#clojure – 2009-07-30

2009-07-30

00:11	Jomyoot	Anyone uses BBEDIT?
00:32	mebaran151	any hack to assign metadata to a java object
00:32	mebaran151	it would be extremely convenient in this one case
02:23	Fossi1	freakin damn lazyness
02:23	Fossi1	it got me again
02:23	Raynes	Fossi1: In order to beat the lazy, you must first, think like the lazy.
02:24	Raynes	Kick your feet up, grab a remote and a cup of coffee and put your hands behind your head for an hour or too. Once you have done this, you will be ready.
02:25	_mst	when someone asks you to do something, just immediately say "Yep, done". Then when they actually want to see the work start madly scrabbling...
02:25	_mst	people will respect you for it :)
02:26	Fossi1	hell, i can think lazy ;D
02:27	Fossi1	normally it's called procrastination though
02:30	albino	_mst: I like that advice
02:30	Fossi1	lisppaste8: url
02:30	lisppaste8	To use the lisppaste bot, visit http://paste.lisp.org/new/clojure and enter your paste.
02:31	lisppaste8	Fossi pasted "ref trouble" at http://paste.lisp.org/display/84421
02:32	Fossi1	when i call this, i get [ref{} ()] back, then when i run *1 on the repl again, ref suddently has a value
02:34	hiredman	are you sure you want commute there?
02:34	Fossi1	could be ref-set, i don't care what color it is, if i have 2 such events
02:35	Fossi1	just why does the ref change after the dosync is done?
02:35	Fossi1	because everything around it is lazy?
02:35	Fossi1	do i need to doall the map?
02:35	hiredman	yeah
02:36	hiredman	you might just put (vec left-events) in the returned pair
02:38	Fossi1	ok, i'll try
02:45	Fossi1	ok, didn't help much with my error, i'll have a look later
02:45	Fossi1	off to more professional clojure ventures ;)
02:45	Fossi1	bbl
03:32	lbj	Top of the morning gents
03:44	AWizzArd	Hi lbj
03:49	lbj	Any interesting new developers in Clojureland lately? (disregarding new new entirely)
03:50	lbj	hehe, developers = developments
03:52	Chousuke	well, the new forkjoin stuff looks promising, and "batch mode"
03:52	lbj	What are they?
03:54	lbj	@ Chousuke
03:54	Chousuke	http://paste.lisp.org/display/84027 <- automatically parallelised vector operations and http://paste.lisp.org/display/84027 (no timing data there though) :/
03:55	Chousuke	I wonder which branch the batch stuff was in.
03:56	lbj	Interesting - Does vec then determine wether or not the task at hand is suited for paralization ?
03:56	Chousuke	vec does nothing different. it's pvreduce and pvmap that do the parallelisation
03:58	Chousuke	rhickey says he has a quad core machine, so the results are rather impressive.
03:58	lbj	Indeed
03:58	Chousuke	I have a dual core and wasn't able to get 2x improvement :/ I suspect my memory is a bottleneck
03:58	lbj	But if you replace vec with pvreduce, what would the answer to my question be?
03:59	Chousuke	pvreduce always divides the work I think
03:59	lbj	There's a certain threshold before which paralization would only slow the process down with scheduling. If this is automated it has to be aware of that somehow
04:00	Chousuke	the batch mode stuff is also interesting; it allows you to get a local "mutable" version of a collection for cases when you know you will be doing lots of modifications in one go and don't need the intermediates.
04:00	Chousuke	lbj: it's not that far automated :)
04:01	Chousuke	regular reduce and map will remain (the pv* stuff is not lazy anyway)
04:05	Chousuke	lbj: There's also the chunked seqs branch, but you may have already heard of that :)
04:06	lbj	That I'm familiar with
04:18	Chousuke	lbj: I hope this doesn't explode your ERC
04:18	lbj	Thanks friend - I actually got one of those cool Jaunty notifications :)
04:19	lbj	Now I just need to work out the hook for private msgs, seems thats handled differently
04:26	blbrown_win	morning
04:26	lbj	Morning :)
04:37	blbrown_win	I wish I got paid for the work I did outside of work
04:38	hiredman	heh
04:39	Raynes	I wish I got paid for the work that hiredman does outside of work...
04:39	blbrown_win	nice
04:39	blbrown_win	me too
06:00	angerman	I have two vecotrs, the first one consists of bools and has exactly one set to true. The second one has values.
06:00	angerman	now I want to get the value that corresponds to the true item
06:01	angerman	any ideas?
06:01	angerman	,(let x [true false false] y ["a" "b" "c"] (for [x x y y :when x] [x]))
06:01	clojurebot	java.lang.IllegalArgumentException: let requires a vector for its binding
06:07	jdz	,(mapcat (fn [x y] (when x (list y))) '[true false false] '["a" "b" "c"])
06:07	clojurebot	("a")
06:07	angerman	,(:true (zipmap [:true :false :false] ["a" "b" "c"])
06:07	clojurebot	EOF while reading
06:07	angerman	,(:true (zipmap [:true :false :false] ["a" "b" "c"]))
06:07	clojurebot	"a"
06:07	angerman	does not feel right :/
06:08	lbj	angerman: Why are you using a vector?
06:08	angerman	lbj: ?
06:09	hiredman	,((comp second first (partial filter first) (partial map vector)) '[true false false] '["a""b" "c"])
06:09	clojurebot	"a"
06:10	lbj	angerman: It sounded like a job for a map, but anyway, merge the two, maybe applying partition 2
06:10	hiredman	,(zipmap '[true false false] '["a" "b" "c"])
06:10	clojurebot	{false "c", true "a"}
06:10	hiredman	,((zipmap '[true false false] '["a" "b" "c"]) true)
06:10	clojurebot	"a"
06:10	hiredman	I think I win
06:10	angerman	hiredman: ahh, right
06:10	angerman	hiredman: :) I tried (true (zipmap ...)) and that didn't work
08:45	lbj	...
10:51	dysinger	moing!
10:53	cgrand	dysinger: shhh everybody is asleep
11:26	Drakeson	how would you find the element in a list that has the maximum value of a function of the element? (reduce max (map f coll)) only gives you the maximum value, not the one that maximizes f.
11:28	stuartsierra	I think there's a contrib fn that does that
11:28	rhickey	,(doc max-key)
11:28	clojurebot	"([k x] [k x y] [k x y & more]); Returns the x for which (k x), a number, is greatest."
11:29	Drakeson	thanks :)
13:09	rhickey	new new branch is up
13:10	stuartsierra	cool
13:10	rhickey	still work in progress, but can create new instances - all method need full type hints (missing hint means Object)
13:11	rhickey	will eventually try to match superclasses' sigs
13:11	rhickey	does accept non-interface superclass with no-arg ctor
13:12	rhickey	primitives can flow in and out
13:12	rhickey	no indirection on method calls
13:13	rhickey	(new [super? interfaces] thisname? {:flags :here}? (methodname [args] body))
13:14	stuartsierra	"thisname" is a class name?
13:14	rhickey	hints are metadata on method and arg names
13:14	rhickey	stuartsierra: no, the name of this object, usually this
13:14	stuartsierra	ah, ok
13:16	rhickey	(defn foo []
13:16	rhickey	(let [y (int 17)]
13:16	rhickey	(new [Object clojure.lang.Seqable] this {:foo :bar}
13:16	rhickey	(#^String toString [] "foo")
13:16	rhickey	(#^int hashCode [] 42)
13:16	rhickey	(#^int baz [#^int x] (+ x y))
13:16	rhickey	(#^clojure.lang.ISeq seq [] (seq ["woo hoo"])))))
13:17	Chousuke	are the flags purely for the compiler or can you access them somehow?
13:18	rhickey	Chousuke: flags go in bitbucket right now
13:18	Chousuke	heh.
13:20	stuartsierra	So this returns an instance of a new dynamically-generated class, which is not a proxy, correct?
13:20	rhickey	right
13:20	rhickey	should be much faster than proxy
13:20	stuartsierra	Gotcha.
13:20	stuartsierra	And it is not intended to replace gen-class for statically-named classes?
13:20	rhickey	not right now
13:20	stuartsierra	ok
13:20	duck1123	will this replace proxy, then?
13:21	rhickey	duck1123: most usage will move to this, but not removing proxy
13:21	rhickey	proxy does support some dynamic fiddling which this won't
13:22	rhickey	not much used afaik
13:22	stuartsierra	Right, this "new" creates a class which cannot be changed, correct?
13:22	rhickey	right
13:22	stuartsierra	I didn't realize proxies could be changed.
13:22	rhickey	but the enclosing 'factory' fn can, since no one depends on the name of the generated class
13:23	rhickey	,(doc update-proxy)
13:23	clojurebot	"([proxy mappings]); Takes a proxy instance and a map of strings (which must correspond to methods of the proxy superclass/superinterfaces) to fns (which must take arguments matching the corresponding method, plus an additional (explicit) first arg corresponding to this, and updates (via assoc) the proxy's fn map. nil can be passed instead of a fn, in which case the corresponding method will revert to the default behavior
13:25	stuartsierra	Wow, never saw that before. Can't think I'd use it much; I'd rather call a normal fn and rebind that.
13:25	Chouser	rhickey: so cool. Are the return type hints required for some reason?
13:25	rhickey	still todo - figure out signatures from superclasses, prevent additional overloads of superclass methods (overrides only, local helper methods must have unique names), volatile declarations, non-reflective calls to this...
13:26	Chouser	ah
13:26	rhickey	stuartsierra: the advantage of that is it updates existing instance
13:26	stuartsierra	Cool, like prototype classes.
13:27	rhickey	stuartsierra: well, it's per instance
13:27	stuartsierra	Ah, like Ruby's instance methods, then. :)
13:27	Chouser	update-proxy is per instance, delegating do a var and redef or binding that would be per class, and newnew won't do either.
13:27	rhickey	like I said, I don't know if anyone uses it
13:28	Chouser	to a var
13:28	rhickey	new new will be what most people imagine proxy is
13:28	lbj	Good evening gents
13:29	rhickey	anyway, what's up there should be usable, and needs a good workout
13:29	stuartsierra	Same usage pattern as an anonymous inner class in Java, right?
13:30	rhickey	stuartsierra: sort of, more emphasis on the lexical binding, less on nested class scope
13:30	stuartsierra	ok
13:30	rhickey	but supports multiple supers, unlike aic
13:30	stuartsierra	right, forgot that detail
13:31	rhickey	no local member not from lexical scope, i.e. no explicit fields, field inits, class init...
13:32	rhickey	will support volatile declaration making for mutable local in 'this' only
13:32	rhickey	I would like a good name for these things other than "kind of like an anonymous inner class" though
13:32	rhickey	:)
13:33	rhickey	lexical objects?
13:33	duck1123	so does "old" new still work the same way, or will usage of it need to be changed?
13:34	duck1123	not that I ever use new, I always do (Object.)
13:34	rhickey	,(macroexpand-1 '(Object.))
13:34	clojurebot	(new Object)
13:35	lbj	The questions still good :)
13:35	rhickey	(new classname ...) works as ever, does something different only when it sees vector as first arg
13:36	Chouser	lexical object isn't bad. I was thinking Clojure object would work.
13:36	rhickey	Chouser: I want to avoid Band-Aid/Kleenex syndrome
13:36	lbj	Ok, sounds good
13:37	Chouser	other languages have lexical objects, but it doesn't seem like there's a whole lot of agreement on what they actually are.
13:37	duck1123	I knew that (Object.) expanded to (new Object) but presumably if the way to call (new Object) changed, the macro expansion would be changed to the new format
13:37	stuartsierra	"temporary-class object"?
13:37	rhickey	Chouser: right, completely different meanings
13:37	stuartsierra	"generated object"?
13:37	y-combinator	Hello. What is the best way to comnvert lazy seq to vector?
13:37	Chouser	y-combinator: vec
13:38	stuartsierra	"anonymous instance"?
13:38	Chouser	you actually get both a class and an object
13:39	rhickey	please try replacing a proxy or two with new new, admittedly the required hints make it a bit of a pain right now
13:39	stuartsierra	"generated class instance"?
13:39	rhickey	Chouser: right, that's the problem
13:39	Chouser	ooh, sure, I have proxy all over the place. hmm, where to start...
13:40	Chouser	hm. need ctor args
13:40	lbj	~proxy
13:40	clojurebot	proxy is <Chouser> proxy teases with its ease of use, then suddenly betrays.
13:41	lbj	To me proxy was downright unuseable, because it could extend far enough
13:47	Chouser	'read' requires a PushbackReader; PushbackReader requires an arg for its ctor; have to keep using proxy here I guess.
13:55	Chouser	Second attempt, this time extending sun.misc.Signal:
13:55	Chouser	Exception in thread "SIGINT handler" java.lang.AbstractMethodError: clojure.contrib.repl_utils$eval__335$start_handling_break$obj__339.handle(Lsun/misc/Signal;)V
13:57	Chouser	(new [sun.misc.SignalHandler] this (#^Void handle [sig] (prn :got-it) nil))
13:59	Chousuke	Void? not void?
14:00	Chouser	,#^Void []
14:00	clojurebot	[]
14:00	Chouser	,#^void []
14:00	clojurebot	java.lang.Exception: Unable to resolve symbol: void in this context
14:00	Chousuke	hmm
14:00	Chousuke	,(str Void/TYPE)
14:00	clojurebot	"void"
14:01	stuartsierra	(.getName (.getParent Void))
14:01	stuartsierra	,(.getName (.getParent Void))
14:01	clojurebot	java.lang.IllegalArgumentException: No matching field found: getParent for class java.lang.Class
14:06	rhickey	#^void in new new
14:08	Chouser	hm. same error though.
14:08	Chouser	(.handle (new [sun.misc.SignalHandler] this (#^void handle [sig] nil)) nil)
14:08	stuartsierra	do you need to type-hint sig?
14:09	rhickey	Chouser: sig is an Object?
14:09	Chouser	yes
14:09	Chouser	(.handle (new [sun.misc.SignalHandler] this (#^void handle [sig] nil)) (sun.misc.Signal. "INT"))
14:09	rhickey	there's no default method for methods yo udon't define
14:10	rhickey	if they get called you'll get AbstractMethodError
14:10	Chouser	(.handle (proxy [sun.misc.SignalHandler] [] (handle [sig] nil)) (sun.misc.Signal. "INT")) ;works fine
14:10	Chouser	rhickey: ok, good to know. sun.misc.SignalHandler has only the one instance method
14:12	rhickey	Chouser: sig is Signal, no?
14:12	rhickey	#^Signal sig
14:12	Chouser	yes, sun.misc.Signal
14:12	rhickey	then it's not Object
14:12	Chouser	oh, sorry. I thought you meant non-primitive
14:13	rhickey	sorry I said an Object before, the sig is not Object
14:13	Chouser	ah, so I do have to hint the arg
14:13	rhickey	right now you have to specify exact signatures, I'm not looking at superclasses at all
14:13	Chouser	ok
14:13	rhickey	it won't end up being this hard, sorry
14:14	Chouser	ok! works perfectly.
14:15	Chouser	so proxy will continue to be useful for when you need to extend classes that require args in their ctors?
14:16	rhickey	Chouser: I need to see when/why people are doing that. Deriving from concrete classes is always a questionable design, java.io a perfect example of how bad it can get
14:17	Chouser	ok
14:17	Chouser	in this case (repl-utils/get-source) I only need it because core/read needs a PushbackReader
14:18	Chousuke	hmm
14:18	Chousuke	are there any OO languages that allow only interface inheritance?
14:18	rhickey	But looking at the difference between new new and anonymous inner classes is instructive, all that implicit complexity we become inured to
14:19	rhickey	Chouser: It's a real pain that there aren't interfaces for java.io
14:19	rhickey	but when does it stop?
14:20	Chousuke	is java.nio better in this regard? :/
14:20	rhickey	even the no-arg concrete ctors are insidious
14:22	Chouser	Does java.nio have anything line-based?
14:24	Chouser	oh, I guess that's not what 'read' cares about. It wants an unread() method.
14:25	stuartsierra	What's the alternative to no-arg ctors? Factories?
14:26	rhickey	stuartsierra: it's actually more about concrete derivation than the ctors
14:26	stuartsierra	right, ok
14:27	Chouser	I guess clojure could define an interface with unread(), and by default use a custom class that extends PushbackReader. That would allow me to use newnew to implement a different class that 'read' would be happy with.
14:27	Chousuke	IClojureReader? :P
14:27	rhickey	Chouser: it could, yes
14:27	rhickey	IMTiredOfJavaIo
14:28	Chousuke	:)
14:28	Chouser	with all that, repl-utils could abandon proxy
14:28	Chousuke	Probably it would help with clojure-in-clojure as well?
14:28	stuartsierra	I remember when I was trying to write c.c.fnmap, I discovered I couldn't derive from PersistantHashMap. Maybe I wasn't supposed to.
14:29	rhickey	stuartsierra: no, there are interfaces + abstract bases
14:29	hiredman	clojurebot: a man of means by no means is <reply>da da king of the road
14:29	clojurebot	Alles klar
14:30	Chousuke	if you define the clojure reader in terms of Clojure interfaces, it would be more portable across hosts too, I guess.
14:30	rhickey	I'd love for a different solution than abstract bases, but they don't have (most of) the problems of concrete derivation
14:31	rhickey	and I can make it so all of mine have no-arg ctors. The other set of very valuable ones, for j.u.collections, also have no-arg ctors
14:34	Chouser	oh, so 'this' is not required, but if not given you have no way to refer to the current instance.
14:34	clojurebot	this is not a bug
14:35	Chouser	no indeed
14:35	rhickey	right, it can get as small as: (str (new [] (#^String toString [] "foo")))
14:35	rhickey	and once I get the superclass stuff in, lose the type hint
14:35	hiredman	ooo
14:36	stuartsierra	that's nice, that will be really useful.
14:36	stuartsierra	You could really do prototype-style objects then, right? Just define a "constructor" function that calls newnew.
14:37	stuartsierra	Well, I guess you can't change the definition of existing objects, but everything else.
14:37	hiredman	stuartsierra: I doubt that
14:37	Chouser	you can do even closer to that now with proxy, I'd think
14:37	hiredman	seems like new has to be a compile time operation
14:38	rhickey	stuartsierra: it's more a rendition of known good practice - factory fns hiding implementation class details, programming to interfaces only
14:38	hiredman	which makes it hard to monkey with at runtime
14:39	stuartsierra	But you could gen-class a superclass if you need static names or mutable definitions, right?
14:39	stuartsierra	e.g., (gen-class "x.y.z") ... then later (new [x.y.z] ...)
14:40	hiredman	gen-interface :P
14:40	Chouser	exactly
14:40	stuartsierra	cool!
14:40	Chouser	proxy already gives you a whole lot of this
14:41	stuartsierra	True.
14:41	rhickey	stuartsierra: right gen-class will do all of that. I'm still thinking about both stable anonymous names (for serialization) and AOT-only static names, but the best code will gen-interface + new new, possibly + gen-static which will make Java-friendly static-member-only classes
14:41	Chouser	Mmmm.. I want gen-static.
14:41	stuartsierra	Nice.
14:41	kotarak	what I don't like about proxy, is that it's saves its .class file in a common place for different namespaces. This makes packaging a nightmare. :(
14:42	Chouser	I have a gen-class with a whole lot of #^{:static true}
14:43	rhickey	people make such a righteous mess of OO, but if you only ever have immutable instances of anonymous classes manipulated via interfaces, life is good
14:43	rhickey	so I'm trying to encourage that
14:44	rhickey	and most of Clojure's Java is that way, except the few places where I need to define constructs to manage mutation or laziness
14:44	stuartsierra	Great idea. A pity it breaks down as soon as you need to do I/O. :)
14:44	hiredman	java.lang.NoSuchMethodError: clojure.lang.RestFn.<init>(I)V (fnparse.clj:1)
14:44	Chouser	kotarak: interesting. would you want a classpath entry per namespace, and somehow tell proxy that?
14:44	hiredman	:(
14:44	kotarak	chouser: just a sec
14:46	rhickey	stuartsierra: I think Chouser has a good idea replacing PushbackREader in Clojure's interfaces. HAving a concrete class in a signature is always a bad sign. But the drawback is that PushbackReader itself won't implement our interface, people will always have to implement our thingy, which is also bad, but not as bad
14:46	rhickey	and supports implementing the interface via composition, which is good
14:47	Chouser	I suppose 'read' could still accept PushbackReader as well
14:47	stuartsierra	I suppose it's trivial to wrap java.io.PushbackReader (or whatever) in clojure.lang.BetterReader.
14:47	rhickey	as soon as there is an interface you can compose
14:48	rhickey	and with new new implement macros like def-this-on-that
14:48	stuartsierra	Right, I was just thinking about some kind of "thin proxy" that would delegate from an interface to a concrete class with matching methods.
14:48	rhickey	and automate all the forwarding associated with composition
14:49	Chousuke	what methods does the clojure reader need? just read and unread?
14:51	Chouser	Chousuke: I think that's right.
14:52	Drakeson	liebke: a suggestion re incanter, could you please remove dependencies from lib/ and put them in cloud.github, as compojure does? thanks :)
14:56	hiredman	oooh
14:57	hiredman	new-new works in clojurebot's sandbox
14:57	rhickey	,(str (new [] (#^String toString [] "foo")))
14:57	clojurebot	"foo"
14:57	rhickey	cool
14:58	rhickey	let the AbstractMethodErrors ensue
14:59	hiredman	,((new [clojure.lang.IFn] (#^Integer invoke [#^Integer x #^Integer y] (+ x y))) 1 2)
14:59	clojurebot	java.lang.VerifyError: (class: sandbox$eval$obj__1939, method: invoke signature: (Ljava/lang/Integer;Ljava/lang/Integer;)Ljava/lang/Integer;) Wrong return type in function
14:59	hiredman	:\|
14:59	rhickey	use #^int etc
14:59	hiredman	oh
14:59	hiredman	,((new [clojure.lang.IFn] (#^int invoke [#^int x #^int y] (+ x y))) 1 2)
14:59	clojurebot	java.lang.AbstractMethodError: sandbox$eval$obj__1945.invoke(Ljava/lang/Object;Ljava/lang/Object;)Ljava/lang/Object;
15:00	hiredman	ah
15:00	hiredman	doesn't match
15:00	hiredman	I see
15:00	rhickey	,((new [clojure.lang.AFn] (invoke [x y] (+ x y))) 1 2)
15:00	clojurebot	3
15:01	Chousuke	hmm
15:01	hiredman	,(import '(clojure.lang IDref AFn IFn))
15:01	clojurebot	java.lang.ClassNotFoundException: clojure.lang.IDref
15:01	hiredman	,(import '(clojure.lang IDeref AFn IFn))
15:01	clojurebot	clojure.lang.IFn
15:01	rhickey	I came up with a plan for supporting long and double args and returns in IFn earlier this week, but there are only so many hours in a week
15:01	rhickey	IDeref
15:01	hiredman	Excellent
15:02	hiredman	~rhickey
15:02	clojurebot	is_rhickey_is_a_minor_god? is yes
15:02	Chousuke	funky predicate
15:02	Chousuke	:P
15:03	gjohnson	howdy rich. I'm wondering about pcalls in clojure.core
15:03	rhickey	gjohnson: hi
15:04	gjohnson	At first glance, I would have imagined it was part of jsr166y.
15:04	gjohnson	but apparently not. so I'm guessing you're doing some magic with threadpools or somesuch?
15:04	hiredman	rhickey: so new-new methods aren't backed my fns?
15:04	rhickey	gjohnson: it's just piggybacking on the agent thread pool
15:05	gjohnson	ok, that was my guess.
15:05	rhickey	hiredman: no, pedal to the metal methods
15:05	rhickey	gjohnson: there is work going on in the par branch to leverage jsr166y, including a fjtask macro
15:05	hiredman	do they have a frame to recur to?
15:05	rhickey	returns something that can be forked and joined
15:06	rhickey	hiredman: yes, should do
15:06	gjohnson	hmm...so will that ultimately give pcalls a dependency on jsr166y and move it out of core?
15:06	hiredman	I guess I could have just tried that out :P
15:07	rhickey	gjohnson: I haven't decided on tying the two together, so probably some new features for fj
15:07	gjohnson	I had built a little tool for splitting my work over multiple processors back in January or so, but some of it was guesswork because I don't know the formula you're using to calculate the size of the fixed thread pool.
15:07	stuartsierra	number of cores reported by the runtime
15:07	gjohnson	,(defn distribute-load-over-processors
15:07	gjohnson	[action-fn arg-seq]
15:07	gjohnson	(let [num-processors (.availableProcessors (Runtime/getRuntime))
15:07	gjohnson	agents (map agent (replicate (* agents-per-processor num-processors) ()))]
15:07	gjohnson	(println "Sending Tasks to" (count agents) "Agents...")
15:07	clojurebot	EOF while reading
15:07	gjohnson	(dorun (map #(send %1 action-fn %2) (cycle agents) arg-seq))
15:07	gjohnson	(println "Waiting for Agents to Finish...")
15:07	gjohnson	(apply await agents)
15:07	gjohnson	(apply concat (map deref agents))))
15:07	gjohnson	bah
15:08	hiredman	:(
15:08	gjohnson	looks like the code didn't make it through unscathed. sigh
15:08	Chousuke	lisppaste8: url
15:08	lisppaste8	To use the lisppaste bot, visit http://paste.lisp.org/new/clojure and enter your paste.
15:08	stuartsierra	try a pastie
15:09	gjohnson	alrighty. the , prefix only works for one-liners then?
15:09	stuartsierra	yep
15:09	Chousuke	and you can't defn anyway :p
15:10	lisppaste8	gjohnson pasted "distribute-load-over-processors" at http://paste.lisp.org/display/84459
15:10	gjohnson	indeed, that would be a bit of a security hole.
15:10	gjohnson	;)
15:11	Chousuke	well, you can get around the protections anyway
15:11	Chousuke	but as far as I can tell, all you can do is DOS clojurebot :P
15:11	gjohnson	hmm...
15:11	stuartsierra	,(loop [x 1] (recur (inc x))
15:11	clojurebot	EOF while reading
15:11	Chousuke	to take over the machine itself, you'd have to break java's sandboxing.
15:12	stuartsierra	ok, I'll play nice, won't retype that
15:12	rhickey	,(.down (new [] (down [x] (if (zero? x) x (recur (dec x))))) 42)
15:12	clojurebot	0
15:12	gjohnson	well, now that that's pasted, I was wondering if someone (rich?) could take a look at the code and tell me how different (for better or worse) this function is as compared to pcalls.
15:12	hiredman	rhickey: that is awesome
15:12	Chousuke	,(loop [x 1] (recur (inc x)))
15:12	Chousuke	:p
15:12	hiredman	♥ new new
15:12	clojurebot	Execution Timed Out
15:13	stuartsierra	Clever clojurebot.
15:13	hiredman	~def pcalls
15:13	gjohnson	apparently so.
15:13	gjohnson	ah, pretty.
15:13	hiredman	uh oh
15:14	gjohnson	that url's not working for me.
15:14	Chousuke	gjohnson: you might want to use futures instead of agents; agents have a thread pool anyway.
15:15	kotarak	Chouser: sorry. Had to operate the musicbear. I'm not in the details how proxy works. But if it generates a class clojure/proxy/my/Super.class, my naive view would say that it should be possible to generate a class named foo/bar/clojure/proxy/my/Super.class, no?
15:15	gjohnson	as noted. that's why I don't spawn any threads.
15:16	gjohnson	I just create a number of agents that are a dynamically bound multiple of the thread pool, and map calls to send across a cycled sequence of the agents.
15:17	hiredman	ah
15:17	gjohnson	this should just queue up all the functions more or less evenly across my agents and let the fixedthreadpool handle them.
15:17	hiredman	because that rev is on the new-new branch
15:17	hiredman	or not
15:17	Chouser	kotarak: ah! hm, I don't see why it couldn't.
15:18	kotarak	Then packaging would be easy. foo/bar/* goes to that .jar, foo/baz/* to that other one.
15:18	Chouser	rhickey: ,(.no-frickin-way (new [] (#^String no-frickin-way [] "no, really!")))
15:18	clojurebot	new Class(x) is (Class. x)
15:18	kotarak	Without fiddling with the proxy classes. (Which namespace calls which proxy? Hmm... Bookkeeping ahead...)
15:18	Chouser	,(.no-frickin-way (new [] (#^String no-frickin-way [] "no, really!")))
15:18	clojurebot	"no, really!"
15:18	gjohnson	my concern is that if I make less agents that the thread pool size, I'll end up with unused threads.
15:19	Chouser	rhickey: will the ability to create new methods like that go away?
15:20	hiredman	grr
15:24	hiredman	kotarak: it's lisp! it has eval! and macros! YEAH!
15:26	rhickey	Chouser: no, that will stay, to allow for the definitions of helper methods. Calling them from outside will always involve reflection though
15:27	rhickey	they will have to have names different from those of the supers' methods, and overloading on arity only
15:27	Chouser	oh, interesting.
15:28	Chouser	runtime reflection required to call other instances of the same class?
15:28	rhickey	Chouser: yeah, anything other than this
15:30	rhickey	the dashes won't work on all JVMs
15:30	rhickey	so I'll have to filter those
15:31	rhickey	,(class (new []))
15:31	clojurebot	sandbox$eval$obj__1918
15:37	Chouser	,(new [] (#^String toString [] (str "what's wrong?")))
15:37	clojurebot	java.lang.VerifyError: (class: sandbox$eval$obj__1923, method: toString signature: ()Ljava/lang/String;) Wrong return type in function
15:38	hiredman	,(class (str "what's wrong?"))
15:38	clojurebot	java.lang.String
15:38	hiredman	:/
15:38	hiredman	,(new [Object] (#^String toString [] (str "what's wrong?")))
15:38	clojurebot	java.lang.VerifyError: (class: sandbox$eval$obj__1932, method: toString signature: ()Ljava/lang/String;) Wrong return type in function
15:39	Chouser	the compiler should even know str returns a String
15:39	hiredman	,(new [] (#^String toString [] (.toString (str "what's wrong?" "foo"))))
15:39	clojurebot	#<sandbox$eval$obj__1952 what's wrong?foo>
15:41	hiredman	,(str (new [] (#^String toString [] (str "what's wrong?")))
15:41	clojurebot	EOF while reading
15:41	hiredman	,(str (new [] (#^String toString [] (str "what's wrong?"))))
15:41	clojurebot	java.lang.VerifyError: (class: sandbox$eval$obj__1981, method: toString signature: ()Ljava/lang/String;) Wrong return type in function
15:41	trotter	are there any good code coverage tools for clojure?
15:42	hiredman	,(.foo (new [] (#^String foo [& bar] "foo")) 1 2)
15:42	clojurebot	"foo"
15:43	hiredman	,(new [] (#^String toString [] #^String (str "what's wrong?")))
15:43	clojurebot	java.lang.VerifyError: (class: sandbox$eval$obj__2016, method: toString signature: ()Ljava/lang/String;) Wrong return type in function
15:43	hiredman	:(
15:43	hiredman	it's too bad
15:45	hiredman	lousey Object specifying interface
15:47	rhickey	Chouser: fixed: http://github.com/richhickey/clojure/commit/3c15d0eb6bdc279c801dd984a3524666d7c5cfbe
15:53	Chouser	rhickey: Thanks! That's every proxy in contrib, except the read thing: http://n01se.net/paste/knS
15:55	rhickey	wow
15:56	rhickey	does it still work?
15:59	Chouser	yes
15:59	Chouser	er, every proxy of mine
16:00	rhickey	will be much nicer without the hints
16:00	Chouser	there are still proxies in swing_utils and singleton
16:00	Chouser	yes, it definitely will.
16:21	Fossi	so, to come back to my question of tomorrow, if i deref a ref, it's contents doesn't get visited, right?
16:21	Fossi	so i get whatever state the ref might be in?
16:22	Fossi	s/tomorrow/this morning
16:22	hiredman	eh?
16:22	Fossi	http://paste.lisp.org/display/84421
16:22	hiredman	"it's contents doesn't get visited" what does that mean?
16:23	Fossi	"hiredman: you might just put (vec left-events) in the returned pair"
16:23	hiredman	Fossi: I remember saying that
16:23	Chousuke	Fossi: camelCase is not very clojurey :p
16:23	Chousuke	Fossi: you should use hyphens with lisp functions.
16:24	Fossi	so if on the left, i return @state-ref, i get {} mostly
16:24	Fossi	Chousuke: ups, all the Java in that app tends to confuse me :)
16:24	Chousuke	also, I'd recommend not putting any actual dosync logic in your functions.
16:24	Fossi	that i don't understand
16:25	hiredman	Fossi: I still want to know what "it's contents doesn't get visited" means
16:25	Fossi	well, i deref the ref, but i get {} (for example on the repl)
16:25	Chousuke	Fossi: instead of going a dosync in change-color, write all the transitions as pure functions and then call (dosync (alter state change-color)) in some function that contains most of the "non-functional" logic
16:25	hiredman	and?
16:25	Fossi	if i use *1, i suddently get a value
16:26	hiredman	Fossi: print out the second item in the return vector of calcGameState forces the map
16:26	hiredman	map being lazy
16:26	Fossi	yeah, somehow feels weird
16:26	Fossi	same for try catch really
16:26	hiredman	the functio being mapped as a side effect of changing the ref
16:27	hiredman	so the ref does not get changed unless you force the map
16:27	Fossi	but i guess that's Chousuke's point
16:27	Chousuke	Fossi: What I'm advocating is isolation of side-effects. ref alterations are side-effects :)
16:27	Fossi	Chousuke: well, i could have another layer there, yes
16:28	Fossi	but then, i don't really know what implementors of dispatchEvent might want to do to the state
16:29	hiredman	anyway
16:29	hiredman	you should force the map
16:30	Chousuke	why are you returning a new state-ref though?
16:30	Chousuke	in calcGameState
16:31	Fossi	i return @state-ref now actually
16:31	Fossi	state-ref was just to test it from the repl
16:32	Fossi	the idea is to have a load of stm calculations being done to get to the new game state for the "mutable" game objects
16:33	Chousuke	do you really need a ref though? hmm.
16:33	Chousuke	I mean, perhaps you could reduce the state over the events with dispatch-event?
16:34	hiredman	yeah
16:34	alrex021	I know this is not really clojure q, but hope someone could help... I need to include Incanter lib jars into my emacs env. How should the (setq swank-clojure-extra-classpaths ..) look? Sorry but I'm new to clojure and lisp too :(
16:35	Chousuke	alrex021: (setq swank... '("file:///some/path/here/javalib.jar" "file://more/paths/")) ought to work
16:35	Fossi	later on the plan is to do it in parallel
16:36	hiredman	,(doc pvreduce)
16:36	clojurebot	"/;nil; "
16:36	hiredman	clojurebot: curse you!
16:36	clojurebot	Excuse me?
16:36	Chousuke	but if future events depend on the earlier state, it's not very parallelisable :/
16:37	alrex021	Chousuke: thank you. Do I need to specify each required jar of the lib then of course? Or is it possible to say...all *.jar(s) in following dir?
16:37	liebke_	alrex021: technomancy created an incanter.el file, that I just added to the Incanter repository today, that might be helpful. It's in the bin/ directory
16:37	clojurebot	technomancy is to blame for all failures
16:37	Chousuke	Fossi: oh, and the (do ...) in changeColor is not needed. function bodies have an implicit do.
16:37	Fossi	ah, didn't know pvreduce yet
16:38	Chousuke	Fossi: pvreduce is very new; not in master yet.
16:38	Fossi	Chousuke: i know, i still like it for reminding me of side effects
16:38	Fossi	i'll wrap each game object in it's own stm then
16:38	alrex021	liebke_: ahh, gr8 the author, what a gr8 community :) ...thanks
16:39	Chousuke	Fossi: hmm, yeah :)
16:39	Fossi	and i don't care much about order of things, or whether they fail at first, with games you do this calculation 60/s, so it doesn't have to be overly correct
16:39	liebke_	alrex021: good luck, I wish I could be more help, but I'm not an emacs user :)
16:40	Fossi	alrex021: (directory-files "~/projects/android/bannerwars/libs/" t ".*\.jar")
16:41	alrex021	liebke_: out of curiosity, what do you use for your clojure dev?
16:41	liebke_	I just use vim and a repl
16:42	liebke_	emacs is the way to go though... if you can get it configured :)
16:42	lbj	liebke_: I'll recommend Emacs + SLIME to you, its a fairly effecient combo, and easily installable thanks to technomancy's clojure-mode
16:42	Fossi	it's infecting
16:42	gjohnson	k
16:43	Fossi	i begin to select text with ctrl space in firebird
16:43	liebke_	lbj: I know, but I've used vi/vim for too long to switch at this point
16:43	Fossi	and try to swap desktops with shift-left
16:43	Fossi	liebke_: viper-mode? :)
16:44	gjohnson	how about vimclojure/gorilla
16:46	hiredman	http://technotales.wordpress.com/2007/10/03/like-slime-for-vim/
16:47	kotarak	ieck do yourself a favour and use vimclojure. (shameless advertisement)
16:47	kotarak	the screen has problems.
16:48	kotarak	It works but is tricky.
16:48	hiredman	shrug
16:50	hiredman	if you ever open a file that contains function calls, not just function definitions, vimclojure will execute those function calls
16:50	hiredman	I find that to be unacceptable
16:51	hiredman	heck
16:51	kotarak	IMHO, you shouldn't do that. Providing an entry point brings flexibility and advantages when debugging.
16:51	hiredman	if you try and edit a .clj file that is not valid (not all the parens are closed, etc) vimclojure will puke
16:51	hiredman	kotarak: I have an entry point, and I have a .clj file that contains a call to that entry point
16:52	hiredman	so if I open up my scratch clojure file, vimclojure pukes
16:52	Chousuke	also, if the file contains (shell-out "rm -rf /") or something :P
16:52	kotarak	You can disable the dynamic part at any time and just use static vim with <C-n> and friends. Works also very well.
16:52	hiredman	yeah, I have vim clojure installed and use the static part
16:53	hiredman	I never use gorilla
16:53	hiredman	I use slime.vim works great, and it works for things besides clojure
16:58	Fossi	yay. it works again. now that i have that set up, i can start doing more business logic :)
17:11	gjohnson	okay, so I know this is a pretty remedial java question...
17:11	gjohnson	but how can I ask a threadpool for its size (number of threads)?
17:11	hiredman	~jdoc java.util.concurrent.ThreadPool
17:11	gjohnson	I see that clojure.lang.Agent uses 2 + the number of cores detected by the runtime for the fixed threadpool size
17:11	hiredman	damn
17:12	gjohnson	?
17:12	hiredman	~jdoc java.util.concurrent.ThreadPoolExecutor
17:12	hiredman	still no dice
17:12	hiredman	http://java.sun.com/javase/6/docs/api/java/util/concurrent/ThreadPoolExecutor.html
17:12	hiredman	there is a getPoolSize method
17:13	Chouser	show works great for this
17:13	gjohnson	ah, thanks.
17:13	Chouser	(show clojure.lang.Agent) --> discover pooledExecutor and soloExecutor
17:14	Chouser	(show clojure.lang.Agent/pooledExecutor) --> discover getPoolSize and getMaximumPoolSize
17:14	Chouser	,(.getMaximumPoolSize clojure.lang.Agent/pooledExecutor)
17:14	clojurebot	3
17:39	Drakeson	how can I make a memoized recursive function?
17:41	hiredman	,(macroexpand '#'foo)
17:41	clojurebot	(var foo)
17:46	Drakeson	hiredman: that is (defn f1 [n] (if (zero? n) 0 (+ (#'f1 (dec n)) n))) (def f1 (memoize f1)) ?
17:46	hiredman	I am not sure
17:47	Chousuke	that would work, but it'll blow the stack if the recursion is too deep.
17:48	Chouser	that shouldn't be necessary
17:50	Chouser	(defn f1 [n] (prn n) (when (> n 0) (f1 (dec n))))
17:50	Chouser	(def f1 (memoize f1))
17:50	Chouser	(f1 5) ; prints 5 to 0
17:50	Chouser	(f1 10) ; prints 10 to 6
17:51	Drakeson	Chouser: thanks. That is an interesting way to test :)
17:59	alrex021_	I have been trying to pluggin this .el script into my .emacs for the past hour with no luck :( http://github.com/liebke/incanter/raw/fcbcdcc788916c204aca32e6b7810de4a6d0e6d2/bin/incanter.el
18:05	alrex021_	Could someone enlighten me on how one can configure "properly" full clojure support for emacs?
18:05	alrex021_	I am really struggling with this
18:05	hiredman	~emacs
18:05	clojurebot	emacs is best configured for Clojure with instructions at http://technomancy.us/126
18:05	hiredman	shrug
18:15	alrex021_	hhm: Clojure install failed .. git clone git://github.com/kevinoneill/clojure.git when following http://technomancy.us/126 instructuins
18:16	durka42	hmm, those instructions are old, the git repository is under richhickey
18:16	durka42	http://github.com/richhickey/clojure/tree/master
18:17	technomancy	alrex021_: getting the latest version of clojure-mode will fix that problem.
18:18	alrex021_	hhmm so is ELPA option not gonna work?
18:19	alrex021_	I used ELPA to install latest clojure-mode .. then ran clojure-install ...that didn't work
18:19	alrex021_	so I take it I need to manual install
18:20	technomancy	alrex021_: clojure-mode 1.3 is in ELPA
18:20	technomancy	that will do it
18:21	alrex021_	(I am extremely new to Emacs) ... do I have to restart after installing 1.3 via ELPA before I run clojure-install?
18:22	technomancy	you shouldn't have to, but if you're having problems you might try that.
18:22	alrex021_	also, I take it I don't have to add anything to my .emacs file after I install the clojure-mode
18:23	technomancy	no, but it will give you some code to add to your .emacs after M-x clojure-install finishes.
18:23	technomancy	but it should be pretty clear from the instructions
18:26	alrex021_	ok I see, thx. I restarted and ran clojure-install...now its getting it from new git location
18:28	alrex021_	is there a way ti autoreload the .emacs file without restart of emacs?
18:28	alrex021_	off hand :)
18:28	technomancy	sure; open it and do M-x eval-buffer
18:29	alrex021_	thx
18:31	alrex021_	hhmm ok the clojure-install worked and it just gave me 1 liner to add to my .emacs and M-x slime doesn't work (slime not an option)
18:31	alrex021_	line it gave me is just (clojure-slime-config)
18:33	alrex021_	so I added (clojure-slime-config) to my .emacs and restarted and then ran: M-x slime
18:33	alrex021_	but no luck
18:38	rhickey	new new: added signature inference, overload detection: http://github.com/richhickey/clojure/commits/new
18:38	rhickey	so now you don't need to hint except to disambiguate
18:38	hiredman	:O
18:38	hiredman	:D
18:42	rhickey	,(str (new [] (toString [] "hello")))
18:42	clojurebot	"sandbox$eval$obj__2040@945b95"
18:43	rhickey	hrm
18:46	hiredman	,(str (new [] (toString [] "hello")))
18:46	clojurebot	java.security.AccessControlException: access denied (java.lang.RuntimePermission accessDeclaredMembers)
18:46	hiredman	whelp
18:46	hiredman	it was a wild and crazy ride
18:46	rhickey	oh no
18:47	rhickey	did clojurebot not like proxy either?
18:47	hiredman	yep
18:49	mebaran151_	is there a super type for Float and Double?
18:49	mebaran151_	other than Number
18:49	rhickey	is permission java.lang.RuntimePermission "accessDeclaredMembers"; dangerous?
18:50	hiredman	,(str (new [] (toString [] "hello")))
18:50	clojurebot	java.security.AccessControlException: access denied (java.lang.RuntimePermission accessDeclaredMembers)
18:51	hiredman	I dunno
18:51	hiredman	,(str (new [] (toString [] "hello")))
18:51	clojurebot	"hello"
18:51	hiredman	,(ancestors Float)
18:51	clojurebot	#{java.lang.Object java.lang.Number java.io.Serializable java.lang.Comparable}
18:55	rhickey	cool
20:31	iBong	is there a (sugary) way to set flags like case insensitivity when using #regex literals?
20:44	sillycloud	yes
20:44	sillycloud	regexp : /pattern/i
20:45	sillycloud	or var rgx = new RegExp(); rgx.ignoreCase = true
20:47	iBong	thx
20:49	sillycloud	um
20:49	sillycloud	sorry
20:49	sillycloud	wrong channel
20:49	sillycloud	lol
20:49	iBong	er
20:49	sillycloud	my answer is obviously wrong
20:49	iBong	yeah that didnt work
20:49	sillycloud	>.<
20:49	iBong	I was like, cool, just like ruby
20:49	sillycloud	lols
20:50	iBong	the java way is super ugly
20:50	iBong	hoping clojure had a nice alternative, like not having to escape escapes
20:50	Raynes	sillycloud: Think you're in #scala? :p
20:50	sillycloud	worse
20:50	sillycloud	i'm currently in #flash
20:50	iBong	lol
20:50	sillycloud	helping the noobs
20:51	sillycloud	personally
20:51	sillycloud	i call zomg_var_to_lower_case() on all my string literals
20:51	iBong	is there a way to set case insensitive on #"regex" ?
20:55	arbscht	iBong: #"(?i)..."
20:58	iBong	thats it, nice, thank you
21:11	Raynes	,(apply str (repeat 10 "(") (repeat 10 ")"))
21:11	clojurebot	"clojure.lang.LazySeq@b0c04bc1))))))))))"
21:11	hiredman	prn-str
21:12	Raynes	Cool.
21:29	gjohnson	hi there.
21:30	gjohnson	I had a lovely time reading through the code for pcalls, pmap, clojure.lang.Agent, Executors, and ThreadPoolExecutors, and I've just got one question at this point.
21:31	gjohnson	what is the rationale for the fixed threadpool's size being 2 greater than the number of cores?
21:32	gjohnson	anyone?
21:33	gjohnson	anyone alive out there?
21:33	rhickey	gjohnson: there isn't any special rationalization, other than it isn't worthwhile to create many more computation threads than you have procs
21:34	gjohnson	rhickey: indeed. I was wondering specifically why you chose to create 2 more than the number of cores.
21:34	gjohnson	rhickey: i.e. is there a performance benefit I'm missing from not having the same number or perhaps 1 or 3 more?
21:35	rhickey	gjohnson: no, it's still subject to change
21:35	gjohnson	rhickey: alright. until I finally took the time to poke into clojure.lang.Agent, I had been creating 2*cores agents and queueing up all my work evenly across them.
21:35	clojurebot	clojure is far closer to perfection then python
21:36	gjohnson	rhickey: that worked great on my 2 core machine (since it incidentally creates the same number of agents as your thread formula creates worker threads)
21:36	gjohnson	rhickey: but when I scaled up to more agents (say 4*cores), performance was degraded considerably.
21:37	rhickey	gjohnson: using send or send-off?
21:37	Chouser	,(.getMaximumPoolSize clojure.lang.Agent/pooledExecutor)
21:37	clojurebot	3
21:37	gjohnson	rhickey: just send, so they were all using the fixedThreadPool
21:38	gjohnson	rhickey: seemed like the extra agent overhead was just bogging down my processor without getting any advantage since there were no free threads.
21:38	rhickey	the number of agents shouldn't matter much then
21:39	gjohnson	rhickey: well, I need at least at many as there are threads in the thread pool, otherwise I'll have idle threads.
21:39	rhickey	it doesn't work that way, there are just jobs in the queues getting pulled off by threads, agents don't do anything themselves
21:39	gjohnson	rhickey: right, I understand that, but each agent can only have one thread running on it at a time.
21:39	rhickey	yes, I mean when more than numThreads
21:40	gjohnson	rhickey: so if there are less agents(work queues) than threads...well, you know.
21:40	rhickey	so 4 threads is 4 threads even if 100 agents
21:40	gjohnson	rhickey: right, but 4 threads is 3 threads if you only have 3 agents.
21:40	hiredman	…
21:40	rhickey	but you said you added more agents and it slowed down
21:40	rhickey	4 *cores
21:41	gjohnson	rhickey: yes, that was my observation. sorry, these are two different issues.
21:41	gjohnson	rhickey: what I noticed was that as I added more agents, my processors were being utilized less.
21:41	rhickey	do the agents do any io?
21:41	gjohnson	rhickey: usage was dropping maybe 10% or so on each one.
21:42	gjohnson	rhickey: no, they are doing CPU bounded work (agent-based simulation, no punning intended)
21:43	gjohnson	rhickey: perhaps I was just having a strange day at my machine. shrugs
21:43	rhickey	and was each job the same size?
21:44	gjohnson	rhickey: no, they were exploratory tree searches on variable-sized trees.
21:44	gjohnson	rhickey: but each time I ran the simulation, it was on the same set of trees.
21:44	rhickey	well, than some agents could be done and doing nothing and the work serializes on the others
21:44	gjohnson	rhickey: the data didn't change, just possibly the order that the agents got around to processing them.
21:44	rhickey	then
21:45	rhickey	forkjoin was made for this
21:45	rhickey	because it does work stealing
21:45	rhickey	you should try par
21:45	gjohnson	rhickey: hmm...I see.
21:45	gjohnson	rhickey: the code pcalls (i.e. pmap) looked very promising for my application actually.
21:46	rhickey	agents are really for stateful asynchronous things, for threading you can just use futures, or ideally, try the par branch
21:46	gjohnson	rhickey: since I designed my algorithm to break all the work up into a sequence of fns, the futures via pcalls seems like the best idea.
21:46	rhickey	in the par branch you can put you job data in a vector and call pvmap, get work stealing etc for free, no coordination
21:47	gjohnson	rhickey: when I wrote this code, I don't believe futures had yet been added to the language spec.
21:47	gjohnson	rhickey: I've been at these algorithms for the past few years.
21:47	rhickey	gjohnson: right, I saw so many people using agents because they were a path to threads, but they weren't doing anything actor-like with them
21:48	hiredman	futures weren't present
21:48	gjohnson	indeed
21:48	gjohnson	ack, I seem to have missed that.
21:49	gjohnson	rhickey: well, the truth of the matter is IMHO that once you're living and breathing clojure all day and night, it's really a thorn in my side to have to call out to the java Thread and Executors classes to get my concurrency off the ground.
21:50	rhickey	gjohnson: I'm not recommending that
21:50	gjohnson	rhickey: I had written my code with those initially, then switched it over to evenly queueing the tasks over a bunch of agents and passing it off to the threadpool transparently.
21:50	gjohnson	rhickey: I've been off in academic development land for quite awhile, and I just came back around and reread all the docs on the website, so futures were a new thing for me this week.
21:50	gjohnson	(yes, I know that's silly sounding, hiredman)
21:51	gjohnson	and I didn't find the parallel library terribly transparent in the website docs.
21:52	gjohnson	and the jsr166y link just led to a page that ultimately led me on to the ForkJoin API without a high-level explanation of what it was supposed to be doing.
21:52	gjohnson	rhickey: so I've been having a bit of trouble getting traction with it.
21:52	gjohnson	rhickey: any pointers on reading references to grok this stuff?
21:53	rhickey	http://github.com/richhickey/clojure/blob/26f5aed73c9cc2959beee0dd43a0d434fce83631/src/clj/clojure/par.clj
21:54	rhickey	there is a par branch on github, try it
21:54	rhickey	gotta run
21:54	gjohnson	alright, thanks
22:38	bpattison	I'm looking for a clojure function that tells me if a symbol is in a list -- I know it something obvious but I can't find it in the Clojure API or clojure.contrib
22:38	Chouser	are you sure you want a list and not a set?
22:39	Chouser	,(#{:a :b :c :d} :c)
22:39	clojurebot	:c
22:39	Chouser	,(#{:a :b :c :d} :e)
22:39	clojurebot	nil
22:39	hiredman	(.contains '(:a :b :c) :a)
22:40	hiredman	,(.contains '(:a :b :c) :a)
22:40	clojurebot	true
22:40	bpattison	yep, that's perfect -- I'll use a set
22:40	hiredman	,(.contains [:a :b :c] :a)
22:40	clojurebot	true