#clojure – 2008-03-03

2008-03-03

09:47	middayc	is clojure currently more as an experimental on in development language or it can it be used for practical use?
09:48	rhickey	I think it's practical, but as the author I'm biased
09:49	middayc	:) well as the author you know the best where the focus is
09:50	rhickey	it is in development, but most changes are for fixes, completeness, or new features, not experimentation
09:51	middayc	nice, I don't mind changes is language is still evolving
09:56	middayc	could I use something like LWJGL from clojure instead of java?
09:57	rhickey	I don't see why not
09:57	rhickey	although you can't target applets due to a Java restriction
09:59	middayc	great.. well opengl applets are very unstable anyway
09:59	middayc	I mean in regards to general public
10:01	middayc	I used java for some things and in the meantime I got consumed into these FP langs and paradigms .. now I would need to do some jvm related things but programming in java does not intrigue me that much
10:01	middayc	clojure seems very interesting from what I saw
10:02	rhickey	lisp + fp + JVM
10:02	rhickey	+ stm
10:05	middayc	One use for java and lwjgl is for something like 2d action game... is clojure somewhere near java speedwise? I think I noticed somewhere groowy for example was quite slower at some tests (I think it was raytracing)
10:07	rhickey	The only place Clojure should be noticeably slower than Java is in numerics - all numbers are boxed in Clojure. But Clojure + a small Java vector lib should do the trick.
10:08	middayc	aha.. so I could write computational libs in java and use them in clojure
10:08	rhickey	right
10:09	middayc	cool :)
10:09	Chouser	rhickey: do you have an opinion on conditional forms that capture a name for the test result, like arc's aif, acond, etc.?
10:09	Chouser	do you hate the idea?
10:09	rhickey	anaphorics - I dislike them
10:10	Chouser	ok
10:10	rhickey	I will be adding if-let and when-let, which will take the name of the thing to be bound
10:10	Chouser	ah! ok.
10:10	Chouser	so the source of the dislike is the unspecified name showing up in the enclosed block?
10:11	rhickey	yes, they don't nest, cause problems for macros etc
10:11	rhickey	too cute
10:11	Chouser	well, #() does that a bit now...
10:11	Chouser	with similar problems, of course.
10:12	rhickey	yes, and should be reserved for the simplest cases
10:12	Chouser	ok
10:12	rhickey	nesting #() is bad style
10:12	Chouser	I'll be looking forward to when-let.
10:12	Chouser	would a cond-let be too convoluted?
10:13	rhickey	one name for all conditions or a name per condition?
10:14	Chouser	hm... I hadn't thought of one name for all. That might be quite nice.
10:14	rhickey	a lot of conditions are just boolean, not much to bind
10:14	Chouser	I was thinking one name per condition would be pretty wordy.
10:14	Chouser	yep, but isn't that generally true for all these (if-let, when-let, etc.)
10:16	rhickey	depends on how you use cond
10:16	Chouser	the syntax would be: (when-let bindform testexpr block) ?
10:16	rhickey	yes
10:16	Chouser	I could probably write that... ;-)
10:16	rhickey	10 minutes tops
10:16	Chouser	Ok, anyway, thanks for the discussion.
10:17	Chouser	Not sure about cond -- just poking around is all.
10:23	Chouser	hmph. doto just betrayed me. I can't think of any easy way to run one of two different methods inside a doto block (based on some test).
10:24	Chouser	I'll have to refactor into "let" and a bunch of (. foo blahs)
10:25	rhickey	it would have to be more of a code walker to find . calls inside of ifs etc
10:26	Chouser	yep. more magic.
10:26	Chouser	It just suggests perhaps I shouldn't have used doto at all.
10:26	Chouser	it seemed so nice at first. ;-)
10:26	rhickey	probably not
10:27	Chouser	I shouldn't complain, though. Making Java interop possible is hard enough. Demanding pretty is too much.
10:30	arbscht	clojure is reasonably pretty, even
10:31	arbscht	(I say that having used Foil :-)
10:33	rhickey	doto is already more concise than the equivalent java
10:34	rhickey	but I think the next level up is something like with-slots for Java objects
10:34	rhickey	naming the object and its pieces
10:37	ericthorsen	Rich: I've come across a couple of scenarios where I need to return iterators / enumerations on clojure seqs to a java framework.
10:37	rhickey	all the collections are Iterable
10:37	ericthorsen	I wrote a simple wrapper that uses a ref to modify the state of a proxy since I always have to return the same reference back on the java side
10:37	ericthorsen	hmm...
10:38	rhickey	there's also SeqIterator which turns any seq into an Iterator
10:38	ericthorsen	That's much nicer of course...but I needed an Enumeration
10:39	rhickey	for what?
10:39	ericthorsen	it does bring up a general question about plugging java objects into existing frameworks implemented in clojure
10:39	ericthorsen	TreeNode.children returns an Enumeration
10:41	rhickey	aargh - I wish they had gotten it right the first time
10:41	ericthorsen	i know...
10:42	rhickey	2 'standard' iterators is silly
10:42	rhickey	I can provide equivalent bridges to/from Enumeration
10:43	ericthorsen	it would be useful since everyone is going to come up against this
10:44	ericthorsen	In cases where I'm handing off a java object to a framework that is expecting side effects to change the state of the object, the model would be to use references and transactions in the proxy yes?
10:48	rhickey	that's one way.
10:49	rhickey	it would allow those objects to participate in enclosing transactions
10:52	rhickey	SeqEnumeration is up
10:57	ericthorsen	thanks!
10:58	ericthorsen	what is another way ?
10:58	ericthorsen	no transactions?
10:59	rhickey	without transactions you devolve into Java's mutable state model, can just use clojure.lang.Box - don't call me with problems :)
10:59	ericthorsen	i know that...that is why I asked what is another way? I did not think that is one you were referring to
11:01	rhickey	that's pretty much it, safety or not, there's not really a per-thread-binding notion in normal Java objects, so no advantage of Var over Box
11:01	ericthorsen	...shy or storing the state somewhere else I mean...I'm fine with this option. I was just curious as to whether there was another option you had in mind for this scenario
11:01	ericthorsen	ok
11:03	rhickey	EnumerationSeq and support for Enumerations in seq is up
11:09	Chouser	(let [[foo bar] [1]]) --> java.lang.IndexOutOfBoundsException
11:09	Chouser	was it silly for me to expect bar to be nil?
11:11	rhickey	it is documented to use nth, which is documented to throw on out-of-bounds
11:11	Chouser	ok
11:12	rhickey	if you might have more stuff, you can use &
11:12	Chouser	yep, thanks.
11:13	rhickey	(let [[foo & [bar]] [1]] [foo bar])
11:13	Chouser	I have a cond-let working, with a single binding form. But so far I'm only using it like an "if", so I'm not sure yet if a single binding form makes sense in general or not.
11:13	rhickey	[1 nil]
11:13	Chouser	oh! fantastic.
11:15	rhickey	full destructuring is available for the & parameter
11:16	Chouser	makes sense, just not an idiom I had thought of. nice.
12:19	arbscht	does it make sense to define a multimethod with no args?
12:22	arbscht	would it be too non-functional to have arbitrary dispatching functions rather than only functions of the args? as it is that's possible for args>0
14:20	rhickey	dispatching function of no args? does it call rand or something?
14:33	arbscht	or something
14:33	arbscht	let's say it depends on a mutable variable
14:35	arbscht	there's nothing stopping me defining a dispatching function with non-zero arity and ignoring the arguments to do something like that
14:36	rhickey	or put an ordinary function in front of it, that reads var, passes something to multimethod
14:36	arbscht	right, that's how I've got it now
14:36	rhickey	probably easier for the next guy to understand that way
14:39	arbscht	maybe. my first instinct was to try it with no-arg multimethods, but that might just be me
19:41	rhickey	sets are coming, any syntax preference:
19:41	rhickey	#[a b c]
19:42	rhickey	#{a b c}
19:42	rhickey	something else?
21:40	Chouser	Running out of brackets... <a b c>
21:41	rhickey	still haven't decided to take away <, >
21:41	Chouser	yeah.
21:42	Chouser	#[] isn't bad. # is starting get pretty diluted.
21:43	rhickey	ever used CL?
21:43	Chouser	[\|a b c\|]
21:43	Chouser	not really. I "learned" lisp in CL from "On Lisp", but never wrote more than a few functions in it.
21:43	rhickey	I think the double chars are tricky for editors
21:43	Chouser	ah, good point.
21:44	Chouser	I was thinking \| suggests alternation which is sorta related to sets.
21:44	Chouser	\|[ ]
21:44	Chouser	or maybe not.
21:46	Chouser	^[]
21:46	rhickey	right now it's #{a b c}
21:47	rhickey	#[a b c] seemed to imply that nth would work, to me
21:49	Chouser	ok
21:49	Chouser	also (make-set a b c) or something?
21:49	rhickey	hash-set, sorted-set, contains?, conj, disj, count etc
21:50	rhickey	it's up now
21:51	Chouser	so fast. makes my head spin.
22:19	albino	in python they had to go to a more literal translation set()
22:19	albino	{} for dictionarys [] for lists and () for tuples
23:22	Chouser	I'm so used to using hashes for that, I just forget to reach for sets.