freenode/#clasp - IRC Chatlog
Search
23:12:22
Bike
my basic-ass transform for that just looks for fixed-to-multiples (so, VALUES calls) but shoudl be fairly easy to extend with type inference like sbcl does
23:13:04
karlosz
i was thinking propagating the values types might be a good first step towards doing local type propagation
23:15:34
Bike
but i would be pretty ok with dropping tha thonestly, you can have extended sequences without making FIND etc generic
23:15:47
karlosz
so i just added it for things like remove-if because i was able to shave megabytes of consing off with that as dx-call
23:16:59
Bike
i mean some idiot user could do (defmethod find-if (predicate ...) (setf *h* predicate) ...)
23:18:49
karlosz
its not a problem if i mark only the cl standard library sequence functions as dx-call right
23:21:33
Bike
but like i said, i'm open to dropping that, since the ability to define methods on sequence functions isn't needed to implement extended sequence classes
23:21:44
Bike
https://github.com/clasp-developers/clasp/blob/master/src/lisp/kernel/clos/sequences.lsp#L189-L204 here's my mega pedantic comment
23:22:04
Bike
like, it's fine. just mark them as dx-call and we'll deal with it when people actually use ext sequences
23:55:26
karlosz
oh, you made set-or-bind-variable take a thunk instead of a next-ast, but i think that's no longer necessary
0:34:11
Bike
everything about those thunks is ugly, if you have a better way to write it that still works please put it in
1:49:02
drmeister
I blame you Bike and karlosz that I wrote so much code missing (declare (ignore whatever)) declarations. You had to add these wonderful compiler warnings so late.
2:53:47
karlosz
Bike: i finally made the change where now cst-to-ast directly converts LETs. its a fairly big efficiency win in terms of compilation time and consing