freenode/#lisp - IRC Chatlog
Search
10:34:09
makomo_
edgar-rft: regarding your comment about feedback paths -- yes, it's possible to model those as well since pins can be used as input, output or both
10:37:32
mrcom
What's the type declaration for (defun baz (f) (declare (type <<??? something that accepts foo>> f))) ?
10:49:38
mrcom
That's what I wound up doing :/ But the better really, really hates losing to the good.
10:51:00
trittweiler
I'd be surprised if there's a compiler that does anything more with that specification in praxis that it wouldn't do with (declare (type function f))
10:52:42
mrcom
I'm looking more for type checking than optimizing. SBCL definitely checks when it compiles apply or funcall.
10:54:04
trittweiler
When you declare a parameter to be a function, and you funcall that parameter, the compiler doesn't need to go through an implicit call to fdefinition. (As symbols and lists of form '(setf foo) are also callable)
14:13:37
flip214
trittweiler: with SBCL and (optimize speed) the disassembly with the different declare statements is quite different...
15:55:51
beach
random-nick: There is usually no point in asking "why" in such situations. This is true both for programming languages and natural languages. At some point, there was a mysterious consensus that one was better than the other. That's all there is to it usually.
15:57:00
beach
random-nick: Why is it called "tooth brush" but "dental floss" in English? Why not "dental brush" and "tooth floss"?
15:59:27
beach
random-nick: You also have define-compiler-macro, define-method-combination, define-modify-macro, define-setf-expander, define-symbol-macro.
16:01:29
jackdaniel
random-nick: one (made up) reason could be, that defun and defmacro were heritage from older lisps, while "modern" way for naming macros when CL was standarized was using define-* and the were to add conditions to the language
16:02:55
jackdaniel
random-nick: I've given you a made up excuse, there could be others. I wouldn't bother with figuring that out
16:03:04
jackdaniel
there are more annoying inconsistencies in CL, like argument order in aref and gethash for that instance
16:05:30
jackdaniel
I haven't seen even one project which would use CL21; I've only seen people praising or ignoring it ;-)
16:05:46
jmercouris
I've never used it, simply because the problems it addresses don't bother me enough
16:25:13
jackdaniel
or if I use named-readtables and I have #f as lambda dispatch - is it inconsistent?
16:25:25
pjb
(and there is unfortunately a lot of code that is not conforming: that is dependent on specificities of eg. sbcl).
16:31:02
JuanDaugherty
without expecting them to debug or develop around shortcomings/lack of completion in CLIM
16:31:51
JuanDaugherty
the why not is because some years ago it was not in that state, at least as far as use with sbcl
16:31:53
jackdaniel
usually writing complete applications requires some debugging and developing around shortcomings of used libraries. if that's your definition then it is not production ready ;)
16:33:37
jackdaniel
well, that depends then solely on the programmer. if he exposes interfaces which allow corruping application from the end-user point of view, then they will be able to corrupt it. but I truly don't think it is any library-specific problem, just a matter of testing and programming / qa effort
16:35:20
jackdaniel
what's based on use of it? I've stated an opinion about all software above, not CLIM particularily
16:36:45
jackdaniel
biggest shortcomming one could find is that McCLIM depends on X server running currently to have something displayed on the screen
16:38:00
jackdaniel
limitation is more painful when you consider distribution for Windows or OSX - that requires installing X-server by hand