[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] Fix the R6RS exact-integer-sqrt and import into core guile

From: Detlev Zundel
Subject: Re: [PATCH] Fix the R6RS exact-integer-sqrt and import into core guile
Date: Wed, 13 Apr 2011 17:52:08 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.0.50 (gnu/linux)

Hi Mark,

> Since you don't want to continue this discussion on a theoretical basis,
> can you please provide a concrete example of how the addition of
> scm_exact_integer_sqrt might be a maintenance burden in the future,
> given that our public C interface already consists of approximately 2K
> functions, including most (maybe all?) of the other numeric operators?

Ok, ok, I have to admit that my reply extended the context from the
specific question of exact-integer-sqrt to the more broader question of
"write Scheme rather than C".  As such most of what I wrote should be
read in that broader context.

> What I'm asking for is a specific example of how some future change in
> the internal workings of Guile's implementation might result in us
> saying "Damn, if only we hadn't exported those last few numerical
> operators, this would've been less of a pain!"

Returning to the question at hand, rather than publishing lots and lots
of C interfaces, I would much rather see generic approaches like what
Andy wrote in his reply:

> Nota bene: this doesn't have to make the lives of C users difficult!  On
> the contrary, we have to make it easy to invoke the Scheme API, and in
> this case to get some number of expected return values.  For example in
> this case, one could do a:
>   scm_call_1_2v (scm_variable_ref (exact_integer_sqrt_var), n, &root, &rest);

This looks very nice and reduces lots of function calls to a more
generic interface that will be easier to maintain and is completely
orthogonal to what is done in scheme.

The opposite side of this coin - calling C from scheme - is nicely
demonstrated by the sqlite3 bindings that Andy come up with[1].  In
there, _everything_ is done in scheme.  Not a single line of C is
needed.  This is due to the genericity of 'pointer->procedure' plus

In my opinion it is a worthwhile goal to strive for such genericity also
in the C calling to Scheme domain.



Practice random senselessness and act kind of beautiful.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]