#clojure – 2008-12-30

2008-12-30

00:20	knabb	i figured it out
00:20	knabb	amazing you have to use xml for the buildsystem
00:42	knabb	hmm i m thinking
00:42	knabb	when implementing a stack in clojure
00:42	knabb	push could just cons to a list
00:42	knabb	but pop
00:43	knabb	should it then return the first element of that list and the rest of the list?
00:43	knabb	a stack seems naturally represented with state
00:43	knabb	so a macro to redef the passed variable
00:44	vogelrn	there's already peek that returns the head and pop that returns the rest
00:46	knabb	yes but pop should do both i mean
00:47	vogelrn	well the point is that it doesn't when you're thinking functionally
00:48	vogelrn	instead, you would likely use the head and recurse on the rest, or something of that nature
01:06	knabb	java really is the greatest eample of really poor problem solving
01:06	chrisn	Is there an error function (like failwith or fail in other languages)?
01:06	knabb	solve a problem by fixing it with an even more complex and idiotic problem
01:07	chrisn	nm, found them in contrib
01:07	vogelrn	knabb: What are you referring to out of curiosity?
01:11	knabb	ant
01:12	knabb	xml
01:12	iamslash	I'm a newbie of clojure... using emacs slime... can evaluate expressions... but
01:12	iamslash	slime-edit-definition doesn't work.. on emacs...
01:12	iamslash	what can i start from..?
01:12	knabb	what is the name of the symbol that looks like a weird E and is used a lot in math notation
01:13	vogelrn	you mean the sum one?
01:13	knabb	iamslash: if you use windows try clojure-box, otherways clojure-mode for emacs is easy to install.
01:13	iamslash	I'm using clojure-mode...
01:14	knabb	not lisp-mode?
01:20	iamslash	I installed... git://github.com/jochu/clojure-mode.git git://github.com/jochu/swank-clojure.git
01:21	iamslash	clojure expressions work well... but when I run 'slime-edit-definition'
01:22	iamslash	Emacs debunk "source definition not found' message...
01:22	vogelrn	to be honest I don't actually know what that is, I don't know much about emacs aside from the various tutorials I've read. What are you trying to do?
01:22	vogelrn	slime-edit-definition, that is
01:22	iamslash	yeah right...
01:26	iamslash	http://groups.google.com/group/clojure/browse_thread/thread/e0ece17449ab24ec/5decf3d10d69616f?lnk=gst&q=%22Source+definition+not+found%22#5decf3d10d69616f
02:45	Lau_of_DK	Goood morning gentlemen
02:48	drewr	Good morning.
02:49	Lau_of_DK	Rough night? :)
02:50	drewr	Fever, sore throat. Doesn't look good... :-)
02:51	Lau_of_DK	It sure doesnt, although I know many people having a hard time with that these days
02:53	drewr	Fortunately germs don't travel well over IRC.
02:53	Lau_of_DK	Scientists may disagree on that
10:19	jgracin	just updated SLIME and swank-clojure and now I'm getting an error during startup. (wrong-type-argument sequencep t)
10:19	jgracin	I guess those are related to slime-repl changes. Has anyone managed to fix it?
10:20	jgracin	Or perhaps, does anyone else have this problem?
10:21	aking	jgracin: for a working slime, I had to rollback my git slime to dec23
10:22	jgracin	aking: I see. Thanks, I'll do that.
10:33	Lau_of_DK	Good afternoon gentlemen
10:33	danlarkin	mornin'
10:33	Chousuke	evening
10:34	Lau_of_DK	You guys are either confused, or your nations have not yet been assimiliated by the Empire of Denmark
10:35	karmazilla	My suggestion to allow an alternative doc-string placement have not spured much discusion. I wonder if people are happy with the current allowed syntax
10:35	Lau_of_DK	karmazilla: What was your suggestion?
10:36	karmazilla	to allow the doc-string to come after the [params*]
10:36	Chousuke	karmazilla: wouldn't that create ambiquity?
10:37	danlarkin	but isn't that ambiguous?
10:37	Lau_of_DK	karmazilla: that would be ambiguous
10:37	Lau_of_DK	:)
10:37	Chousuke	what if you wanted a function like (defn foo [] "string")
10:37	karmazilla	no, I defined it in a way that limited it to unambigues usages: http://groups.google.com/group/clojure/browse_thread/thread/fcde7a631a033b7b/faf1a89ab490f582
10:37	Lau_of_DK	But karmazilla, I agree that where it is now, is not very nice, Ive thought about an modification myself
10:38	karmazilla	Chousuke: that should work as it does today'
10:39	danlarkin	I like where the docstrings are now
10:39	Chousuke	I think the docstring is just fine between the symbol and the parameters
10:39	karmazilla	but (defn foo [] "spam" "string") is already allowed syntax, but the "spam" part isn't used for anything
10:40	Lau_of_DK	karmazilla: Is the only advantage of this change prettier code?
10:40	danlarkin	karmazilla: but (defn foo [] (println "side effect!!") "string") is allowed too, same thing as (defn foo [] "spam" "string")
10:40	karmazilla	Lau_of_DK: yes, just prettier code
10:41	Lau_of_DK	ok
10:41	karmazilla	danlucraft: not the same thing... side-effecting thingy != no-op string
10:41	Lau_of_DK	Isnt there already an option of doing something like (defn #^{:doc "This does nothing"} myfunc [x y] nil) ?
10:43	danlarkin	karmazilla: well what I mean is that "spam" in that example isn't a no-op string... it's a form in an implicit do
10:43	danlarkin	it gets evaluated
10:44	Chouser	I think karmazilla is correct that his proposed format wouldn't break anything.
10:45	karmazilla	I don't have the fantasy to imagine a cenario where it would break, anyway
10:45	Chouser	But personally I don't think it matters much -- one more optional way to format defn, slightly more complex implementation and docs, very minor (or no) benefit.
10:46	Chouser	generally, I'm either writing one-off functions and provide no doc string at all and only one body per function. Or I'm trying to write good clean maintainable functions, and I provide doc strings and am much more likely to provide multiple arities.
10:48	Chouser	karmazilla: not to be discouraging, but that's why I've not weighed in on the g.group.
10:49	danlarkin	I'd prefer the possibility for """this style string""" more than this change :)
10:50	rhickey	karmazilla: I'm not interested in supporting that syntax as it adds little and doesn't scale to multiple arities
10:51	rhickey	obviously it was considered and rejected
10:51	karmazilla	Maybe it's just a habit that have this bother me more than most. When I write library code, I try to be rather thorough with my doc-strings, putting in code examples and other stuff, and then it just.... would feel better if I could put the arg list on top
10:51	rhickey	I have pages and pages of variants of fn in my initial design docs
10:51	karmazilla	ok
10:51	rhickey	design is about tradeoffs - multiple arities won here
10:53	Lau_of_DK	karmazilla: I think it was cool that you actually went the extra-mile and provided a patch though, I hope you keep the creativity flowing :)
10:53	duck1123_	It takes a while to get used to (coming from other lisps) but once you see the reason for why it is, it's not that bad.
10:55	karmazilla	I guess I have to keep my python out of my clojure, like I have to keep my haskell out of my python, and my java out of... no, wait.. nothing
10:56	Lau_of_DK	:)
10:56	Lau_of_DK	I agree, keep Python away
10:57	duck1123_	keep the peanut butter and the chocolate together though
10:57	Lau_of_DK	Wise words
11:12	StartsWithK	does anyone know can clojure now be used with osgi? if not, maybe example of using ServiceLoader, it looks like ServiceLoader uses one class loadef for all bundles, so maybe that could work for clojure
11:13	StartsWithK	i looked at cloujredev, eclipse plugins are osgi bundles, but it is in java nad clojure is not separate bundle, and there is no interaction from clojure code with object from other bundles
11:13	StartsWithK	clojuredev*
11:14	duck1123_	forgive me if I'm wrong, but isn't OSGI kinda like a servlet container?
11:16	StartsWithK	something like that
11:17	duck1123_	I know clojure works with Tomcat, so there's hope at least
11:17	StartsWithK	problem last time i had is that every bundle gets his own class loader (or somethin even more complicated than that), so i couldn't..
11:17	StartsWithK	.. create bundle without his private clojure.jar
11:18	StartsWithK	and even with that, if proxy created with clojre would be sent to some other bunde it would fail, i think it remeber its old class loader
11:19	StartsWithK	is this resolver maybe after AOT was added
11:20	duck1123_	you were trying to create classes for a different system with proxy?
11:21	StartsWithK	if object was created in bundle A, then passed as argument to method on object created in bunlde B, then it would fail
11:22	StartsWithK	only 'solution' was that every bundle has everything it needs inside it, but then there was no point in using osgi
11:22	duck1123_	I would definitely try again post-AOT
11:23	StartsWithK	so RT changed the way how it handles class loader?
11:24	duck1123_	I know some things changed in that area... don't know if it solves your problem without looking deeper at it
11:26	StartsWithK	uf, don't want to go on that experimenting road with osgi like last time.. wasted too much time on it before
11:26	StartsWithK	there is also use-context-classloader, and it was in clojure before
11:26	StartsWithK	but its use is undocumented
11:27	StartsWithK	it sugest that something could use another proxy, and not the one in RT
11:44	Lau_of_DK	Yo Pupeno
11:47	Pupeno	Hello Lau_of_DK.
11:48	Raynes	I find it amusing that even in the infancy of this language, there is more people in this channel than there is in #Scala.
11:48	Raynes	:>
11:49	Lau_of_DK	Muhahaha :)
11:49	Chouser	Friendlier ones too.
11:49	Lau_of_DK	Aww Chous', was that directed at me? blush
11:49	Chouser	I imagine having the author here increases the relevence of the channel quite a it.
11:49	Chouser	quite a bit.
11:49	cemerick	Raynes: definitely a good sign, but don't take irc users as an indication of anything like an indication of broader trends.
11:50	cemerick	The masses like the curly braces, regardless of what is found within them...
11:50	Raynes	cemerick: I said nothing about popularity of either language, it's just funny.
11:50	Chouser	Lau_of_DK: take all the credit you'd like. But I was actually uncharitably directing it at others who appeared to intentionally make me feel like an moron while I was trying to make the jump to FP.
11:51	Lau_of_DK	cemerick: perhaps we should consider replacing all these parentheses with curly braces then
11:51	Lau_of_DK	ehe
11:51	Chouser	heh
11:51	Raynes	{+ 2 2}
11:51	Lau_of_DK	{defn add [x y] {+ x y}}
11:51	Lau_of_DK	Looks a bit like barbed wire, kinda cool
11:51	Chouser	I suppose we'd use () for literal maps then?
11:52	Lau_of_DK	Chouser: Really? That sounds awful
11:52	cemerick	it's sorta rubyish, really. { somefn arg arg }
11:52	Lau_of_DK	FP is a big enough steak on its own
11:52	cemerick	=> reader syntax, baby!
11:52	Lau_of_DK	Perl!
11:53	vogelrn	I think #[] is open for literal maps, isn't it? :P
11:53	vogelrn	avoid using the parenthesis altogether
11:54	Chouser	vogelrn: truly evil, my friend.
11:54	ericlavigne	We can have <+ x y> and replace the <,> functions with lt,gt.
11:56	danm_	scala seems strange to me, like they don't need a macro system because any language feature you could ever hope for is either already in the language, or slated for the next release
11:57	ericlavigne	Or rather...
11:57	ericlavigne	<
11:57	ericlavigne	+ x y
11:57	ericlavigne	>
11:57	ericlavigne
11:58	danm_	call by name parameters are an interesting way to avoid the need for macros as well
11:58	ericlavigne	Because not giving parentheses/braces their own line is a dead giveaway that you're using Lisp :-P
11:59	Lau_of_DK	If we could somehow implement this it would be great: http://99-bottles-of-beer.net/language-perl-737.html
12:01	Lau_of_DK	hehe, you should see the Malbolge solution then
12:01	Lau_of_DK	HAHA: http://99-bottles-of-beer.net/language-lolcode-1544.html
12:04	Lau_of_DK	And an extra fact, Clojure is actually on the top10 on that site
12:21	mibu	does anyone here know how to use the clojure.parallel lib?
12:25	aking	When inside a proxy function, how can I do a "return this;"?
12:30	mibu	aking: I think just "this".
12:32	aking	mibu: thanks - seems to compile - I test in a sec. So, 'this' is a reserved word? I don't see it mentioned on the java interop page.
12:32	rhickey	aking: no, 'this is just bound by the proxy macro
12:32	mibu	aking: it's not reserved. it's just implicitly bound to this
12:33	rhickey	hmpf, no java.beans in Android
12:34	aking	ahh.. thanks - rereading the proxy doc mentions it - sorry for the noise.
12:34	inertia-	rhickey: clojure on the llvm, possible? certainly seems worth it
12:34	rhickey	inertia-: with what libraries?
12:35	inertia-	itself - perforamnce is what i am thinking of
12:35	rhickey	inertia-: better to target a fast CL
12:36	rhickey	only perf benefit over JVM is tagged fixnums
12:36	rhickey	without polluting Clojure with primitive types
12:36	inertia-	isn't there a greater potential for perf increases over time?
12:37	inertia-	i'm not saying i'm a performance whore but i certainly think about it
12:37	rhickey	LLVM is not nearly the same infrastructure as JVM/CLR
12:37	inertia-	(mainly that java is always 1-4 times worse than C on alioth benchmarks)
12:37	rhickey	I'm completely not interested in reinventing wheels
12:37	inertia-	:)
12:38	inertia-	i like clj very much, thanks rh
12:38	rhickey	inertia-: you're welcome!
12:42	zakwilson	I think it would be cool if somebody implemented languages similar to Clojure for other platforms, but Clojure is very clearly tied to the JVM. It took me a long time to come around, but I think that's a Good Thing.
12:43	rhickey	zakwilson: Clojure could easily target the CLR, and once did
12:43	rhickey	could probably target any OO host with MT primitives
12:44	rhickey	or drop MT, as ClojureScript does in targeting Javascript
12:45	rhickey	ugh, jar utility cannot remove files ?
12:55	Chouser	(complement neg?) vs. #(not (neg? %)) vs. #(>= % 0) ...opinions?
12:58	mibu	chouser: not neg if you have a literal style, >= if you're the compact type.
12:59	mibu	chouser: for me, it's a matter of mood.
13:08	rhickey	Hello Android World! (from Clojure)
13:08	mchurch	What is the best/preferred way to spawn a thread in Clojure?
13:10	karmazilla	mchurch: if not with an agant, then maybe (doto (Thread. #(println "pokemon")) .start) ?
13:15	mchurch	karmazilla: I haven't gotten to agents yet in my reading, but they sound cool.
13:15	mchurch	karmazilla: I'm just looking for very basic thread functionality now (prototyping something) so that will probably be enough.
13:17	rhickey	danlarkin: looks like bean is the only function that uses an unsupported API (java.beans)
13:19	rhickey	I'll check in as soon as I figure out how to handle bean (had to comment it out to get Hello Android to run)
13:19	aking	cool - will try it as soon as it's in
13:19	aking	rhickey: what did you do for the classloader? That patch I submitted - or a proper fix?
13:20	rhickey	aking: a proper fix :)
13:20	aking	heh :)
13:20	rhickey	the problem with the patch is it ignores the fact that the classloaders should have delegated
13:21	rhickey	rather than just workaround, I dug into it and found how Android was responding to contextLoaders etc
13:21	rhickey	so now the baseloader is properly rooted on Android and delegation works
13:22	aking	aye - thought that should have been the case, but I'm not familar enough with how clojure handles the classloaders - but hope it at least pointed at what was wrong
13:22	rhickey	also got the bridge method resolution fixed up, plus special handling for the source being present but the .class not
13:24	aking	once you make an announcement on the list, it'll be time to progit/digg it - it's been a few days since there's been a clojure story
13:27	rhickey	aking: I'd rather it worked smoothly before it got a lot of attention
13:27	rhickey	but support is up (svn 1192) - you'll have to comment out bean in core_proxy.clj manually if targeting Android, for now
13:28	rhickey	have fun!
13:32	aking	rhickey: great! I'll update and try it out after lunch
13:36	Nafai	rhickey: You got it working?
13:36	rhickey	FYI, what I got running was this: http://github.com/Nafai77/helloandroid/tree/master, with no changes to it other than adding <arg value="-JXmx1024M" /> to the dex target in build.xml since I was getting out of mem errors there
13:36	Nafai	Sweet!
13:37	rhickey	Nafai: you'll need to comment out bean in core_proxy.clj before building Clojure
13:38	Nafai	Just the bean method?
13:44	Nafai	rhickey: The bean method?
13:46	rhickey	the entire bean function in core_proxy.clj
13:46	rhickey	no bean on Android
13:46	Nafai	Ok
13:46	Nafai	Right
13:48	replaca	rhickey (et al.): congrats on getting Android working!
13:53	Nafai	Still getting a no class def error
13:54	aking	Nafai: I'm updating now...
13:59	aking	Nafai: works for me :)
14:00	aking	woohoo - time for some Android clojure coding..
14:00	aking	did you copy the updated clojure.jar file to hellowandroid/lib ?
14:00	Nafai	aking: Which Android sdk version are you using?
14:00	Nafai	aking: Yeah, it's symlinked
14:01	aking	I'm running it under WinXP - android-sdk-windows-1.0_r2
14:01	Nafai	I'm on Linux, but same version
14:01	Nafai	I'll have to play with it later, at work :)
14:02	aking	Nafai: if it's the same class def error, sounds like it's not picking up the updated jar file
14:02	Nafai	Yeah, it does sound that way
14:18	Lau_of_DK	Good evening Mr. Schulz
14:18	aking	Nafai: got a ddms screen capture of it running if you want to see it. Memory usage isn't to bad - a basic clojure android app seems to take about 2.6Mb
14:19	Nafai	aking: Yeah, I'd like to see it :)
14:19	Lau_of_DK	Me too
14:21	AWizzArd	clojurebot: max people
14:21	clojurebot	max people is 116
14:21	aking	Nafai: see it here: http://antimass.org/aking/clojure-android.png
14:22	Nafai	Yay!
14:24	AWizzArd	aking: is this running in a Java ME or does your phone support a full JRE?
14:24	aking	It's running under Google's Android vm
14:24	aking	the emulator - but should work fine on an Android phone
14:25	mibu_	rhickey: do you plan in the future to implicitly integrate operations from clojure.parallel in the standard operations?
14:26	AWizzArd	aking: I ask because to my knowledge Clojure currently does not run inside a ME as those guys have too strict memory settings, something like that. So it seems that the Android vm uses some settings that make more sense for todays hardware.
14:27	cemerick	Chouser: did anything concrete come out of the pretty-printer discussion last month?
14:27	cemerick	(I just want to use one, too :-) )
14:28	Chouser	mibu_: requires a third-party .jar, so it'd surprise me. But there's a pmap in core
14:28	aking	AWizzArd: I haven't tried under ME - but I don't see why it shouldn't work. rhickey has to remove bean support for android, which I believe ME is also missing, so it might start working.
14:28	Chouser	cemerick: nope, we've got two (unfinished?) implementations.
14:28	replaca	cemerick, Chouser: I was just thinking about digging in on a pretty printer last night
14:28	aking	AWizzArd: That clojure android app is running Nafai's code here: http://github.com/Nafai77/helloandroid/tree/master
14:29	mibu_	chouser: I understood the ForkJoin lib will be included in Java7
14:29	replaca	I'm pretty much done with common lisp format and that leads kind of naturally to the XP pretty printer
14:29	Chouser	mibu_: ah, I didn't know that.
14:29	replaca	are others doing active work?
14:29	Chouser	replaca: I was only working on pprint because I thought nobody else was. I've dropped it for now.
14:30	cemerick	replaca: not in this corner. All of my tinkering has fallen by the wayside :-(
14:30	Chouser	I must say, though, that the CL-style format feels like quite a burden.
14:30	replaca	I just started reader the MIT tech note on the XP pretty printer last night (the basis of the PP in CL)
14:30	replaca	Chouser: in what sense?
14:31	Chouser	replaca: it's like a whole language in itself -- not even an s-expression based one -- and so to use it or read code that uses it I need to learn the parsing rules, functions, the whole works.
14:32	cemerick	yeah, FORMAT is pretty rough for the user -- mini languages aren't fun
14:32	replaca	Chouser: indeed. it's oonly for those that like it. But it does giive you better control and simplifies your printing code. So it's a trade-off.
14:33	cemerick	stuff like this scares the bejeezus out of me, for instance: http://cybertiggyr.com/gene/fmt/
14:33	Chouser	When I first started on a pprint, I asked if XP as a good model and rhickey_ said it seemed overly complex. iirc.
14:33	replaca	The nice thing about XP is that it's already thought out -> therefore easier to implement
14:34	cemerick	Java already has a text formatter -- if that can be folded in transparently, then we'll all be better off.
14:34	replaca	(Cause you don't have to worry about unforeseen corner cases)
14:34	Chouser	perhaps we could come up with something that has the internal structure and feature set of FORMAT, but uses an s-expr API instead?
14:35	replaca	One thing that would be pretty easy is to sexpr the bracket constructs
14:35	replaca	It would make it more verbose, but easier to understand the flow
14:36	replaca	The nice thing about format is that you don't have to explicitly traverse your structures
14:36	replaca	This argument has been going on since the eighties, of course
14:36	Chouser	:-)
14:37	replaca	I've always liked format cause it get printing "out of the way"
14:37	Chouser	but clojure gives us new tools that CL doesn't have, so it may be worth at least skimming over the argument one more time.
14:37	replaca	And I mostly don't care that the formats are kind of write-only
14:37	cemerick	replaca: even when you need to fix a bug?
14:38	replaca	Well, the bug would only be in the format itself, so you can focus on that. But the point is valid.
14:39	replaca	I mostly did the format implementation just cause I wanted to (god knows why)
14:39	replaca	But there's lots of raw material there to support other mechanisms
14:39	replaca	and of course we have the java mechanism "built-in"
14:40	replaca	and certainly one reason to have format "lying around" is that it makes porting CL code easier
14:40	replaca	cause converting the prints seems like a silly way to spend time
14:41	replaca	all of which is basically orthogonal to pretty printing :-)
14:44	AWizzArd	cemerick: why are you scared by format? I mean, it's basically a DSL just as regexps are. Or maybe you try to avoid those too?
14:44	gnuvince	regexes should be avoided when possible :)
14:44	Chouser	heh. well, I certainly don't mind having a lib around that is 100% backward-compatible with CL for porting. Seems very nice for that use case.
14:45	Chouser	AWizzArd: I was thinking about that -- if we could come up with an s-expr-based text matching language that was no more than 2 or 3 times as verbose as regex, I'd probably prefer that too.
14:45	replaca	gnuvince: yes, but we don't avoid them when they make our intent clearer.
14:46	replaca	Chouser: yes, but good luck with that. sexprs are pretty verbose relative to these tight string languages
14:46	cemerick	AWizzArd: format does way, way more than I've never needed such a facility to do. It's too clever for its own good, IMO. regexes solve an immediate and constant problem, and no better option has presented itself, so they remain.
14:46	Chouser	it might be worth looking at how perl6 approaches regex -- they redesigned to be much better at composition of patterns and such, at some cost in verbosity.
14:47	replaca	The advantage of both regex and formats (both CL and Java/printf-style) is that they work in the target domain: strings.
14:48	AWizzArd	Chouser: fair enough
14:48	AWizzArd	cemerick: well, in todays world it's probably really not too common anymore to print out texts into a shell or something like that. Most programs come with a gui and for those there are other tools.
14:49	AWizzArd	I don't see it as a real loss to not have format.
14:49	gnuvince	format can return a string which you can put in a message box, label, whatever.
14:50	cemerick	I'm sure one's opinion is informed by the problems that one needs to solve. At least in my world, simple parameterization of a string is almost always sufficient.
14:50	cemerick	ah, parameterization, and (str (interpose "," seq)), etc.
14:54	drewr	What am I missing here? (map Math/sqrt [1 2 3])
14:55	Lau_of_DK	(map #(Math/sqrt %) seq)
14:57	drewr	Lau_of_DK: Thanks. My brain was failing me there.
14:58	drewr	I knew I had to promote it to an IFn but I was doing identity for some reason.
14:59	Lau_of_DK	No probs, imho Clojure should eat Java directly :)
14:59	AWizzArd	But this is only because Math/sqrt is a Java method and not a Clojure function.
15:00	AWizzArd	(map clojure-fun col) is equivalent to (map #(clojure-fun %) col)
15:00	AWizzArd	or not?
15:01	Lau_of_DK	Yes sir
15:07	Chouser	as long as you're only providing a single col argument to 'map'
15:09	drewr	I wonder if it would be too magical to automatically promote things like Math/sqrt.
15:09	drewr	Maybe static methods only?
15:10	drewr	They act functionally.
15:16	Chouser	that's been brought up before. I think rhickey_ was open to the idea, though I'm not finding a reference right now.
15:16	jgracin_	I have what is probably a silly question, but I've got to ask it anyway. Is there any way to compile a clojure file which references some Java classes, but without having those Java classes in the classpath?
15:17	ericlavigne	Why only static methods? Ball.throw (non-static) could be a function whose first argument is a ball, while ball.throw would be a similar function with one less argument (already received the ball object as its first argument).
15:18	ericlavigne	jgracin_: perhaps you could refer to an interface which those java classes implement?
15:19	chrisn	I like the way ericlavigne is going
15:19	chrisn	That exact feature would have made a lot of my opengl code a bit cleaner
15:19	jgracin_	ericlavigne: I'm trying to see if I can mix Clojure and Java code which have interdependencies and therefore cannot be compiled sequentially.
15:20	Chouser	jgracin_: you definitely need access to any classes you extend or interfaces you implement.
15:20	jgracin_	e.g. how would I compile A.java b.clj C.java in which A.java depends on code defined in b.clj and C.java, and b.clj depends on C.java
15:21	jgracin_	Chouser: I see
15:22	Chouser	gen-class uses introspection to poke around in the superclasses and figure out what it needs to do
15:22	ericlavigne	jgracin_: you can avoid such cyclic dependencies by breaking your code into smaller pieces.
15:22	Chouser	what jgracin_ described isn't cyclic
15:22	chrisn	ericlavigne: perhaps what you describe is how partial should work w/r/t java interop?
15:23	Chouser	jgracin_: I think you might be able to get where you want to go with careful use of ant and tasks for compiling .clj files.
15:24	cemerick	hrm, clojure.xml/emit doesn't do any string escaping at all...
15:24	Chouser	jgracin_: you might also look into using gen-interface (which I think doesn't need any of the classes it references) early, then javac all your .java files, then use 'proxy' in your .clj at runtime.
15:25	Chouser	cemerick: yeah, that came up on the g.group. It's fixed in clojure.contrib.lazy-xml
15:26	jgracin_	Chouser: Interesting idea, I'll try that. Thanks!
15:27	cemerick	Chouser: is lazy-xml being considered as a replacement for clojure.xml, or is an analogous fix going to slip in there eventually?
15:27	Chouser	Perhaps I should post an issue
15:27	Chouser	cemerick: I'd like lazy-xml to replace clojure.xml, but I've got no indication any such thing is planned.
15:29	Chouser	clojure.xml has a couple correctness issues. string escaping, removing whitespace on parse, adding whitespace on emit
15:29	cemerick	yeah, it seems very prototypey
15:30	ericlavigne	chisn: I am saying that we should think of the object to which a method belongs as being simply the first argument in a function. As a shortcut for Java methods, we should be able to say obj.method which implies partial application of obj to Obj.method.
15:30	Chouser	lazy-xml fixes a couple of those, though it still always trims whitespace.
15:31	cemerick	Chouser: my brief run-ins with xml wizards have left me with the impression that whitespace handling is a decidedly tricky/ambigious thing when it comes to serialization
15:31	chrisn	ericlavigne: I am checking out memfn...
15:32	chrisn	doesn't work for statics
15:34	ericlavigne	Maybe memfn expects you to pass something like Math.class as the second argument.
15:35	ericlavigne	I haven't used it.
15:35	cemerick	I can see how lazy xml parsing might not be an ideal standard approach, though. I certainly always want the whole doc, so the laziness is just overhead for me.
15:39	ericlavigne	chrisn: I speak of the way I would do it. However, I don't yet know enough about Clojure to know if a better way has already been implemented. I'm certainly quite impressed by the parts of Clojure that I have used.
15:39	chrisn	I am in the exact same boat.
15:40	chrisn	Once I get my demo project finished and lightly polished, I intend to go through the core source code and then through the contrib source code.
15:40	chrisn	I always ask questions that just require more knowledge of the apis that came with the system.
15:40	ericlavigne	The proxies seemed overly complicated, so when I needed to create Java classes I just wrote some Java instead.
15:41	chrisn	I just forced that bastards to work my way ;).
15:46	chrisn	I wish you could create immutable classes from clojure. I am using defstruct all over the place and every once in a while I mistype a keyword. Which is a runtime error...
15:50	Chouser	chrisn: defstruct may produce a class with fields in the future.
15:51	Chouser	cemerick: don't you do something clever along those lines?
15:51	chrisn	Chouser: Is it possible to get compile time errors for certain classes of problems? I know, dynamic language blah blah, but I am not always using it in a dynamic way.
15:52	cemerick	Chouser, chrisn: yeah, genbean. It's evolved quite a bit since my last post on it some months ago.
15:52	cemerick	I'll probably post an update after clojure goes 1.0 and I've updated it to match.
15:52	Chouser	chrisn: you can throw and exception from a macro
15:52	Chouser	s/and/an/
15:52	chrisn	yep, good point
15:53	chrisn	I have buddy doing work on partially typing lanuages.
15:53	chrisn	Adding in hindley milner to python, but only where you have types
15:53	chrisn	works pretty darn well.
15:53	cemerick	of course, the really nice stuff still requires that you drop any static type checking when creating the objects in question (e.g. when providing a delay for a slot in an object, rather than a strict value)
15:53	Chouser	chrisn: interesting. I've thought for quite a while that i wanted a language with optional static typing.
15:55	chrisn	I personally think types rock. I love Haskell and F#. I just love getting shit done a lot more ;) . Plus Lisp code just looks better.
15:55	Chouser	Scala convinced me I do not want required static types. :-)
15:56	chrisn	Heh. With F#, you can always cast something to an object. So you can easily defeat the type system.
15:56	chrisn	I don't know about Scala.
15:57	Raynes	We were just having a conversation in #Scala about why F# is horrible.
15:58	chrisn	oh jesus
15:58	chrisn	here we go
15:58	cemerick	indeed. It's like a black hole.
15:58	chrisn	why is F# horrible?
15:59	cemerick	maybe we can all just agree that all languages are horrible, and we'd all be better off running movable type presses.
15:59	cemerick	or something :-P
15:59	chrisn	anyone tried newspeak? then you can plug in your own type system.
16:00	durka	hey kotarak, i'm having strange problems with gorilla
16:00	kotarak	Uh? Strange problems? That's no good.
16:01	durka	i start gorilla and do \sr in macvim
16:01	durka	gorilla shows "Connection started"
16:01	durka	macvim freezes
16:01	durka	macvim and java quickly climb to the top of CPU usage
16:02	durka	if i kill gorilla (ctrl-c), macvim unfreezes and shows the repl, which of course doesn't work because gorilla is dead
16:03	kotarak	huh?
16:03	kotarak	That's strange.
16:04	kotarak	Did you change something to get this behaviour?
16:04	kotarak	(Other vim, ruby, clojure, contrib?)
16:04	Nafai	Doh
16:04	Nafai	Figured out my issue. :)
16:04	Nafai	I did 'git svn fetch' instead of 'git svn rebase' :)
16:04	durka	kotarak: i updated clojure and contrib to the latest svn
16:05	kotarak	Hmmm... Will check. I had no problems up to now.
16:15	durka	kotarak: i have to go, but mention my name and i'll see it
16:16	kotarak	ok. Just figuring out.
16:16	kotarak	Some change in Clojure broke it.
16:40	RSchulz	kotarak: You should check out this topic from the group: http://groups.google.com/group/clojure/browse_thread/thread/93e35885b0c0706d?hl=en
16:40	RSchulz	It may be germane to what's happening to you in gorilla.
16:48	kotarak	RSchulz: thanks for the pointer
16:48	kotarak	Will investigate.
17:12	mchurch	I have a noob question that is probably not clojure related. I'm trying to create a socket. This: (def sock (new java.net.Socket "127.0.0.1" 5023)) : throws java.net.ConnectException : Connection refused. Why could that be?
17:13	mchurch	Is it something wrong in my code, or is the problem at the system level?
17:14	karmazilla	apparently, nothing is listening on your localhost port 5023
17:16	RSchulz	mchurch: You're aware that that particular java.net.Socket constructor immediately attempts to connect to the spcified host and port, right?
17:16	mchurch	RSchulz: I think I've figured it out, because ServerSocket could be instantiated.
17:16	mchurch	So I instantiated a ServerSocket on that port, then a Socket, and the socket connected.
17:17	mchurch	So, I guess they're connected. I didn't realize that I needed a ServerSocket before a Socket.
17:17	RSchulz	You're attempting to connect to a listening socket in the same process?
17:17	RSchulz	Naturally, you'll need threads / Agents to make that work.
17:17	karmazilla	or non-blocking IO
17:17	mchurch	RSchulz: No, 2 different Repls
17:17	RSchulz	Yes, there has to be socket listening to accept the connection request.
17:18	mchurch	I'm just trying to get a handle on sockets before I do the next stage of my current projects
17:18	mchurch	I didn't know what a socket was a couple weeks ago.
17:18	RSchulz	It's OK. Sockets didn't know about you, then, either.
17:18	RSchulz	Sockets, meet mchurch. Mchurch, sockets.
17:18	RSchulz	There. Now you've been properly introduced.
17:18	RSchulz	Anyway, what sort of protocol are you working towards?
17:19	RSchulz	You might want to work with slightly higher-level API such as the ones in Commons HTTP Client and HttpServlet.
17:21	mchurch	This is for an RPC engine.
17:23	karmazilla	also, for client code, you'd typically want to set timeouts before connecting, and thus not use a Socket constructor that connects automatically
17:23	mchurch	karmazilla: Ok. I'm just prototyping right now.
17:23	RSchulz	Is th wire protocol already defined? Is it an external requirement? Or are you inventing it? Or using HTTP?
17:24	RSchulz	Distributed programs are so non-tivial, it's almost always a good idea to build upon as much infrastructure / framework / etc. as you can.
17:25	Chouser	...unless your homework assignment is to implement an RPC engine from the Socket level up.
17:25	RSchulz	Oh, to be in school again...
17:29	mchurch	RSchulz: The protocol's already written. We have an RPC engine already written in SBCL and I'm writing a Clojure version of it.
17:29	RSchulz	I see. Well, at least you don't have to think about that part. Is it text or binary?
17:29	mchurch	RSchulz: Binary
17:29	RSchulz	Odd.
17:30	RSchulz	Why?
17:31	mchurch	Oh. One unrelated question. Sorry, but this has been bugging me: regarding import, what's the Clojure analogue of import java.io.*?
17:31	mchurch	In the event that I want to import a whole package rather than just a list of classes.
17:32	RSchulz	There's no "wild-card" import in Clojure.
17:32	mchurch	RSchulz: So you have to list all the classes?
17:32	RSchulz	Yes.
17:33	mchurch	RSchulz: Got it. Thanks.
17:33	RSchulz	I personally don't think wild-card imports are a good idea.
17:34	mchurch	RSchulz: What do you dislike about them?
17:34	RSchulz	You don't know what you're saying. In large systems that use many libraries, there can be clashes and you can get odd compilation errors.
17:34	RSchulz	I think you should say everything you mean, basically.
17:35	mchurch	RSchulz: Ah. Yeah, I can see that being a problem.
17:35	RSchulz	You could conceivably even get sucessful compilation and run-time errors (during dynamic linking, most likely).
17:36	RSchulz	Clojur at least doesn't make you say the package part over and over for each class from that package.
17:36	mchurch	RSchulz: Yeah. That would be horrible.
17:37	RSchulz	Realistically, it's not likely, though I've had the compile-time manifestation quite commonly in my own code.
17:37	RSchulz	I just couldn't see a better name for one of my classes than "Set", and so you can see the problem.
19:06	durka	kotarak: tracing gorilla reveals that after \sr gorilla goes into an infinite loop of de.kotka.gorilla.repl-ln/read-hook called with nil
19:15	lisppaste8	durka pasted "gorilla trace" at http://paste.lisp.org/display/72852
20:57	MarkVolkmann	Is there an absolute value function in Clojure? I've looked but don't see one.
20:57	MarkVolkmann	I know it would be easy to write my own, but don't want to do that if I don't have to.
20:59	pjb3	Math/abs
21:00	pjb3	That's the java one, but I've heard people complain because it doesn't work with all numeric types in clojure
21:00	MarkVolkmann	Oh right! I forgot to just use Java.
21:01	ericlavigne	pjb3: Also, methods aren't functions. It's always nicer to use something from native Clojure.
21:02	pjb3	ericlavigne: Is there a clojure function/macro that will return a function that wraps a method?
21:03	ericlavigne	I think that's what memfn is for, but I haven't used it yet.
21:04	pjb3	user=> (map (memfn Math/abs) [-3 3])
21:04	pjb3	java.lang.IllegalArgumentException: No matching method found: abs for class java.lang.Integer
21:04	ericlavigne	I think it's trying to find an Integer method called abs in the Math package :-(
21:05	ericlavigne	Maybe something like Math.class needs to be provided as another argument to memfn.
21:06	dreish	memfn predates #() notation, and I believe it is generally considered obsolete now.
21:06	Chouser	,(map #(Math/abs %) [-3 3])
21:07	dreish	But that's unrelated to the fact that 3 is an int, and Math/abs takes a double.
21:07	pjb3	(map #(Math/abs %) [-3 3])
21:07	pjb3	Right
21:07	Chouser	Or: (map #(.abs %) [-3M 3M])
21:07	ericlavigne	I thought Math.abs was overloaded to take a variety of possible arguments.
21:08	durka	int, long, float, double
21:09	dreish	,(Math/abs -3)
21:09	dreish	clojurebot is being shy, but that gives me 3 at my repl.
21:16	durka	so shy that it left the channel
21:26	pjb3	What's the easiest way to convert a map to a hash map?
21:27	Chouser	you can use a java.util.Map in a lot of places that you'd normally use a hash map.
21:27	pjb3	In this case, I have a function that will passed a map
21:28	pjb3	and I will need to do some stuff with it, passing through it multiple times
21:28	pjb3	each time I pass through it, it needs to be in the same order
21:28	pjb3	I don't care what the order is, as long as it's the same each time
21:29	Chouser	java.util.HashMap doesn't guarantee that?
21:30	pjb3	The map passed to the function could be an ordinary map
21:30	pjb3	I want users to be able to do
21:30	pjb3	(myfunc {:x 1 :y 2})
21:31	pjb3	Actually now that I think of it, I can probably just use (seq themap) in my function
21:32	Chouser	yes, I would think so.
21:32	pjb3	because I don't need to use it associately while processing it in my function
21:32	pjb3	associatively
21:35	Chouser	(into {} themap) if you really want a PersistentHashMap
21:37	pjb3	Wouldn't it have to be (into (hash-map) themap)?
21:38	Chouser	that's fine too
21:39	pjb3	What's the difference between array map and hash map?
21:39	pjb3	or right, the array map preserves insertion order
21:40	pjb3	In clojure {} gives you an array map?
21:40	pjb3	Has it always been like that?
21:41	pjb3	I thought the default was hash map
21:41	Chouser	hm...
21:42	rhickey	pjb3: array map preserves initial order, multiple insertions will cuase it to transition to hash-map
21:43	rhickey	cause
21:44	pjb3	multiple being some large number?
21:44	pjb3	user=> (class (assoc (assoc (assoc {} :a 1) :b 2) :c 3))
21:44	pjb3	clojure.lang.PersistentArrayMap
21:45	rhickey	user=> (def m {1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1})
21:45	rhickey	#'user/m
21:45	rhickey	user=> (class m)
21:45	rhickey	clojure.lang.PersistentArrayMap
21:45	rhickey	user=> (class (assoc m 9 1 10 1))
21:45	rhickey	clojure.lang.PersistentHashMap
21:45	pjb3	I see
21:46	pjb3	That makes sense, has it always been like that or was that a change recently?
21:46	rhickey	always been a transation, recently made sure small maps were array maps, including {}
21:46	rhickey	transition
21:47	pjb3	For some reason I remember doing (array-map :x 1 :y 2) for small maps that I needed to have a specific order for
21:48	pjb3	What happens if you have a large map that you need insertion order preserved for?
21:48	Chouser	you can still do (apply array-map coll)
21:51	pjb3	Ok, and then you just need to make sure to know that if you assoc onto that, you'll get back a hashmap
21:51	Chouser	right
22:31	rhickey	large array maps are a bad idea perf-wise
22:51	gnuvince	Grrr... I like Haskell and all, but sometimes they can make things so hard to understand...
22:52	Nafai	rhickey: Thanks for all of the hard work getting Clojure working on Android, you rock.
22:52	Nafai	rhickey: I verified it also works on an actual device :)
22:53	Nafai	I've updated the README that is on github that includes the svn revision number and also to comment out the bean function
22:56	gnuvince	Nafai: you mean, Clojure can be used to create application on Google's Android?
22:56	Nafai	gnuvince: Yup!
22:56	gnuvince	That.s so cool :)
22:57	Nafai	http://github.com/Nafai77/helloandroid/tree/master
22:58	aking	Nafai: you might also want to add a comment about increasing the heap size when compiling with dx
22:59	aking	Nafai: or just update the build.xml :)
22:59	Nafai	aking: I didn't have to do so
22:59	aking	Rich did - and so did I - runs out of mem on XP
22:59	Nafai	aking: What change did you have to make? I can make it and upload it
22:59	Nafai	Yeah, I'm on a 64-bit vm on Linux :)
23:00	aking	From rhickey earlier: FYI, what I got running was this: http://github.com/Nafai77/helloandroid/tree/master, with no changes to it other than adding <arg value="-JXmx1024M" /> to the dex target in build.xml since I was getting out of mem errors there
23:02	Nafai	Ok, cool
23:02	Nafai	I'll push that change
23:02	Nafai	I might start porting over some of the sample applications to Clojure
23:03	aking	Nafai: I'm running emacs/slime under XP with a remote repl to my 64bit linux box for most of my clojure work - but used XP only for ANdroid testing. Eventually I'll get Android setup under linux
23:03	Nafai	Cool
23:06	Nafai	Unfortunately, ddms doesn't run on 32-bit Linux at the moment, since it links against 32-bit swt librarie
23:06	Nafai	Change pushedd
23:07	aking	Nafai: you mean "ddms doesn't run on 64-bit linux"?
23:08	Nafai	Yeah, :)
23:15	gnuvince	Hmmm
23:16	gnuvince	From Real World Haskell "Fewer parentheses leads to reduced mental juggling while reading a function."
23:16	gnuvince	I'd disagree with that: fewer parentheses means you need to remember the proper operator priorities
23:19	mattrepl	and mentally decoding $
23:21	Chouser	looking back at my Scala code, I'd have to agree.
23:23	Chouser	Reducing parens does give an overall look of simplicity -- it does not appear to be tangled or deeply nested.
23:23	Chouser	but both are illusions. It may be tangled and deeply nested, you just can't tell. In fact it is quite difficult to tell what the heck is going on at all.
23:25	falconai	any one have a quick example of how to use the "read" function? (read "(doc read)") obviously doesn't return any data structure
23:25	rhickey	I find Haskell particularly painful to read due to precedence etc - I'd much rather have sexprs
23:25	Chouser	(read (java.io.PushbackReader. (java.io.StringReader. "(doc read)")))
23:26	rhickey	falconai: (read-string "(doc read)")
23:30	falconai	so it looks like a sequence is returned by read-string...but it has to be a tree where as sequences are usually linear ... how do i traverse this tree?
23:31	Chouser	falconai: no, it's returning whatever literal object the text represents
23:32	Chouser	list, map, vector, etc.
23:33	falconai	i see...what i am trying to do is the following:
23:34	falconai	scheme's frtime (a CELLs like data flow framework) let's one do something like this: (even? 2) which returns true (like clojure) (even? seconds) switches true/false every second
23:35	falconai	scheme does this by extending its evaluator...i guess in clojure it would mean that between reader's output and the compiler, i would have to get access to the forms, manipulate them and send them on to the compiler...
23:36	falconai	does it make sense?
23:37	Chouser	I suppose.
23:38	Chouser	there's code in clojure.contrib.walk for walking arbitrary clojure s-expr trees and making replacements.
23:38	Chouser	you could run the reader yourself, I suppose, and then process the results and finally eval.
23:39	Chouser	...or you could write macros, which are given the already-read s-expr trees to work with
23:39	falconai	does walk operate on character based s-expr or forms returned by reader?
23:40	Chouser	...or you could take advantage of built-in clojure features to get what you want without messing the the structure of the programs
23:40	Chouser	'walk' operates on the data-structures returned by 'read' which are the same as what are passed to macros
23:40	Chouser	falconai: there has been some discussion of "cell" implementations and "reactive programming" on the google group, including a couple of implementations.
23:42	falconai	right, i saw that (I think my original message might have started the main discussion)...but as far as I can tell, frtime is a more robust system...several years of use, several papers and a dissertation behind it
23:42	Chouser	ok
23:44	Chouser	starting with 'read' seems like a hard way to go.
23:45	falconai	so a reactive expression such as (even? seconds) needs to be converted to a non-imperative expression involcing agents and watchers...but the end user should only see the reactive version (and not code related to agent watchers)...
23:45	falconai	what is a good way to do that? (keeping in mind that reactive expressions can turn programs "inside-out")
23:46	Chouser	would a macro be unacceptible? (react (even? seconds))
23:46	falconai	...i have never programmed in clojure, scheme, CL, etc....so I am starting as ignorant as can be
23:47	falconai	I think (react (even? seconds)) may end up being the pragmatic solution...but what about when these reactive values are nested: (react (even? (+ (* temperature 100) seconds))) where temp and sec are reactive...will i be able to use
23:48	falconai	macros to 'walk' the expression tree and get at the reactive values?
23:48	Chouser	yes
23:49	falconai	good, encouraging...is there a way to make the REPL go through my macro for each new expression? (or is that the same as messing with the reader?)
23:49	Chouser	a macro named react in the above example would be given the entire tree of s-expressions, and would be expected to return a new tree of s-expressions to be evaluated in their place.
23:49	falconai	ideally, i'd still like to avoid telling clojure explicitly about reactive values...i'd like them to be transparent (which may not be possible in a clean way)
23:51	Chouser	well, you could write your own repl (not terribly hard) and process all expressions after 'read' and before 'eval'. But wouldn't it then be impossible to provide any non-reactive expressions?
23:51	falconai	Chouser: is textjure some sort of graphic REPL? for some reason i get that impression...and i'm very interested in something like it (lifting clojure's environment to something similar to spreadsheets)
23:52	Chouser	I plan for textjure to include a repl and a text editor using Swing widgets instead of a plain text terminal.
23:54	falconai	sounds good, do you have an ETA on it? :)
23:55	Chouser	well, the repl largely works already, but without a good text editor I don't think it provides much benefit.
23:56	Chouser	I want vi-like keybindings, and passable syntax highlighting and indentation support.
23:57	Chouser	right now I'm spending most of my "clojure time" fixing up the web page docs, but I hope to get back to textjure in another week or two.
23:58	falconai	speaking of which, http://clojure.org/repl_and_main is empty
23:59	Chouser	yeah, rhickey_ added a few blank pages. I'm not sure exactly why. :-)
23:59	gnuvince	probably to have people fill them up ;)