup module patch (Issue 2026)

[David Kastrup]

Ian Hulin <address@hidden> writes:

> For example, there is a lot of discussion on guile-developers on how
> to support a feature which has been deprecated (local-eval), and it
> turns out it was useful to David K in his parser spring-clean, so I
> flagged it with Guile and it's now in their bug-tracking system. It
> looks like they're going to provide us with a solution, but the
> discussion on the thread is way over my head. . .

For the record: what they are currently discussing is _way_ more general
and exciting than the amount of headaches it is going to end up saving
us, because the peculiarities of Lilypond's syntax mean that the savings
we get in code complexity, robustness, and efficiency will not be all
that impressive when compared to the solution currently in place.

So at one point of time in the discussion, in particular when it turned
to "perhaps we can code some inefficient toy implementation for the sake
of Lilypond", I more or less took out Lilypond from the discussion and
argued for the sake of Guile itself, since we are talking about
something fitting a similar need as GCC plugins (which obviously _are_
asked for) while being much more logical to use.

And you can bet that _if_ they manage to integrate this sort of
functionality in Guilev2, I _will_ adapt our current code as soon as
possible.  Because our current solution is a good hack, but a good hack
does not beat a turnkey solution.

> My current woes with the markup facility are mainly on the LilyPond
> side, and how it's currently implemented, and how to get from "here"
> (Guile V1.8) to "there" (Guile V2.0). I don't think we need to involve
> the Guile guys on this bit yet.

Well, I took our problems from trying to get a multi-file project going
(those _are_ rather basic assumptions) without a usefully documented
interface of include-at-compilation-time as an example of why promoting
Guile as _the_ GNU extension language is not without problems, to a
private mailing list of GNU maintainers.  So expect Andy and possible
some others to not be oblivious of our problems.

I obviously don't agree with your assessment that it is all Lilypond's
fault for not getting it, so I might have a better opinion of your
skills and assessment (or a different opinion about just what kind of
skills should be required before you are allowed to touch Guilev2) than
you yourself.

If you, after the amount of work you put in, experience problems with
what amounts to basic integration tasks in Guile, most people will have
them.  It does not matter whether or not someone like Andy could solve
them in five minutes, since Andy is not included in the Guile
distribution.  So I really would recommend not being shy about asking
for advice (including "where is this documented") on their developer
lists timely whenever you get stuck.  I am sure they are by now in a
state of mind where they will greatly prefer that over having to deal
with me, and be appropriately responsive.

-- 
David Kastrup