The main issue I've encountered is the occasional compiler bug. If
something doesn't work, more-so than when using a more mature
implementation, one should suspect that perhaps a compiler bug is
involved. I've noticed, for instance, cases where the interpreter and
the compiler give different results (closures being one case, though
Past that the biggest thing I miss is the ability to save the state
of the lisp, as I can in most other native implementations. So
startup time on complex systems can border on unacceptable for wide
deployment, because of the need to load many files. There's been some
discussion of this on the list in the past.
Debuggability is another issue, with integration with slime
substantially less so than with other lisps - one can't view or
change local variables in frames, for instance, even when running
interpreted, there isn't a facility to restart a frame, no stepping
The documentation notes issues with CLOS performance - what can be
more disruptive more my development is the occasional and
unpredictable tendency for it not to process redefinitions, leading
to confusing situations where you recompile and then don't see your
changes having effect.
But, as you note, the ability to write completely cross-platform
(modulo some issues with path name formats) lisp code that integrates
easily with the many available JAVA libraries is a total win,
outweighing, for me, the other issues.
On Jun 17, 2007, at 12:36 AM, Chas Emerick wrote:
> After grabbing ABCL v0.0.10.5 from CVS, I was pleasantly surprised to
> see cl-ppcre's test suite runs (a) with the same results as reported
> by SBCL v1.0.2, and (b) with very respectable speed (sbcl v1.0.2
> finishes (cl-ppcre-test:test) in 19s here, ABCL finishes in 45s).
> I was particularly surprised (pleasantly so!) by ABCL's speed.
> Running at ~50% of SBCL is entirely acceptable given ABCL's Java
> integration, attendant deployment options, and relative youth.
> So, I'm pressing on. I didn't see a TODO file in the source tree,
> and nothing similar was apparent in a couple of searches online. I'd
> be very interested in learning what others view as the key weaknesses
> of ABCL (besides, or perhaps more specifically than those cited on
> the project page). Such information would be very helpful in keeping
> me (and others, I would suspect) from playing hunt-the-wumpus too
> much before deciding to seriously invest in ABCL.
> - Chas
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> armedbear-j-devel mailing list