Subject: Re: the state of the art of predicate dispatching
From: Craig Chambers (chambers@cs.washington.edu)
Date: Mon Jul 09 2001 - 17:29:36 PDT
Doug Orleans wrote:
> 1. Publications: As far as I can tell, there are still only two
> published papers on predicate dispatching: "A Unified Theory of
> Dispatch" from ECOOP98, and "Efficient Multiple and Predicate
> Dispatching" from OOPSLA99. Am I missing anything? Aaron, did you
> finish your master's thesis on adding predicate dispatching to
> CLOS? Is it available online?
There's the Predicate Classes paper from ECOOP'93, which led to part of
predicate dispatching's functionality. There are several other languages with
features similar to predicate classes, and there was a paper at a CLOS workshop
at ECOOP'93 where someone (I don't remember who) presented a paper on Predicated
Classes in CLOS.
>
> 2. Code: The GUD implementation is avalable from the Cecil project
> site, but there is no source code. Can this be made available?
This is a prototype version of Dubious that Craig Kaplan put together, which
reflects his ideas about the language. I don't think this implementation
implemented the full predicate dispatching & typechecking rules presented in the
ECOOP'98 paper. Craig K. decides how public he wants to make his source code.
The Vortex implementation of the dispatching algorithms from the OOPSLA'99 paper
can be made available, if it's needed. But it doesn't focus on the predicate
implication reasoning part, and in fact, I think only the Cecil subset of
predicate dispatching is implemented.
> 3. Functionality: The biggest thing that seems to be missing from
> predicate dispatching as described in the two papers is
> call-next-method/resend. Has anyone thought of an elegant way to
> handle this? I think it boils down to providing a way to select a
> subset of the methods that were applicable to the current set of
> arguments, such that there is another single most specific method
> in this subset. Basically some generalization of Cecil's directed
> resend is what I'm aiming for.
I haven't thought about it much, but I'm not sure what more you need. Is
Cecil's possibly-directed resend not all that's needed? Are you asking about
semantics, or implementation?
>
> 4. AOP: Has anyone else thought much about the connections between AOP
> and predicate dispatching? The similarities between predicates and
> AspectJ's pointcuts seem significant to me, even down to both
> producing a set of bindings to be used in the method/advice.
> Basically, both systems let you say "when a function is called, and
> condition X is true, do Y". My toy predicate dispatch system has,
> instead of generic functions containing methods, a single global
> set of methods that dispatch on join points, where a join point
> object contains the generic function and the list of arguments.
> Most method predicates just compare the jp's gf to the gf that the
> method "belongs to", but aspectual predicates can make more
> interesting decisions based on the gf (like, testing if the gf is
> in a set). Right now my stumbling block is how to get an aspectual
> predicate to be considered more specific than the predicate for the
> actual method implementing the join point; I think I may make
> "before", "after", and "around" be different types of join points
> from regular join points, so that an aspectual predicate only has
> to be more specific than other aspectual predicates. [Sorry if
> none of this makes sense-- I can post some code examples if you want.]
Interesting. I haven't considered the connection beyond thinking that predicate
dispatching is potentially a flexible base for a variety of composition
mechanisms.
>
> 5. The future: What does the future hold for predicate dispatching?
> It seems like the Cecil team has switched focus to spreading the
> multiple-dispatch word, i.e. to Java programmers, but I assume
> you've still got some predicate dispatch projects on the back
> burner? Jonathan, will you be incorporating some form of predicate
> dispatching into Proto?
I wouldn't say we've switched focus, but we have lots of different foci, some of
which are more on the forefront of our attention than others. We have put less
effort recently on the boundaries of flexibility offered by predicate
dispatching, pursuing instead modularity solutions (on the language design
front) that help in "spreading the word" to Java (and ML, and software
architecture/components) programmers, and on new compiler architectures (on the
implementation front), plus other things. We have limited resources, and
pursuing predicate dispatching doesn't seem to be the biggest bang for our buck
right now, nor the greatest area of student interest.
-- Craig Chambers
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://majordomo.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2b25 : Mon Jul 09 2001 - 17:30:04 PDT