Steel Bank Common Lisp

On 02/17/2012 06:06 PM, Myrosia Dzikovska wrote:
> Hi,
>
> I am considering moving to sbcl from Allegro CL. I work on a project
> that involves a bunch of components written in different languages,
> communicating using ICE middleware (http://www.zeroc.com). There is no
> "native" ICE interface for LISP, and none is planned anytime soon.
> There are C and Python interfaces. The protocol requirements are
> complex enough that I don't have time to implement these in LISP.
>
> My solution so far has been as follows: have the ICE server code
> running in C++, accepting requests. When a request comes in, the C
> code calls a lisp call handler, eventually gets the results back, and
> passes them to the caller. Calling LISP from C is possible, if
> awkward, in Allegro CL.
>
>
> Sbcl manual says that calling LISP from C is hackish and poorly
> supported, and does not provide any detail. In practice, doing this in
> ACL is extremely hackish as well, with various issues and
> restrictions. But it was documented enough for me to make it work.
>
> Has anyone had any experience with calling lisp from C in sbcl? I'd
> like to know if this is worth my time, or I should look for a
> different LISP distribution. I'd consider alternatives to directly
> calling LISP from C, e.g., if it would be possible to have a thread
> sleeping in LISP and woken up from C (but it's not obvious to me that
> it would work from the manuals, either).
Have you tried using CFFI ? CFFI provides macros for defining callbacks;
but more importantly provides a uniform API between different lisp
implementations for interfacing with foreign code.

Akshay

On Fri, Feb 17, 2012 at 12:39 PM, Akshay Srinivasan
<akshaysrinivasan@...> wrote:
>> it would work from the manuals, either).
> Have you tried using CFFI ? CFFI provides macros for defining callbacks; but
> more importantly provides a uniform API between different lisp
> implementations for interfacing with foreign code.
>
> Akshay
>
>


Thanks! That library looks very interesting, especially from the point
of view of interoperability. But the docs don't say much
about callback in sbcl: "Not all platforms support callbacks." So that
doesn't help me decide whether sbcl is going to work at all.

Someone else suggested in private mail that I try a pipe trick, with
LISP polling a pipe stream, and C writing notifications as necessary.
That's the option I'd go for, I think. If I can make it work with
CFFI, even better ;-)

Myrosia

On 02/17/2012 09:21 PM, Myrosia Dzikovska wrote:
> On Fri, Feb 17, 2012 at 12:39 PM, Akshay Srinivasan
> <akshaysrinivasan@...> wrote:
>>> it would work from the manuals, either).
>> Have you tried using CFFI ? CFFI provides macros for defining callbacks; but
>> more importantly provides a uniform API between different lisp
>> implementations for interfacing with foreign code.
>>
>> Akshay
>>
>>
>
>
> Thanks! That library looks very interesting, especially from the point
> of view of interoperability. But the docs don't say much
> about callback in sbcl: "Not all platforms support callbacks." So that
> doesn't help me decide whether sbcl is going to work at all.
I use CFFI with SBCL; it does work :)

>Sbcl manual says that calling LISP from C is hackish and poorly
>supported, and does not provide any detail. In practice, doing this in
>ACL is extremely hackish as well, with various issues and
>restrictions. But it was documented enough for me to make it work.
>
>Has anyone had any experience with calling lisp from C in sbcl? I'd
>like to know if this is worth my time, or I should look for a
>different LISP distribution. I'd consider alternatives to directly
>calling LISP from C, e.g., if it would be possible to have a thread
>sleeping in LISP and woken up from C (but it's not obvious to me that
>it would work from the manuals, either).

My experience with using FUSE with Lisp (it is _all_ about callbacks)
shows that a callback into lisp done by C code inside a Lisp-launched
thread generally works. C-launched threads plus callbacks are a recipe
for pain.

> My experience with using FUSE with Lisp (it is _all_ about callbacks)
> shows that a callback into lisp done by C code inside a Lisp-launched
> thread generally works. C-launched threads plus callbacks are a recipe
> for pain.
>

Definitely true in Allegro CL, too. I jump through quite a few hoops
in my ACL interface because of that ;-)

Le 17 février 2012 13:36, Myrosia Dzikovska <myrosia@...> a écrit :

> Sbcl manual says that calling LISP from C is hackish and poorly
> supported, and does not provide any detail. In practice, doing this in
> ACL is extremely hackish as well, with various issues and
> restrictions. But it was documented enough for me to make it work.
>
> Has anyone had any experience with calling lisp from C in sbcl? I'd
> like to know if this is worth my time, or I should look for a
> different LISP distribution
>

The manual is pretty out-of-date on that point: it hasn't been updated
since alien callbacks were merged 6 years ago or so. You could either use
CFFI's facility, or SBCL's, which isn't fully exported and documented:
sb-alien::define-alien-callback. In either case, you need to let Lisp start
the process, and you should only call from C into Lisp in threads created
from Lisp. That last point was already mentioned, but isn't strictly true;
you can set things up to call into Lisp from a thread created by C, but it
probably isn't work the effort (and that *is* hackish).

All in all, C-to-Lisp support is about equally workable but unpleasant in
both ACL and SBCL.

-Thomas