From: Olivier Dion via General Guile related discussions <guile-user@gnu.org>
To: Tim Meehan <btmeehan@gmail.com>
Cc: guile-user <guile-user@gnu.org>
Subject: Re: C extensions
Date: Sun, 21 Feb 2021 17:41:16 -0500 [thread overview]
Message-ID: <87wnv159ir.fsf@clara> (raw)
In-Reply-To: <CACgrOxL3TZ=bpRCuDuGztYq8b9GFfOD6hf9k6xVs7++c5WYT8g@mail.gmail.com>
On Sun, 21 Feb 2021, Tim Meehan <btmeehan@gmail.com> wrote:
> Hello Olivier, thanks!
> Those functions helped, and now I know where in the source of Guile to look.
>
> Is "inexact" the same thing as "floating point number" in Guile-speak?
I don't think so. If I remember correctly, exact numbers are numbers
that can be represented by a quotient of two integers. In other words,
any rational number is an exact number. I might be wrong here. I
highly suggest that you read the numerical tower from the Guile's
texinfo manual for more informations on that.
>
> Here is one of the functions in my extension - it returns a lognormal
> variate. There are a bunch of things like this that could be added to
> "random.c" - but I'm imagining that the only user of stuff like that might
> be ... me ... anyhow:
>
> Right now, it doesn't give really helpful errors when you give it a
> negative standard deviation:
> scheme@(guile-user)> (random:lognormal 0 -1)
> $1 = #f
In my example, I returned false because this is how I would approach the
problem from a C perspective. i.e., returning an error code. I'm still
learning on how to work with Scheme. I guess that if you want something
more Pythonic, you would throw an exception.
> SCM_DEFINE (scm_random_lognormal, "random:lognormal", 2, 1, 0,
> (SCM _mu, SCM _sigma, SCM state),
> "Return an inexact real in a Log Normal distribution.\n"
> "\n"
> "@var{mu} - mean of the underlying normal distribution.\n"
> "@var{sigma} - standard deviation of the same.\n"
> "Optionally, a random @var{state} can be provided.\n"
> "See @code{seed->random-state}.")
> #define FUNC_NAME s_scm_random_lognormal
> {
> if (SCM_UNBNDP (state))
> state = SCM_VARIABLE_REF (scm_var_random_state);
> SCM_VALIDATE_RSTATE (1, state);
>
> if (scm_is_false (scm_positive_p (_sigma)))
> return SCM_BOOL_F; // I'd like to return some meaningful error
> here ...
You can use `scm_throw()` to throw an exception.
----------------------------------------------------------------------
/*
* This create a symbol that has static linkage in C. You might want
* to declare it globally instead.
*/
SCM_SYMBOL(lognormal_error_key, "invalid-argument");
if (scm_is_false (scm_positive_p (_sigma))) {
scm_throw(lognormal_error_key, scm_list_n(SCM_UNDEFINED));
/* Unreachable path! */
assert(0);
}
----------------------------------------------------------------------
Note that throwing an error might be easier to do in Scheme itself:
----------------------------------------------------------------------
(throw 'invalid-argument)
----------------------------------------------------------------------
>
> double mu = scm_to_double (_mu);
> double sigma = scm_to_double (_sigma);
> return scm_from_double (exp (mu + sigma*scm_c_normal01
> (SCM_RSTATE(state))));
> }
> #undef FUNC_NAME
>
> On Sun, Feb 21, 2021 at 9:59 AM Olivier Dion <olivier.dion@polymtl.ca>
> wrote:
>
>> On Sat, 20 Feb 2021, Tim Meehan <btmeehan@gmail.com> wrote:
>> > I'm trying my hand a writing C extensions.
>> > I've done this for stuff like Matlab before, and was wondering how you do
>> > the usual checking of the arguments that are passed in from Guile.
>> >
>> > In the manual, 6.13.13.1 "C Support" has a few functions.
>> > libguile/numbers.h has a bunch more ...
>> >
>> > What I have is an extension function, sort of like the bessel function in
>> > the tutorial:
>> >
>> https://www.gnu.org/software/guile/manual/html_node/A-Sample-Guile-Extension.html
>>
>> Usually I do this in 2 or 3 steps.
>>
>> 1. Define the primitive in C with raw arguments.
>>
>> 2. Make a wrapper of this primitive for Guile to use. Optionnaly do
>> the arguments checking here or go to step 3.
>>
>> 3. Make a second wrapper in Scheme and check the arguments there.
>> This wrapper will call wrapper in made in 2.
>>
>> Here's a full example:
>> ----------------------------------------------------------------------
>> float c_sin(double x)
>> {
>> return sin(x);
>> }
>>
>> SCM_DEFINE(c_sin_wrapper, "sin", 1, 0, 0,
>> (SCM x),
>> "My sine")
>> {
>> if (scm_is_number(x)) {
>> return c_sin(scm_to_double(x));
>> }
>>
>> return SCM_BOOL_F;
>> }
>> ----------------------------------------------------------------------
>>
>> > What I would like to do is verify that the first argument is an inexact
>> > number, larger than 0. How would I go about that? Perhaps some of it
>> > could
>>
>> For your example, use the following predicates:
>> ----------------------------------------------------------------------
>> scm_is_inexact(x) && scm_is_true(scm_positive_p(x))
>> ----------------------------------------------------------------------
>>
>> > be:
>> > SCM_REALP
>> > scm_misc_error
>> >
>> > Any tips?
>> --
>> Olivier Dion
>> PolyMtl
>>
--
Olivier Dion
PolyMtl
prev parent reply other threads:[~2021-02-21 22:41 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-02-21 3:56 C extensions Tim Meehan
2021-02-21 4:10 ` Nala Ginrut
2021-02-21 16:00 ` Olivier Dion via General Guile related discussions
2021-02-21 21:58 ` Tim Meehan
2021-02-21 22:18 ` Tim Meehan
2021-02-21 22:41 ` Olivier Dion via General Guile related discussions [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/guile/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87wnv159ir.fsf@clara \
--to=guile-user@gnu.org \
--cc=btmeehan@gmail.com \
--cc=olivier.dion@polymtl.ca \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).