[elephant-devel] Re: elephant and sbcl

[Ben]

Elliott --

Thanks for the info.  Glad to hear people are using Elephant (and
finding it useful!)

Christophe told me that the way I do string serialization should
continue to work -- with some adjustments.  I'm guessing that elephant
is dying because it is not correctly computing the number of bytes
underlying a string.  Probably there are other issues with the new
SBCL unicode stuff as well.

Another big headache for Unicode strings is going to be key sorting
functions.  One would like for string keys to be alphabetically stored
-- in other words, one probably wants a (C?) function to sort strings.
For Unicode this is of course rather difficult.  In Lispworks / ACL,
since they use UTF-16, I managed to steal some code from IBM's ICU
project.  Maybe a Lisp solution will be better in the long term --
though this will require good callback support.

I've been waiting for the unicode stuff to settle down a bit more
before attempting to get it to work.  If people want this sooner
rather than later, let me know.  Of course I'm always willing to
accept patches.....

So -- it's on my TODO list for the next release.

take care, B

On Tue, 9 Nov 2004, Elliott Johnson wrote:

> Hi Ben,
>
> I've played with elephant quite a bit.  It makes life pretty nice :)
>
> Last night I coupled 0.2.1 with the latest SBCL (0.8.16.37) and it gave off some pretty
> weird type errors that look like they are due to a mix of serialization and the new
> unicode branch of SBCL.  I showed in #lisp, but I though you might also like to see
> what I got.
>
> Unfortunatly I couldn't get any real backtraces because it tossed me to the max
> debugger level.  You might find it helpful: http://paste.lisp.org/display/3668
>
> HTH!
>
> -Elliott
>