Re: [Chicken-users] Noob question: (re)compile FFI-accessing code within REPL

[Graham Fawcett]

On 6/28/07, Martin Percossi <address@hidden> wrote:
> Now, supposing I wanted to change the c-test to:
>
> (define (c-test)
>   ($ printf "BOO!")

> The question is: am I obliged to recompile outside of the REPL loop?
> In other words, is there a `compile' command which lets me recompile
> test.scm from within the REPL?

Not within the interpeter, no. Instead you could try one of these:

* (define (my-printf . args) ($ printf args)) in your compiled code
 and then (define (c-test) (my-printf "BOO")) in a non-compiled
 Scheme file (or just type it in the interpreter). In other words,
 compile the "primitives" only.

* Don't use the REPL. Define your code to be compiled, then write a
 test script (think "unit tests") that exercises your code. Use a
 Makefile or other script to build compile your code as needed and
 automatically run the tests. When it's sufficiently cooked, load it
 in your REPL and play around.

* Use the lazy-ffi instead. It works nicely at the REPL:

 (use lazy-ffi)
 #~"libc.so.6"
 (define (c-test) (#~printf "BOO!"))

* The "inline" egg has support for compiling C on the fly, into a
 temporary shared-library that is then loaded into the interpreter. A
 similar approach could work for Scheme procedures, I think, though
 I've never tried. The approach would be to

 * write out the procedure definition to a temporary file,
 * csc -s /tmp/the-file.scm -o /tmp/tempXXX.so
 * load the /tmp/tempXXX shared library.

 It might not be a one-size-fits-all solution, but it might just
 work. Of course, it doesn't exist yet. :-)

> - Is quack generally preferred over hen in emacs? Quack looks pretty
> good; haven't tried hen yet. Keep in mind that I will be using windows
> eventually.

I haven't used hen either; I do run Quack and haven't had any problems
with it. I find that paredit-mode is indispensible, and mmm-mode can
help with literal sections (#<<EOF) and some other syntactic stuff.

> - When I download the readline egg, I get nice command completion on the
> csi REPL -- sweet! Unfortunately, it doesn't seem to work within emacs
> (quack), either in the REPL or when editing a scheme file. Is there any
> way to make this work?

Hm. You could try running csi in a term buffer (M-x term). You would
have to tell Quack (or cmuscheme?) where to find your csi buffer,
though, if you want them to integrate properly. This might work:

(defun run-csi-in-term ()
 (interactive)
 (term "/usr/local/bin/csi")
 (rename-buffer "*scheme*")))

There might be easier solutions though, other than readline
support. Using a tags-table would be the "Emacs way" I think.

Good luck,
Graham