Re: Integrate Guile with GNU Scientific Library

[Unknown <unknown@unknown.invalid>]

On 4 Oct 2019, 03:10, Matt Wette < matt.wette@gmail.com> wrote:

On 10/3/19 12:28 AM, Freeduck via General Guile related discussions wrote:

> Hi All,

>

> I am trying to integrate Guile with a part of GNU Scientific Library and have found SWIG, which seems very easy to use, but only supports Guile 2.0. I have Guile 2.2 installed.

> The Guile homepage has a tutorial on integrating Guile into C, and the manual has a section on dynamic ffi, but it I find it hard to get started, maybe it is just lack of experience.

>

> My main problem is not integrating with GSL, but doing polynomial regression. I have been working in Racket for a while which has allot of maths integrated. Is there something like numpy/scipy for Guile.

>

> Kind regards

>

> Kristian

>

> Sent with [ProtonMail](https://protonmail.com) Secure Email.

The FFI Helper in nyacc (https://savannah.nongnu.org/projects/nyacc)

provides the functionality of swig.  You can try that.  I'd be

interested in how you like that compared to swig.

A quick hack to see if I could get it to work for gsl failed:

$ cat gsl-sort.ffi

(define-ffi-module (gsl-sort)

  #:pkg-config "gsl"

  #:include '("gsl/gsl_sort.h"))

$ guild compile-ffi gsl-sort.ffi

...

ERROR: In procedure dynamic-link:

In procedure dynamic-link: file: "libm", message: "file not found"

So for some reason the linker is not finding libm.so on my Unbuntu.

I will need to look into that.

You can check the examples in examples/nyacc/lang/ffi-help.

First, sorry if I'm messing up bottom posting and formatting ... I'm responding using a non-free mobile phone app :)

Second, the answer might be totally off-topic, but I always found it easier to generate math programs by Guile and then compile them, then a full, seamless library level integration into a Guile program itself. See, e.g.

https://gitlab.com/codetk/schemetran

for a naive implementation of fortran wrappers and compiler helpers in Guile. It is naive, but I managed to do a lot of sophisticated linear algebra with it.

Anyway, good luck :) .

T

PS also in mine previous experiences when trying to link stuff using ffi, looking at the symbols in libraries directly helped.