[gambit-list] Libraries for day-to-day projects ?

Álvaro Castro-Castilla alvaro.castro.castilla at gmail.com
Tue Jul 8 11:56:03 EDT 2014


On Tue, Jul 8, 2014 at 5:31 PM, mikel evins <mevins at me.com> wrote:

>
> On Jul 8, 2014, at 9:59 AM, Álvaro Castro-Castilla <
> alvaro.castro.castilla at gmail.com> wrote:
> > Hi Eric,
> >
> > I'm trying to work on that problem with the project
> http://schemespheres.org
> > It is sort of Quicklisp for Gambit, with multiplatform support. It is
> not "complete" in the sense that some important parts are missing, but the
> upcoming version brings full support for Linux/OSX/iOS/Android. This
> version is currently beta (the development branch of the github projects).
> > I will work soon also integrating or developing most of the things you
> mention, although I've been more focused on the general aspects of
> integrating several libraries/modules and multiplatform support. Most
> notably, it's lacking database interaction and a web server, but of course
> it's planned in the future. Most of those things can be found as
> independent project, and my intention is to merge some of them in the
> project (with the due credit and licenses). Some times this is
> straightforward and some times it isn't, and that's one of the reasons of
> the project. Spheres are fully compatible with Gambit, so you don't have to
> worry about getting locked into the platform.
> >
> > Hope it helps, and I hope the project grows to better provide that what
> you need.
>
> I'd love to see SchemeSpheres turn into a Quicklisp for Gambit. It might
> be challenging to achieve that. Some perspective about what Quicklisp is
> and what it does might be helpful.
>
> The instructions for installing Quicklisp on a supported Common Lisp
> implementation are:
>
> 1. Download this link: http://beta.quicklisp.org/quicklisp.lisp
>
> 2. Load that file in your Lisp.
>
> That's it; you're done. All 1,000 or so Quicklisp-supported libraries are
> now available to use.
>
> If you want, you can add a small blob of code to your Lisp's init file
> that will automatically initialize Quicklisp any time you start the Lisp.
> Quicklisp itself can do that for you, if you like: eval
> (ql:add-to-init-file) and it will detect which Lisp you're running and add
> its startup code to the appropriate init file.
>
> From some Lisps you can even skip Step 1 above, because they can load
> quicklisp.lisp directly from the URL.
>
> Once you have Quicklisp, using it is about as easy as it could possibly
> be. Run ql:system-apropos to find out the exact name of a library you're
> interested in, then run ql:quickload with the name of the library. Done.
>
> Quicklisp interoperates with ASDF, the system-definition tool, making it
> equally easy to use its libraries in your projects. All you have to do is
> add a :depends-on clause to your system definition, mentioning the names of
> any Quicklisp libraries you want to use, make sure that you've installed
> them using ql:quickload, and you're done.
>
> Quicklisp doesn't have anything like SchemeSpheres' concept of "spheres".
> Its collection of libraries is a completely flat list of a thousand or so
> individual libraries with no sort of hierarchy or organization imposed on
> them. I haven't found that to be a weakness. On the contrary, it's simply
> one less thing I need to learn about in order to use it.
>
> If it has a weakness it's that, with over a thousand libraries available
> and no reliable source of summaries of what they're for, it can take
> significant time to find out exactly which libraries you might want to use.
> That's a pretty small quibble in the scheme of things, though. Quicklisp is
> incredibly useful and convenient. I use it in 100% of my Common Lisp
> projects, and I miss it whenever I'm working in another language.
>
> There are pretty good package managers for some other languages, but it's
> hard to beat Quicklisp for convenience and ease of use. The fact that it
> works with most Common Lisp implementations is just gravy.
>
>

Thanks for the explanation, I appreciate. I might have used the similarity
to Quicklisp too lightly, since I just took the reference regarding the
concept of "set of libraries for day-to-day projects".  I'd say that the
concept is closer to the Yeoman project + a set of generic libraries.

I don't see SchemeSpheres reaching that kind of integration any time soon.
The model is more adapted to Gambit's C/universal code generation (doesn't
intend to be multi-implementation at the moment), and tries to make it
easier to use and automate some of Gambit's features. I don't see the idea
of a flat super-module working well for Gambit, but maybe I'm wrong on this.

This leads me to a question: would Gambit be able to discard unused
function in some way? I guess it's a hard thing to do since the generated
code needs to be linked and thus only a link-time optimization would be
able to do that... and still, that would be for C functions.

Perhaps a post-syntactic expansion done at the Scheme code level could
achieve that, and pave the road towards the concept of Quicklisp. Either
that or a module-per-function approach. But this last approach wouldn't
allow Gambit to provide for the single-host optimization, am I right?

Álvaro
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.iro.umontreal.ca/pipermail/gambit-list/attachments/20140708/a65730d2/attachment.htm>


More information about the Gambit-list mailing list