6 May 2008 17:43
Re: [open-axiom-devel] [Axiom-developer] Am I posing this solve problem wrong?
"Bill Page" <bill.page@...> writes: > I have been thinking specifically about the algorithms that the Axiom > interpreter uses to do type inference and function selection. Some part of > the behaviour is "hard coded" into the interpreter and not easy to change, > but a large part of it depends on interaction with the design of the Axiom > library, i.e. the coercions and other operations that are exported by exposed > domains. It seems to me that in many cases not enough attention has been paid > to "tuning" the Axiom library so that the Axiom interpreter produces fewer > surprises for the new user. Some time ago I posted a question in exactly this direction (something like "why doesn't HOAGG R export coerce: % -> List R and coerce: List R -> %). I can only guess that one answer (which I didn't get, but I give myself) is optimization: coercion done by the library *might* be relatively slow. However, I'm quite sure it would make a lot of sense to replace the contents of i-coerce by appropriate definitions in the algebra. Exploration here is very very welcome -- and in FriCAS also relatively painless, one only has to keep in mind that changes at this level usually make complete recompilation of the algebra necessary. Of course, the insights will be valuable for all three projects. Martin ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference(Continue reading)
[ That is probably because I've tried many radical ideas, only to
discover that there were reasons why they did certain things the way
they did, even if I don't like it. ]
| that any such consideration of optimization was ever applied. I expect
| that the detailed coding in the Axiom interpreter and the coercions
| supplied (or not) by the Axiom library are merely artifacts of the
| constraints to the original Axiom development process.
OpenAxiom already moved many of syntactic transformations and
coercions to library. More reamin to do done; but that would rely on good,
clear, abstract, object models of computations.
-- Gaby
-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
RSS Feed