Re: as for Calc and the math library

[Nicolas Martyanoff <nicolas@n16f.net>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Nicolas Martyanoff <nicolas@n16f.net>
>> Cc: arthur miller <arthur.miller@live.com>,  nicolas@n16f.net,
>>   emacs-devel@gnu.org
>> Date: Tue, 13 Aug 2024 15:10:28 +0200
>> 
>> Eli Zaretskii <eliz@gnu.org> writes:
>> 
>> > AFAIU, there should be no reasons not to be able to load MIT licensed
>> > libraries via the emacs-module machinery.
>> 
>> IIRC the reason I abandoned was that you can load a dynamic module but
>> there is no mechanism to reload it once modified, meaning one cannot
>> develop packages with dynamic modules iteratively.
>
> This is a technical problem with loading shared libraries, so it will
> also happen with FFI, AFAIU.
>
> In general, unloading and reloading doesn't work in Emacs well even in
> Lisp, less so with native-compiled Lisp.  Not sure this can be
> improved and how, but it would be a welcome addition and enhancement.

The main difference is that with a dynamic module, if you want to add a
new C function available in elisp or modify an existing one, you have to
update the C glue code defining the function and restart Emacs to build
and reload it.

With the kind of FFI API we are talking about, once the
foreign library has been loaded (e.g. libpq), one can write Elisp
functions calling foreign functions incrementally.

This is not the end of the world if you are just binding a couple
functions and you know everything you need. When you are binding a
large and complex library, being able to go at it incrementally and
testing everything in Emacs at each step makes it much comfortable.

Ultimately all of this is moot since the political aspect takes
precedence. But thinking about it, it should be possible to build a
dynamic module exposing a couple functions whose job would be to open
arbitrary shared libraries and call functions with libffi (which is MIT
licensed so no license violation here). Some Emacs code could be used to
extract constant values, structure layouts and other information
necessary at runtime, but it is not that hard (see cffi-grovel in Common
Lisp for example).

-- 
Nicolas Martyanoff
https://n16f.net
nicolas@n16f.net