#clojure – 2009-09-30

2009-09-30

00:45	thedoor	wow
00:45	thedoor	a lot of people at this channel :P
00:45	arbscht	it is growing :)
00:46	thedoor	a work friend talk to me about clojure today, now im installing the nescessary thing to star :)
00:46	thedoor	start*
00:51	Raynes	thedoor: And it's the middle of the night for a lot of people. We get new people in here all the time.
00:52	Raynes	thedoor: I hope you have lots of fun with Clojure. :)
00:52	thedoor	here is 02:00 am :)
00:53	thedoor	i need to sleep but i need to star this :P
00:53	thedoor	start*
00:53	thedoor	sorry my bad english :(
01:44	hiredman	clojurebot: ping?
01:44	clojurebot	PONG!
02:56	yason	re
03:06	vy	Reading the Agents documentation... What does "dispatch" in "If during the function execution any other dispatches are made (directly or indirectly), they will be held until after the state of the Agent has been changed." expression?
03:07	hiredman	send or send-off
03:08	vy	Thanks.
03:11	vy	Dispatches on an agent are placed in a queue that are consumed by a thread pool? Right? If so, assuming mutation of the agent is serialized, what's the necessity/role of a thread pool in here? Wouldn't it be more simple to dedicate a single thread for an agent and its dispatch queue?
03:13	vy	Hrm... Or that "thread pool" is dedicated to "whole available agents"?
03:14	hiredman	send sends to a fixed size threadpool (I think it as 2+core count last I checked)
03:14	hiredman	send-off sends to a non-fixed size threadpool
03:15	hiredman	well, have 200 threads try to run on a 2 core machine is kind of, well, not entirely optimal
03:16	jdz	hiredman: optimal for what?
03:17	jdz	when i look in the process browser on my machine, there are 261 threads running currently
03:17	hiredman	thats a lie
03:17	vy	Hrm... Yeah, that's for sure. What I now realize is that: 1) There is a single (actually 2, one for "send" and another one for "send-off") dedicated thread pool and mutations to a single Agent is serialized. 2) No more than one thread tries to write to the same Agent at a time in the pool.
03:17	jdz	hiredman: says who?
03:18	hiredman	the processor is switching back and forth between 261 processes
03:18	hiredman	jdz: I just did (according to my logs)
03:18	jdz	well, then by your logic, what's the point of having 2+core count threads? why 2 more?
03:19	jdz	and JVM starts with how many threads?
03:19	hiredman	jdz: because threads might block for io or whatever
03:19	jdz	so, the original question stands...
03:19	vy	Sun Java System Application Server 9.1 Performance Tuning Guide > Thread Count [http://docs.sun.com/app/docs/doc/819-3681/6n5srlhod?a=view]
03:26	jdz	vy: that's about very specific purpose
03:27	jdz	vy: namely HTTP request handling
03:27	hiredman	everyone has some heuristic
03:28	hiredman	I've seen 1+core count, 2+core count, 5*core count
03:28	hiredman	there was a paper in acm like eight years ago that did some real math on the subject
03:28	jdz	lol, "real math"
03:29	hiredman	well, more than multiplication and addition
03:33	Fossi	hi
03:38	Fossi	cgrand: sorry to bother you, but you seem to be involved into hashmap development: is there a chance to make it serializable?
03:52	cgrand	Fossi: personally I don't care about java serialization so I won't push for it. Plus other clojure data structures doesn't implement serializable either.
03:54	Fossi	cgrand: well, at least in my clojure PersistenHashMap implements Serializable, but BitmapIndexedNode prevents it from actually being it
03:54	cgrand	hmm let me check
03:55	Fossi	and it's a pita to serialize a deep map (for memcache/jcache in this case) either way through serializable (because of maps and keywords) and through the reader because of 'custo,m' java classes that we can't read back
03:56	Fossi	so wanting to cache an object graph, we're stuck with walking it and transforming or such
03:56	hiredman	Fossi: you know about print-dup and #=()?
03:56	Fossi	hiredman: what does #=() do?
03:56	hiredman	#=() allows for limited read time evaluation
03:56	Fossi	i've tried print-dup and we couldn't get the things to read back in
03:57	hiredman	so you hook a custom method into, uh, I think it's called print-method
03:57	hiredman	,(doc print-method)
03:57	clojurebot	"; "
03:57	hiredman	how useful
03:57	Fossi	;D
03:57	hiredman	which prn uses
03:57	Fossi	yeah, i get that part, but how do i read it back in?
03:58	hiredman	it emits a some thing like "#=(java.util.ArrayList (1 3 4 5))"
03:58	hiredman	,#=(java.util.ArrayList (1 3 4 5))
03:58	clojurebot	Can't resolve java.util.ArrayList
03:58	hiredman	,#=(java.utils.ArrayList (1 3 4 5))
03:58	clojurebot	java.utils.ArrayList
03:58	hiredman	bah
03:58	hiredman	,#=(java.utils.ArrayList. (1 3 4 5))
03:58	clojurebot	java.utils.ArrayList
03:58	Fossi	ok, that should prolly work
03:59	hiredman	,#=(java.util.ArrayList. '(1 3 4 5))
03:59	clojurebot	#<ArrayList [quote, (1 3 4 5)]>
03:59	hiredman	,#=(java.util.ArrayList. (1 3 4 5))
03:59	clojurebot	#<ArrayList [1, 3, 4, 5]>
03:59	hiredman	gah
03:59	hiredman	read time arraylist
04:01	Fossi	hiredman: i'll try and snoop around some more in the read and print code
04:05	cgrand	Fossi, for the record, PersistentHashMap implements Serializable through AFn and AFn was marked Serializable in commit "cleanup for findbugs" http://github.com/cgrand/clojure/commit/7cd3b285328e7e7e71b23080303d66640e0f21e8#diff-0
04:14	LauJensen	Morning cgrand
04:14	cgrand	morning LauJensen
04:24	Fossi	cgrand: ok, so i guess that's a smallish 'bug' then
04:24	Fossi	since a lot of stuff is marked serializabel since a lot of things implement ifn :)
04:24	cgrand	,(ancestors (class #(42)))
04:24	clojurebot	#{clojure.lang.IMeta clojure.lang.IFn clojure.lang.IObj java.lang.Object clojure.lang.Fn clojure.lang.Obj java.util.Comparator java.util.concurrent.Callable java.lang.Runnable :clojure.contrib.generic/any java.io.Serializable clojure.lang.AFunction clojure.lang.AFn}
04:26	Fossi	also a bit weird to have something callable marked as serializable :)
04:36	serp_	,(funcall)
04:36	clojurebot	java.lang.Exception: Unable to resolve symbol: funcall in this context
04:37	serp_	,(#(42))
04:37	clojurebot	java.lang.ClassCastException: java.lang.Integer cannot be cast to clojure.lang.IFn
04:37	serp_	,#(42)
04:37	clojurebot	#<sandbox$eval__4764$fn__4766 sandbox$eval__4764$fn__4766@1324de2>
04:37	serp_	I don't get it =(
04:37	serp_	why would the second one work.
04:39	funkenblatt	well i think a better question is why the first one didn't work
04:39	funkenblatt	the second one is supposed to work
04:41	serp_	I'm talking about ,(#(42))
04:41	funkenblatt	oh
04:41	serp_	shouldn't #(42) return a lambda function that simply returns 42, and is later evaluated
04:43	Fossi	,(macroexpand-1 '(#(42)))
04:43	clojurebot	((fn* [] (42)))
04:44	serp_	oh I see...
04:44	Fossi	a smallish annoyance and source of common confusion
04:44	Fossi	but better than having to type #(( all the time
04:45	serp_	I don't know... seems like magic to me... makes the language harder to understand
04:45	opqdonut	you can always use #(int 42) ;)
04:45	serp_	unless...
04:45	serp_	,(macroexpand-1 '#42)
04:45	clojurebot	No dispatch macro for: 4
04:46	serp_	,#42
04:46	clojurebot	No dispatch macro for: 4
04:46	serp_	nope
04:46	Fossi	# is not the reader macro #( is
04:46	opqdonut	anyways, (constantly 42) is better
04:46	Fossi	sind # is used for a lot of things
04:46	Fossi	*since
04:46	opqdonut	, ((constantly 42) 0)
04:46	clojurebot	42
04:47	serp_	so a construct with #( has an implied function call in it
04:47	Fossi	otherwise you would have to type #((some-fn %)) all the time
04:48	serp_	sure, but then it is very clear what's going on =)
04:51	jdz	no, if just don't use the #() construct at all
04:51	jdz	then it's very clear what's going on
04:57	Chousuke	#() is a very useful shortcut sometimes, but it's not intended to replace fn
04:58	Chousuke	you can think of # as transforming a single expression into a function, if you want :P
04:59	Chousuke	(and that expression must be a list form)
05:27	Fossi	sometimes i miss #([(:a %) (:b %)]), but destructuring works better anyway
05:41	Chousuke	Fossi: or the new juxt
05:41	Chousuke	,((juxt :a :b) {:a 1 :b 2 :c 3})
05:41	clojurebot	java.lang.Exception: Unable to resolve symbol: juxt in this context
05:41	Chousuke	hm
05:42	Chousuke	clojurebot: you're old :(
05:42	clojurebot	No entiendo
05:43	Chousuke	Fossi: well, anyway, a new function was added recently, and #(vector (:a %) (:b %)) = (juxt :a :b)
05:44	cgrand	,(#(map % [:a :b]) {:a 1 :b 2 :c 3}) ; not a vector though -- and juxt is better
05:44	clojurebot	(1 2)
05:44	serp_	,(#(list % %) 1 2)
05:44	clojurebot	java.lang.IllegalArgumentException: Wrong number of args passed to: sandbox$eval--4786$fn
05:44	cgrand	(#(list %1 %2) 1 2)
05:45	serp_	what does just % mean?
05:45	Chousuke	same as %1
05:45	serp_	oh
07:42	Drakeson	how can I use incanter on a headless machine? incanter imports processing.PApplet, which apparently asks for a graphical environment. Is there a standard way to deal with graphical packages when you run on a headless machine and don't need them but they are imported in the package?
07:46	dliebke	Drakeson: I shouldn't import PApplet in incanter.core, I'll try and remedy that today
07:47	Drakeson	thanks
07:49	Drakeson	one more thing, incanter places the dependencies in the git repos. maybe what compojure does is a bit better. compojure zips and puts the deps in github using the `upload' feature, and just places a link in the main repo.
07:50	dliebke	Drakeson: there has been discussion about using Maven to manage dependencies
07:52	Drakeson	that would be cool :) (I often follow discussions about package managers, and dependency resolvers).
08:38	Fossi	i'm getting a "Caused by: java.lang.NoSuchMethodError: clojure.lang.Namespace.importClass(Ljava/lang/Class;)Ljava/lang/Class;" when building from ant, but not from emacs using the same clojure.jar
08:39	Fossi	it's behaving really weird
09:00	ar1	It's so quiet.
09:16	Fossi	ar1: so?
09:16	Fossi	btw: my problem was solved by running ant clean and thus deleting our target dir. no idea what went wrong inbetween
09:38	AWizzArd	Is there an easier way to do (.toString (.format (java.text.SimpleDateFormat. "yyyy-MM-dd_HH-mm-ss-SSS") (java.util.Date.) (StringBuffer.) (java.text.FieldPosition. 0)))?
09:38	AWizzArd	,(.toString (.format (java.text.SimpleDateFormat. "yyyy-MM-dd_HH-mm-ss-SSS") (java.util.Date.) (StringBuffer.) (java.text.FieldPosition. 0)))
09:38	clojurebot	"2009-09-30_06-43-03-925"
09:39	rsynnott	heheh, poor clojure hasn't quite escaped the terrifying Java date classes, then? :)
09:39	rsynnott	someone should write a nice wrapper
09:39	chouser	someone did
09:39	chouser	they got shouted down for not just using JediDate or whatever it's called
09:40	chouser	,(.format (java.text.SimpleDateFormat. "yyyy-MM-dd_HH-mm-ss-SSS") (java.util.Date.))
09:40	clojurebot	"2009-09-30_06-45-36-693"
09:41	rsynnott	(presumably there on the basis that the world might arbitrarily start using a different calendar)
09:41	AWizzArd	chouser: oh good, this is much nicer
09:42	AWizzArd	strange i didn't see this on the api page of SimpleDateFormat
09:44	manic12	this may sound like backward steps to clojurites, but it is true that ordinary single-object dispatch (smalltalk/flavors style) is quite useful to engineers...
09:45	chouser	manic12: sure. Clojure uses it all the time.
09:45	manic12	I haven't learned enough clojure to do that in clojure
09:45	chouser	because the JVM can do it fast, mostly.
09:46	manic12	and I also haven't learned how to create generative style extensions to that
09:47	chouser	conj is single-dispatch on the type of the first arg
09:48	chouser	for example
09:50	manic12	another thing, I haven't learned how to write declaratively in clojure, just functional with procedural aspects mostly talking to java libs
09:52	Fossi	re serialization: now we have some trouble serializing structs. it seems that they are treated as an persistentMap and that has a "create" method, but the structs don't
09:53	Fossi	so naturally, things break
09:57	manic12	I would like to create an 'object' which behaves a bit like a map, that instead of simply fetching the value, it calls a function on itself to compute and memoize the value, seems a bit anti-immutable
09:59	manic12	but maybe that's ok, because it would be easier to implement 'undo'
09:59	manic12	anybody there?
10:00	chouser	did you ask a question?
10:00	manic12	i am asking how to do that sort of thing with clojure
10:01	chouser	one way would be to implement clojure.lang.IPersistentMap using proxy
10:01	manic12	what about just deriving a actual java class?
10:03	chouser	,(let [m (proxy [clojure.lang.APersistentMap] [] (valAt [_] 5) (seq [] nil))] [(get m :foo) (:bar m) (seq m)])
10:03	clojurebot	[5 5 nil]
10:04	tmountain	chouser: that's pretty nifty
10:06	jdz	why the hell does slurp declare the f parameter as String?
10:07	tmountain	jdz: it takes a string pointing to the file
10:07	jdz	and why does it care if i pass it an instance of java.io.File?
10:07	tmountain	$ echo "foo" > /tmp/foo.txt
10:08	tmountain	(slurp "/tmp/foo.txt")
10:08	tmountain	"foo\n"
10:09	jdz	tmountain: and notice that (slurp (new java.io.File "/tmp/foo.txt")) does not work :/
10:10	tmountain	jdz: yeah, looking at its implementation, slurp creates a FileInputStream in the background with f as it's argument
10:10	jdz	despite the fact that java.io.FileInputStream accepts java.io.File in the constructor
10:10	tmountain	jdz: yeah, that does sort of limit it's utility
10:13	tmountain	jdz: slurp uses the two argument form for the FileInputStream constructor
10:14	tmountain	jdz: to allow specification of encoding
10:14	jdz	tmountain: your point being?
10:14	jdz	tmountain: it uses it anyway even when passed only one parameter
10:15	tmountain	jdz: point being, you'd need another declaration to make it work with both a File and String
10:15	jdz	no
10:16	jdz	just drop the #^String thing
10:16	jdz	easy as that
10:16	manic12	chouser: how do I access the proxy object if valAt is being defined in the proxy definition?
10:17	manic12	i need the keys to be able to reference other keys
10:17	tmountain	jdz: my mistake, I see it's the InputStreamReader that takes the encoding argument
10:18	jdz	tmountain: my gripe with this is that somebody added the declaration although it does no good.
10:18	jdz	tmountain: same for the encoding
10:19	tmountain	jdz: yeah, you're right, I removed the type hint, and it works with a File as well
10:20	Chousuke	I suppose the type hint gives no measurable performance gain. :P
10:21	Chousuke	so it'd be better to remove it.
10:21	tmountain	yeah, I'd imagine there's miniscule difference in performance
10:22	Chousuke	if you wanted performance you wouldn't use slurp anyway
10:22	Chousuke	it's wasteful ;P
10:22	tmountain	using string builder
10:22	Chousuke	str uses a string builder internally. I mean storing an entire file in a string
10:23	tmountain	wasteful memory wise
10:24	manic12	chouser: if you're there I want to post something
10:28	AWizzArd	Compojure users: anyone here who handles json requests?
10:36	manic12	i would like to use clojure in a practical application, is there anyone here who can discuss how i might be able to satisfy some of the requirements with clojure?
10:38	manic12	is anyone here at all?
10:39	rsynnott	hiding :)
10:40	manic12	must be a coffee break
10:44	cemerick	we're all working :-P
10:44	arbscht	manic12: state your question and anyone who can answer it may...
10:46	manic12	clojurebot: lisppaste
10:46	clojurebot	It's greek to me.
10:46	manic12	is there a specific lisppaste for this irc channel?
10:47	ambient	M-x package-install lisppaste ;)
10:47	manic12	tehe
10:49	Fossi	lisppaste8: url
10:49	lisppaste8	To use the lisppaste bot, visit http://paste.lisp.org/new/clojure and enter your paste.
10:50	Fossi	manic12: there you go
10:51	lisppaste8	manic12 pasted "how to do in clojure?" at http://paste.lisp.org/display/87950
10:51	manic12	thx
10:58	raek	how do I add my own directory to the CLASSPATH and still have the library paths in it?
10:58	raek	or -- how do I find my system default CLASSPATH?
10:59	raek	I ran "find /usr -name \*.jar -print" to find the jars in my system
10:59	raek	but they're evererywhere!
11:05	liwp	raek: on unix with bash: export CLASSPATH=path/to/foo.jar:path/to/classes:$CLASSPATH
11:05	raek	yes, I know
11:06	raek	but I sort of found a solution to my problem
11:06	liwp	ok, so what's the problem exactly
11:06	liwp	ok
11:06	raek	I specified the path to my own clj files with the -cp option to java
11:06	liwp	raek: that'll work
11:06	raek	but then I also had to add the path to the system jar files
11:07	liwp	you shouldn't have to do that. What do you mean with system jar files?
11:08	raek	I'm trying to get compojure running
11:08	raek	but it turned that I missed a dependency
11:09	raek	java.lang.ClassNotFoundException: org.mortbay.jetty.servlet.Context (jetty.clj:0)
11:09	raek	this is what I get when evaluating (use 'compojure)
11:10	raek	oh wait
11:10	raek	forgot to put the jar files produced by "ant deps" into my classpath
11:11	raek	you may ignore my previous 10 lines
11:12	liwp	heh
11:12	raek	java.lang.SecurityException: sealing violation: package org.mortbay.jetty.servlet is sealed (jetty.clj:0)
11:13	raek	ok, what have I done now?
11:21	raek	ok, working now
11:22	raek	removed jetty and firends with my package manager and used the jars supplied by compojure
11:29	kunjan	Is Clojure dynamically scoped in some cases?
11:36	yason	kunjan: I'm not an expert by any means but IIRC no. You could probably simulate dynamic scope with vars and (binding ...)
11:40	raek	(let) in clojure introduces lexically scoped variables
11:40	raek	and (binding) dynamically scoped ones
11:44	rhickey	raek: they aren't dynamically scoped, just dynamically bound. If you do (defn foo [] (use-some free-var)) it will fail at compile time if free-var isn't in scope, vs true dynamic scope where someone could establish free-var then call foo and have it find it
11:45	sproingie	http://onclojure.com/2009/05/06/simulating-dynamic-scoping/
11:46	raek	ah, please bear with me. i've only used clojure for a week
11:46	sproingie	doesn't look like a general solution tho
11:47	raek	:rhickey ah, I see.
11:49	raek	these things are new to me, as I come from languages that doesn't have these things
11:49	raek	or I wasn't aware of them
11:53	raek	is it un-ideomatic to use variables with dynamic bindings as a way to not have to pass a certain arg to function?
11:53	raek	i'm coding some things that calculates the width and height of texts that will be drawn
11:53	raek	and these things depends on the Metrics of the font in use
11:54	rhickey	raek: if you are just starting out you should avoid dynamic binding
11:54	yason	raek: so what you want is an easy way to pass on state?
11:55	stuartsierra	I think it's idiomatic, but subject to a number of tricky bit when laziness & concurrency are involved
11:55	raek	(binging [metrics (.. label .getFont .getMetrics)]] (calculate-width "text"))
11:56	raek	and (calculate-width) then uses metrics
11:57	raek	but as the width is a function of both the text and the metrics, maybe I should pass along metrics as an arg
11:58	raek	that is the way I do it now
12:04	manic12_	are object systems in general anti-clojure philosophy?
12:05	rhickey	manic12_: http://clojure.org/state
12:05	Makoryu	manic12_: Stateful objects are the bane of both concurrency and common sense.
12:06	rhickey	I don't know that anyone has reconciled OO with the problems enumerated there, but I don't think it is impossible. Whether the result would be considered OO anymore is another question
12:07	rhickey	OO includes a lot of things that Clojure delivers a la carte, like polymorphism and hierarchy
12:07	manic12__	sorry my internet connection dropped, i think the router was rebooted
12:08	Makoryu	manic12__: http://clojure-log.n01se.net/ <- You can always use this to see what you missed.
12:08	manic12__	thanks
12:08	Makoryu	Or, er, not, I suppose
12:08	Makoryu	I thought it updated more frequently
12:09	chouser	every 15 minutes right now
12:09	chouser	20 minutes
12:10	Makoryu	manic12__: Anyway... https://gist.github.com/61e9ffb761b574428cc4
12:10	manic12__	i've got a declarative DSL built into clos, and i am interested in doing the same things with clojure
12:11	sproingie	might try spinoza or categories. not tried either myself so i couldnt say how good they are
12:12	sproingie	dammit..
12:12	sproingie	manic12__: ^^^
12:13	manic12__	are those on github or somewhere?
12:13	sproingie	google code i think
12:13	sproingie	teh googelz knows :)
12:13	Draggor	Are there any lexers for Clojure? Something like cl-yacc?
12:13	manic12__	oh wow
12:14	sproingie	neither spinoza nor categories defines a call-next-method tho, let alone a formal MRO
12:15	manic12__	i actually don't need call-next-method because it makes no sense in the DSL
12:16	manic12__	so categories is a drinking game
12:16	sproingie	hard to reconcile with the predicate dispatch of multifn's anyway
12:16	stuartsierra	Draggor: you could use ANTLR or JavaCC to generate a lexer
12:17	sproingie	manic12__: categories is at http://code.google.com/p/explorersguild/
12:17	sproingie	somewhere in the source
12:17	Draggor	stuartsierra: the JavaCC site appears to be down. Does one have any particular features over the other?
12:18	cemerick	wasn't cgrand working on a slick clojure DSL for grammars?
12:18	sproingie	antlr has oodles of features. also a vertical learning curve, and some annoying conventions around lexer/parser distinctions
12:20	cgrand	cemerick: still working
12:22	cemerick	cgrand: just giving you the intro to do some promotion :-)
12:22	stuartsierra	Draggor: ANTLR is generally considered easier to use, but it requires a runtime library to run your parser. ANTLR generates recursive-descent parsers, JavaCC generates (I think) bottom-up parsers
12:23	cemerick	I got tired of javacc and antlr years and years ago.
12:23	stuartsierra	For very simple grammars (e.g., Lisp) they're overkill.
12:23	stuartsierra	But essential for parsing something like C.
12:25	manic12__	cannot each change to an object be a new structure entirely?
12:25	Draggor	I'm just having fun and writing a parser for a die roller, something like [5d6] or [[1d6]d6]
12:26	Draggor	And then put math operations in it too
12:26	rsynnott	stuartsierra: quite a few C compilers have awful hand-made parsers
12:27	Draggor	Really it's just a calculator with another operator d and using [] for grouping instead of ()
12:27	sproingie	isn't g++'s new parser a hand-rolled recdescent parser?
12:28	stuartsierra	rsynnott: really? ugh
12:32	manic12__	i don't understand the difference between altering a ref and mutating state
12:34	sproingie	they are state
12:34	sproingie	but you can only change them in a transaction
12:34	manic12__	it says that is "identity"
12:34	manic12__	it doesn't matter if it's transactional
12:35	sproingie	clojure doesn't lack state completely, it just shoves it into some well-defined areas
12:36	manic12__	that statement should be put on the clojure.org/state webpage
12:39	sproingie	that page gets a little overly focused on identity. refs are essentially first-class slots, just not necessarily "owned" by some structure.
12:39	manic12__	ok
12:40	manic12__	but oo is a bad idea?
12:40	sproingie	hardly
12:41	sproingie	oo doesn't even necessarily have to throw around uncoordinated mutable state
12:41	sproingie	it's just that most oo languages have few alternatives to doing so
12:42	sproingie	java has some immutable collections in libraries that are pretty efficient, but most people don't use them
12:42	sproingie	C++ STL is pretty much hopeless
12:43	sproingie	scala's the only oo language i can think of that has seriously decent functional objects
12:43	manic12__	I'm looking at the "memoize" example and thinking about declarative oop
12:45	arohner	manic12__: in clojure, refs, atoms etc are a way to cleanly declare and isolate when & how state changes happen
12:45	manic12__	not absolute declarative like prolog though, declarative where it makes sense to do so: storing knowledge
12:45	arohner	yes, they're all state changes, but clojure has much stricter guarantees about how they happen
12:46	sproingie	clojure has all the state of any imperative language, but everything stateful in clojure has some kind of synchronization primitive baked into it
12:46	sproingie	refs are transactional, agents have asynchronicity, and atoms are, well, atomic
12:46	sproingie	clojure has all kinds of state, it just doesn't have "naked state"
12:46	manic12__	i haven't determined the difference yet
12:47	manic12__	that's good
12:47	manic12__	because it will make it easier to implement undo
12:48	sproingie	you could easily write a clojure program in completely imperative style, it's just that clojure steers toward a functional style
12:48	sproingie	by using immutable collections and so forth
12:49	rhickey_	a point of Clojure is to separate identity, state and value
12:49	manic12__	so clojure needs some object systems
12:49	rhickey_	people throw around the term state without considering its meaning
12:50	manic12__	i read the clojure.org/state page, and i'm one of those people apparently
12:50	sproingie	most people who aren't FP geeks will conflate them somewhat :)
12:51	sproingie	"pure" fp geeks don't even go in for the concept of "identity"
12:51	sproingie	insofar as being distinguishable from value that is
12:51	manic12__	fp=functional programming?
12:51	sproingie	yep
12:51	milep	Hello, is there a good/standard way to use structmaps with sorted-set? Or should I create custom compare function and use that with normal set?
12:51	dnolen	sproingle: being the creator of spinoza, I would say don't bother. I haven't touched that project in ages :) it's best to just dive into fp style programming.
12:51	rhickey_	by being atomic, and pointing to immutable composites, Clojure's reference types ensure transitions from value to value, and you can grab one of those values without getting trounced by future state transitions
12:52	rhickey_	contrast that with objects that change in your hands
12:52	sproingie	dnolen: aw. i rather liked the idea of first-class protocols.
12:52	rhickey_	those objects don't have obtainable state at all
12:52	dnolen	sproingie: yes, I think mikel's generic functions project is less half-baked tho.
12:53	sproingie	dnolen: is that the categories thing in explorersguild?
12:53	manic12__	so if one thread grabs the state from the ref, it won't change on it underneath if another thread grabs it next
12:53	dnolen	sproingie: yes
12:53	rhickey_	a state is a value at a point in time, being a value it can't change
12:53	rhickey_	references just manage time
12:54	rhickey_	objects typically conflate all three and become impossible to reason about
12:54	manic12__	the most unfortunate thing about my code is that i use a huge c++ library!
12:56	manic12__	but is my statement correct? about the state not changing from under a function in one thread because it's immutable?
12:57	rhickey_	manic12__: state of what? values don't have state. If you get a value from a ref you are done with state
12:57	manic12__	ok, so yes
12:58	manic12__	the state doesn't change because it's not a state in the first place it's an immutable value
12:59	rhickey_	manic12__: there is a thing with state (the reference). When you dereference it, you obtain its value at a point in time. That value is an immutable thing and has no state/time at all
12:59	manic12__	i understand now
13:00	rhickey_	I don't mean to be pedantic, but I am on a mission to clarify this I guess. So tired of OO muddle...
13:01	manic12__	the c++ library suffers from exactly this stuff clojure is intended to solve
13:01	rhickey_	Clojure disassembles OO and delivers it in orthogonal pieces
13:01	manic12__	that's fine, i get it
13:03	manic12__	but to do what I have coded in clos will require implementing some constructs i didn't have to before, that's what i was wondering
13:03	rhickey_	the prize is obtainable values and the fact that you can write pure functions that work with them, completely independent of things relating to state. In most OO values are relegated to primitives
13:03	manic12__	i should say meta programming in clos
13:04	rhickey_	manic12__: porting from CLOS will require looking at how it is being used, what's stateful, what's inheritance driven etc. It is a rich polymorphism system for sure
13:04	manic12__	it uses memoization heavily
13:06	manic12__	i don't want to introduce uses of 'state' just because that's the way clos does it
13:08	manic12__	the C++ code often destroys objects you pass into functions and you have to explicitly copy them first if you want to keep it
13:10	manic12__	so it will be a challenge to make it appear like native code
13:11	chouser	"Take an old man's word; there's nothing worse than a muddle in all the world. It is easy to face Death and Fate, and the things that sound so dreadful. It is on my muddles that I look back with horror--on the things that I might have avoided." -- http://tinyurl.com/ychwqfp
13:18	rhickey_	Chouser: love it!
13:18	Draggor	Anyone tried http://kotka.de/projects/clojure/parser.html ?
13:19	chouser	rhickey_: I'm not sure he was talking about identity-value conflation, but it'll do.
13:30	drewr	,(#{false} [true false])
13:30	clojurebot	nil
13:30	drewr	shouldn't that be false?
13:31	chouser	,(#{false} true false)
13:31	clojurebot	java.lang.IllegalArgumentException: Wrong number of args passed to: PersistentHashSet
13:31	chouser	hm
13:31	stuartsierra	,(#{false} false)
13:31	clojurebot	false
13:32	chouser	drewr: sets return the matching value, else nil.
13:32	stuartsierra	,(#{:a} :a)
13:32	clojurebot	:a
13:32	stuartsierra	,(#{:a} [:a])
13:32	clojurebot	nil
13:33	stuartsierra	,(some #{false} [true false])
13:33	clojurebot	nil
13:33	stuartsierra	,(some #{:a} [:b :a])
13:33	clojurebot	:a
13:35	drewr	chouser: isn't false the matching value though?
13:36	chouser	#{false} does not contain the vector [true false]
13:36	chouser	,(#{:foo [true false] :bar} [true false])
13:36	clojurebot	[true false]
13:42	hiredman	(#{false} false)
13:42	hiredman	,(#{false} false)
13:42	clojurebot	false
13:42	hiredman	,(#{nil} nil)
13:42	clojurebot	nil
13:42	hiredman	:(
13:43	chouser	,(#{'spam} 'spam)
13:43	clojurebot	spam
14:02	drewr	chouser: oh, ok; I thought it wanted a coll there
14:02	drewr	which is what stuartsierra was demonstrating :-)
14:04	drewr	which makes sense when the set is callable there
14:09	bitbckt	Anybody have an easy way of taking an Array of N^2 items and splitting it into an NxN matrix?
14:11	ambient	x+y*width
14:11	chouser	,(partition 4 (range 16))
14:11	clojurebot	((0 1 2 3) (4 5 6 7) (8 9 10 11) (12 13 14 15))
14:11	bitbckt	chouser: That's what I was looking for. Thanks.
14:16	tmountain	has anyone seen this before? http://pleac.sourceforge.net/
14:17	tmountain	it's a website featuring recipes from the perl cookbook implemented in a variety of programming languages
14:17	tmountain	good resource for finding out how to do something quickly in a given language
14:18	tmountain	I was thinking of starting a clojure module if anyone else is interested
14:19	tmountain	I contributed the ruby section on directories, and it was pretty fun
14:20	sproingie	pleac is pretty nifty
14:20	sproingie	good way to compare expressiveness
14:20	sproingie	adding clojure would be awesome
14:21	tmountain	sweet, I'll contact the project admin and see about getting it added
14:35	drewr	stuartsierra: your some example was also confusing me
14:36	stuartsierra	which one?
14:36	drewr	,(some #{:foo} [:foo false true])
14:36	clojurebot	:foo
14:36	drewr	,(some #{false} [:foo false true])
14:36	clojurebot	nil
14:36	stuartsierra	That happens because "some" tests the result of applying the predicate to an element in the sequence. "false" tests false!
14:37	drewr	that seems like a bug
14:37	stuartsierra	Not a bug, I think, but certainly an edge case.
14:38	stuartsierra	,(some #(false? %) [:foo false true])
14:38	clojurebot	true
14:38	chouser	you just have to be careful dealing with collections that might contain nil or false
14:46	drewr	kotarak: how's ClojureCheck coming?
14:47	kotarak	drewr: uhweuhm oehmm...
14:47	kotarak	drewr: falling off the spare time clip at the moment, I'm afraid.
14:48	kotarak	Although I'd really like to work on it.... I have to cut done my projects...
14:49	drewr	what's it lack?
14:49	kotarak	drewr: The TAP part is more or less complete, but the QuickCheck clone is a mess. I'd really like to clean that up.
14:51	drewr	k
14:51	kotarak	.. and make it work, actually...
14:52	drewr	a coworker started working on his own and didn't know whether to tell him to use yours :-)
14:53	kotarak	drewr: -.- I have to win in the lottery, so that I can quit my day job and start working full time on fun projects. )
14:53	kotarak	:)
14:54	drewr	yep
14:55	kotarak	the job cuts away 43h (effectively 48h) per week, with the recently grown family and soon to be bought apartment the prospects are rather depressing from the clojure project point of view. (Although the last two points have higher priority than clojure... :))
14:57	drewr	please put your family before clojure :-)
14:58	kotarak	Of course! :D
14:59	kotarak	drewr: but now that you asked, I will probably invest some time to drive CC a bit more. VimClojure ate a lot of time lately, too...
16:08	ngoc	Why "Logging str" is output 5 times in the example at http://clojure.org/concurrent_programming?
16:10	kotarak	ngoc: What do you expect? Have you had a look at the implementation of str?
16:10	arohner	lisppaste8: url
16:10	lisppaste8	To use the lisppaste bot, visit http://paste.lisp.org/new/clojure and enter your paste.
16:11	kotarak	ngoc: http://groups.google.com/group/clojure/browse_frm/thread/9f8b14cd7cdf9f4d/90158571284ac1ef?q=loves+ethel+lucy+group:clojure for details
16:12	lisppaste8	arohner pasted "get the value of a local at runtime?" at http://paste.lisp.org/display/87966
16:13	arohner	is there any way to get the value of a local at runtime? and is there any way to implement value-of in a way that doesn't use eval?
16:14	hiredman	uh
16:14	kotarak	arohner: the only way I see is just returning expr from my-fun.
16:14	arohner	is there a way to edit the paste?
16:14	kotarak	you can annotate it
16:14	hiredman	there is no value at runtime
16:14	hiredman	err
16:14	hiredman	at compile time
16:15	lisppaste8	arohner annotated #87966 "untitled" at http://paste.lisp.org/display/87966#1
16:15	hiredman	when is when macro expansion happens
16:15	hiredman	there is just a datastructure
16:15	kotarak	arohner: that's not possible, as hiredman says.
16:16	arohner	the macro turns into code that calls (my-fun 'foo)
16:16	arohner	my-fun is run at runtime
16:16	arohner	ah, the paste is missing a `
16:17	kotarak	And a quote on the expr.
16:17	lisppaste8	arohner annotated #87966 "untitled" at http://paste.lisp.org/display/87966#2
16:17	kotarak	But here expr gets evaluated.
16:17	manic12__	i implemented a "destructive" assoc for structmaps which obviously doesn't destroy anything but does allow for undo and redo, you have to pack the structmap in another map and make a ref to it and pretend the ref is the structmap
16:18	arohner	sorry for the mistakes. I'm trying to create a simplified example. obviously I'm making mistakes in the translation
16:18	arohner	`(my-fun expr)
16:18	hiredman	manic12__: sounds horrible
16:18	manic12__	that would have been a pain in CL because you need to copy the objects yourself
16:18	lisppaste8	kotarak annotated #87966 "quoted expr" at http://paste.lisp.org/display/87966#3
16:19	manic12__	hiredman: why?
16:19	hiredman	manic12__: because then you are going to have a lot of little "mutable" structmaps
16:20	manic12__	but you can undo them
16:20	arohner	interesting kotarak. is it possible to do the work in my-fun?
16:20	kotarak	arohner: which work?
16:20	tomoj	being able to undo doesn't make mutable state a good thing :)
16:20	hiredman	while controled mutation is better than uncontroled, immutability is better still
16:20	arohner	getting the value of expr. doing it in the macro rather than the function would make my code a lot more complicated
16:20	arohner	(or maybe my design is just wrong)
16:21	kotarak	arohner: well you do the work in the function. The macro just wraps everything up in a thunk, which is (maybe) called in the function to obtain the value.
16:22	manic12__	i want to be able for the user of the DSL to def to the ref and keep a constant ref to an dynamic object
16:23	hiredman	:(
16:23	manic12__	for animation and changing inputs/attributes
16:25	Chousuke	wouldn't a simple ref for the input do?
16:25	Chousuke	or inputs.
16:26	manic12__	how do you synchronize the state of the object to acheive the undo/redo?
16:26	lisppaste8	arohner annotated #87966 "untitled" at http://paste.lisp.org/display/87966#4
16:26	Chousuke	hm.
16:27	Chousuke	I'm not quite sure what it is that you are doing now
16:27	Chousuke	how does your scheme allow for undo/redo? :/
16:27	wavis	manic12__: instead of using a ref, just create and undo function that returns the previous map, which you had originally used assoc on
16:28	manic12__	and even if the struct map gets really big I don't think clojure's copying of it would slow it down compared to the fact that the object may need to recompute all of it's values
16:28	kotarak	arohner: that cannot work, you can't eval an expression passed to a function (at least not in general)
16:28	manic12__	generators?
16:28	manic12__	wavis
16:28	Chousuke	no, really, I have no idea what you're thinking
16:28	Chousuke	you might have a good idea but I can't make sense of it.
16:29	kotarak	arohner: besides the macro would need to look like: `(my-fun (quote ~expr)). Your expansion is (my-fun your.name.space/expr).
16:29	ngoc	kotarak: Thank you. This is a pitfall for new comers. Since this example is for new comers, I think it should be explained in the example.
16:30	manic12__	it's like a spreadsheet
16:30	manic12__	but it's a tree of objects
16:30	manic12__	(refering to clos implementation)
16:31	manic12__	inputs and specially declared modifiable attributes can be changed
16:31	tmountain	maniac12__: I'd determine if you really need "objects" or if you can use Clojure's native types to represent whatever data you're referring to
16:32	manic12__	i'm using structmaps from spinoza
16:32	Chousuke	hm
16:32	manic12__	i'm trying to implement a declarative DSL from clos to clojure
16:33	manic12__	i haven't used cells very much at all, but it's like cells but lazy eval
16:35	manic12__	it's not for specially trained computer scientists, it's a DSL for mechanical & civil engineer types
16:35	wavis	,(defn undoable-assoc [the-map key val] (with-meta (assoc the-map key val) {:undone the-map}))
16:35	clojurebot	DENIED
16:36	manic12__	wavis: at some point the user needs handlebars
16:36	wavis	manic12__: anyway, with that you could (^new-map :undone)
16:36	wavis	handlebars?
16:36	manic12__	probably
16:37	manic12__	there has to be at least one mutable thing in the program, and that is the object which gets displayed in the viewer
16:39	Chousuke	the object displayed is probably a function of the inputs, which are the mutable stuff :)
16:39	wavis	assuming I have a clue, you just need to change what the viewer is holding to display. the collections to display can still be immutable
16:39	ngoc	Is there a way to redefine functions? (defn str [& args] 123) will throw error.
16:40	wavis	so represent the whole display as an immutable structure and swap it
16:40	manic12__	the inputs are just going to be entries in the struct map and clojure will copy the whole struct map which gives the ability to synchronize an undo/redo
16:40	hiredman	ngoc: because str is not defined in the namespace you are in
16:40	manic12__	wavis: you are correct
16:41	ngoc	hiredman: How to solve the problem?
16:41	manic12__	but sometimes it's nice to spawn one at the repl
16:41	manic12__	if it's not viewing properly or something
16:42	kotarak	ngoc: you don't want to do this!
16:43	Chousuke	ngoc: redefining core functions is going to break everything
16:43	Chousuke	ngoc: what you might do is temporarily rebind them though.
16:43	hiredman	ngoc: there are many ways to solve it
16:43	ngoc	(ns clojure.core) then (defn str [& args] "123") solved the problem
16:44	hiredman	#$%@#$%
16:44	hiredman	that is not a solution
16:44	Chousuke	and created a thousand more :)
16:44	hiredman	clojure.core/str is part of the core of clojure and depended upon by other functions
16:44	hiredman	changing it breaks all those functions
16:44	kotarak	People sure have funny definitions for "solved" and "problem"...
16:44	manic12__	mutable is not a dirty word
16:44	hiredman	yes, it is
16:45	manic12__	then get rid of refs, agents and atoms
16:45	Chousuke	cutting off your head solved headache but...
16:45	Chousuke	solves*
16:45	kotarak	Chousuke: hmmm? Phantom pain?
16:45	sproingie	it's a dirty word, but sometimes your problem wants you to talk dirty ;)
16:45	ngoc	I want to do alias_method_chain like in Ruby/Rails: http://api.rubyonrails.org/classes/ActiveSupport/CoreExtensions/Module.html#M001219
16:45	technomancy	noooo!~
16:45	sproingie	clojure gives you mutability all over the place. it's just managed.
16:46	technomancy	alias_method_chain is a maintenance nightmare
16:46	hiredman	manic12__: but mutation is needed
16:46	hiredman	so you minimize/control it
16:46	manic12__	exactly
16:46	sproingie	heck atoms arent even all that managed
16:47	ngoc	I save the original str, redefine it, and call the original function inside the new one
16:47	sproingie	that's what let is for
16:47	hiredman	they are managed in the sense that future state is a function of past state
16:47	hiredman	ngoc: don't
16:47	kotarak	ngoc: and broke the whole system....
16:48	kotarak	ngoc: monkey patching is not a good idea. Don't do it.
16:48	ngoc	But Rails plugins all work this way, and they do work
16:48	hiredman	ngoc: when you create a new namespace you have the option of not aliasing bindings from clojure.core
16:48	technomancy	ngoc: no, they break
16:48	technomancy	all the time
16:49	hiredman	so you can leave str out of your namespace and redefine it there
16:49	manic12__	I have to implement a memoization scheme for rest of that attributes, but it has to be finer grained than copying the whole struct map, it has to memoize each attribute individually but be able to unbind itself when dependent things change
16:49	Chousuke	manic12__: copies are cheap
16:49	Chousuke	manic12__: just give the user a reference :P
16:49	manic12__	but for EVERY attribute call?
16:50	hiredman	attribute call?
16:50	manic12__	message
16:50	hiredman	I would not use spinoza
16:50	Chousuke	manic12__: in which situation would you even copy the struct map entirely? :/
16:50	technomancy	ngoc: rails 3 is removing as many alias_method_chain calls as possible due to this
16:50	Chousuke	manic12__: I mean, if you assoc to it, that doesn't copy the whole map
16:51	ngoc	If redefining does not "solve" the problem, is there a way? I want to add feature to an existing function, just like in Rails plugins.
16:51	Chousuke	(doc binding)
16:51	clojurebot	"([bindings & body]); binding => var-symbol init-expr Creates new bindings for the (already-existing) vars, with the supplied initial values, executes the exprs in an implicit do, then re-establishes the bindings that existed before."
16:51	kotarak	ngoc: you want multimethods
16:51	sproingie	"just like in Rails" often means "doing it completely wrong"
16:52	hiredman	I'm pretty sure the original author of spinoza even removed his repo
16:52	sproingie	monkeypatching everything onto Object... OOP, You're Doing It Wrong
16:52	hiredman	or not
16:53	sproingie	there's reasonable use cases for opening up base classes, sure. most of ruby's uses aren't it
16:53	manic12__	Chousuke: yeah but I should only need to copy the struct map when information "changes" that invalidates dependent attributes
16:53	Chousuke	manic12__: I still don't see what "copying" happens if the struct map doesn't change
16:54	sproingie	you don't need to copy the whole map, you just record a sequence of changes
16:54	manic12__	clojure copies just about everything you can look at the source code
16:55	sproingie	when you need to get everything into the same state, you can condense the changes to a minimal set then generate exactly one new object from that
16:55	hiredman	eh?
16:55	Chousuke	huh?
16:55	hiredman	copies?
16:55	Chousuke	manic12__: you need to copy something to keep the data structures immutable, but it doesn't copy everything all the time.
16:55	Chousuke	that'd be slow
16:55	manic12__	sproingie: you're right and that's what i do in clos
16:56	hiredman	http://blog.higher-order.net/2009/02/01/understanding-clojures-persistentvector-implementation/
16:56	Chousuke	hmm, well, that's possible with clojure transients.
16:56	manic12__	go to IPersistentMap assoc method and show me that it's not copying every value
16:56	Chousuke	manic12__: it's not.
16:56	Chousuke	it shares structure.
16:56	Chousuke	with its "parent" maps
16:56	ngoc	How to implement something like (defn f (+ (original-f) + 1)) with multimethods or binding? Binding does not work as expected as the example at http://clojure.org/concurrent_programming shows ("Logging str" is output 5 times).
16:56	manic12__	newVals[i]=val;
16:57	hiredman	ngoc: it does work as expected
16:57	kotarak	ngoc: bindings works as expected, the output depends on the definition of the original str
16:57	hiredman	str is called internally by println and the love function
16:57	Chousuke	manic12__: if you have a map with a thousand objects and you assoc to it, Clojure surely does not copy each of those thousand values
16:57	sproingie	love?
16:57	manic12__	PersistentStructMap.java
16:58	Chousuke	manic12__: assoc is O(log_32 n)
16:58	hiredman	sproingie: if I recall, the example on that page defines a function named love
16:58	sproingie	ah
16:58	hiredman	loves, pardon me
16:58	sproingie	i was digging around in the API looking for it ;)
16:59	manic12__	ok, Chousuke if you're right than that's even better
16:59	tomoj	Chousuke is right
16:59	Chousuke	manic12__: how many items will your struct maps have, anyway?
16:59	stuartsierra	~seen jochu
16:59	clojurebot	no, I have not seen jochu
16:59	ngoc	hiredman: I thought str called itself, not by println
16:59	manic12__	they can have alot, but not enough to make a full copy slow, like 100 to 200 "slots"
17:00	sproingie	log32 n aka "fast enough for you old man"
17:00	kotarak	ngoc: yes, str calls itself, which is your str (it print logging and then calls the orig-str)
17:00	Chousuke	manic12__: maybe struct maps do copy all of their values (because usually you don't have so many of them). but it's hopefully not going to be too slow
17:00	hiredman	ngoc: that is possible as well
17:00	Chousuke	manic12__: after all, every value is just a pointer. copying a hundred pointers is not too expensive :P
17:01	manic12__	it won't
17:01	manic12__	no
17:01	sproingie	i thought structs used a vector underneath
17:01	manic12__	in the clos version, it would have to actually go through and unbind dependent slots, I'm sure that's slower than just copying it
17:02	hiredman	I wonder if there is a point to using structmaps with 200 keys
17:02	sproingie	oh right struct vs structmap. confoozing
17:02	Chousuke	... creating instances of them would be painful :)
17:02	hiredman	structmaps are supposed to be a optimized case for small maps with the same keys
17:02	manic12__	yeah, after a point you would think it would have the offsets in a class object
17:02	sproingie	hiredman: i could see writing a virtual machine with a slot per bytecode
17:03	sproingie	but chances are it'd be a singleton
17:03	hiredman	sproingie: eh?
17:03	sproingie	you know a big table
17:03	manic12__	i'm using structmaps right now because that's what spinoza uses
17:03	sproingie	nevermind my babbling
17:03	hiredman	I wonder at what number of keys a struct map ceases to be an optimization
17:03	sproingie	manic12__: the author of spinoza told me not to waste my time with it
17:04	hiredman	spinoza :(
17:04	Chousuke	hiredman: probably never, as the keys are shared between structmap instances
17:04	manic12__	maybe so, but I don't plan on using it like a clos, it's a learning tool
17:05	sproingie	if it's just to learn, then stop optimizing up front
17:05	manic12__	real clos does all kinds of optimization tricks for fast generic function dispatch and accessor methods
17:05	manic12__	who's optimizing?
17:05	hiredman	multimethods!
17:05	manic12__	i implemented an undo
17:06	sproingie	you're fretting about copies of data structures that never actually get copied
17:06	ngoc	Hmm, I feel that binding forces the callers to explicitly list all the plugins they wants to use all the time, like (with-feature1 (with-feature2 (with-feature3 str))), this is ugly. How to plug all the plugins only once when the program starts?
17:06	sproingie	sounds like optimization concerns
17:06	tomoj	ngoc: the whole point is that you shouldn't do that :)
17:07	hiredman	ngoc: I would really recomend you spend more time learning clojure before you begin worrying about writing something that needs a plugin system
17:07	sproingie	multifns have the start of a nice generic function dispatch, but building on that is still in roll-your-own territory
17:07	Chousuke	ngoc: design a system that allows people to create new functions by combining the plugins with the base functions
17:07	manic12__	there is no declarative programming paradigm in clojure
17:07	sproingie	looking at watches, i would argue otherwise
17:07	Chousuke	ie. (def my-cool-function (plugin1 (plugin2 (plugin3 base-function)))))
17:07	sproingie	it's certainly not declarative like prolog, no
17:08	manic12__	i'm not saying declarative is the best, but it makes sense for my application
17:08	manic12__	mechanical engineers are visual people
17:08	manic12__	spatial
17:09	manic12__	(they're so spatial)
17:10	hiredman	there are a lot of primitives available that a declarative system can be built upon
17:10	manic12__	so far I feel like I can't describe anything geometric with clojure because it feels like some kind of strange scheme with an emphasis on immutability
17:10	hiredman	promises,delays,etc
17:10	manic12__	structmaps and vectors?
17:10	ngoc	hireman: I have been studying Clojure for one month, no big progress at all. Do you have any tips on studying Clojure?
17:11	hiredman	ngoc: write something small first, the sidebar on the clojure.org website has a list of features to try out
17:11	manic12__	as "dumb" as smalltalk and flavors, or even clos is, you can still describe types of objects and how they are contained
17:12	hiredman	manic12__: erm
17:12	sproingie	you do have to think functionally. clos with all its declarative syntax sugar still usually comes down to OO
17:12	manic12__	um, yeah
17:12	hiredman	we may have different understandings of "declarative programming"
17:12	manic12__	declarative and object oriented are not mutually exclusive
17:13	hiredman	what do you mean by declarative?
17:13	sproingie	you can get a lot of declarative behavior out of multifns
17:13	manic12__	car has chassis, chassis has wheels
17:14	sproingie	struct, works fine
17:14	manic12__	you can but you can get a lot of milage of of single object dispatch too!
17:14	ngoc	Chousuke: Do you mean I should study macro?
17:15	sproingie	multifns have multiple dispatch, they just don't have any kind of MRO for call-next-method to work
17:16	manic12__	sproingie: i am very familiar with multiple-object dispatch from clos
17:16	sproingie	multifns are insanely more powerful than just switching on class. you can destructure an object and switch on some arbitrary function of it
17:16	hiredman	manic12__: that seems to relate to the dictionary definition of Declarative in that you are declaring relationships, but I don't see how it relates to declarative programming
17:16	sproingie	that's probably why there's no MRO, how the heck do you determine what the "next" thing is from that
17:17	kotarak	ngoc: consider a simple program, eg. a logfile parser or something overlookable. And start implementing that.
17:17	kotarak	ngoc: get comfortable in Clojure before you start with macros...
17:17	manic12__	sproingie: that's great, get a bunch of gearheads doing it
17:17	sproingie	floating around in my head is the idea of dispatching on a sequence instead of a single value, but i don't know offhand what the code would look like
17:18	sproingie	manic12__: make a bunch of macros that look like autolisp, done
17:18	sproingie	not saying it's a good lisp, just saying that's what they know
17:18	manic12__	they also know icad
17:18	manic12__	which is built on flavors
17:19	hiredman	sproingie: you can dispatch on sequences, just works better with vectors
17:19	Chousuke	sproingie: the dispatch fn would just be (comp vec map your-dispatch-fn) :P
17:19	Chousuke	er, wait
17:19	hiredman	↑ yeah, use vectors
17:19	Chousuke	(partial map dfn)
17:19	sproingie	hiredman: yeah a vector would probably be more efficient, i'm thinking the sequence could be lazy though.
17:20	sproingie	given arguments foo, bar, baz, generate a MRO. now given foo, bar, baz, a MRO, and a current position, find the next method
17:20	hiredman	sproingie: the isa stuff is all built to work on vectors already
17:20	hiredman	(isa? [Object Object] [String Character])
17:20	hiredman	,(isa? [Object Object] [String Character])
17:20	clojurebot	false
17:20	hiredman	bah
17:20	manic12__	it's great that clojure has all of these great features with multifns & metadata, but if simple people have a hard time doing simple programming, well I'm sure you would say they need to re-learn
17:21	hiredman	,(isa? [String String] [Object Object])
17:21	clojurebot	true
17:21	sproingie	it ain't logo yet
17:21	manic12__	I'm trying to apply a simple object system paradigm from flavors and clos to clojure
17:21	sproingie	i'd rather the bike gets built before the training wheels
17:22	sproingie	clos is anything but simple
17:23	hiredman	(defmulti new (fn [& args] (vec (conj (map class (rest args)) (first args)))))
17:24	hiredman	(defmethod new [:box String Integer] [_ name size] {:name name :size size})
17:24	manic12__	yeah, well clos can be simplified with some meta programming and macros
17:24	hiredman	of course you couldn't call it new
17:25	hiredman	it would have to be my-new
17:25	manic12__	make-instance?
17:25	hiredman	and you'd want to map type not class
17:26	hiredman	and you'd want to add :box as the :tag metadata to the resulting {}
17:26	manic12__	you know why java is popular...it creates a framework where programmers don't have to think as hard
17:26	hiredman	that is a lie
17:26	sproingie	snort
17:26	manic12__	icad is the same way
17:27	hiredman	well, they don't think they have to think
17:27	hiredman	so they write crap
17:27	sproingie	i have to think all the time, long and hard, about how to get around the limitations in the language
17:27	manic12__	and i hit a limitation in clos, that's why I'm here
17:27	sproingie	usually i can find a nice library that helps out. which is great til i want to make it work with another library that solved another problem
17:28	manic12__	i would like to create a dead-simple geometric library for the icad style declarative/generative/object-oriented programming
17:29	sproingie	you may be the only one familiar with it. pastebin an example of what it looks like?
17:30	manic12__	sometimes i think computer scientists have lost touch with the problems the science was created to solve
17:31	kotarak	"Computers help us to solve problems we wouldn't have without them."
17:31	hiredman	~clojure
17:31	clojurebot	clojure is a very attractive hammer with a nice heft to it
17:33	manic12__	if you have used parametric cad to design physical objects such as machine parts, it shows that things have progressed, but at the same time the software designers lock out the user from being able to really write flexible applications
17:34	manic12__	that's why I became a programmer
17:35	sproingie	computer scientists have never particularly been in touch with the real world. that's not their job.
17:35	sproingie	engineers create stuff, scientists figure out new principles to create stuff on.
17:36	sproingie	theory, meet practice. practice, theo-- hey theory, stop wandering off, pay attention.
17:37	hiredman	manic12__: I think you might want to get a blog
17:39	manic12__	i don't think anyone would read my blog
17:41	hiredman	I dunno, you could grab the mic and inspire the masses to rise as one and throw off the shackles of the computer science intelligentsia
17:41	sproingie	you have nothing to lose but your boot times!
17:41	manic12__	i think i should be an arborist
17:42	drhodes	there's a race of men who live in the trees
17:43	manic12__	tarzan of the kbe trees here
17:45	manic12__	i am going to have to get this thing to do a deep copy
17:46	manic12__	if it finds a "pointer" to a struct map as a value of an entry, it needs to copy that whole thing too, not just the pointer
17:46	hiredman	manic12__: :(
17:46	hiredman	that is horrible
17:48	manic12__	the structmaps don't need to contain keys for every message, just the ones that are "changeable" (I hate these "")
17:50	hiredman	still sounds horrible
17:52	Chousuke	why does it need a deep copy?
17:53	Chousuke	I mean, maps nest just fine. :/
17:53	manic12__	alibre design is written i java, i should just get a license to that and run clojure as a scripting language and be done
17:54	manic12__	because if you don't deep copy, you when you undo or redo you will get state that belongs to something else
17:54	manic12__	it's a very thin skeleton
17:54	Chousuke	oh, so the pointer is a ref then?
17:55	manic12__	no, by "pointer" i mean (identical? val1 val2)
17:55	Chousuke	but that's not true
17:55	Chousuke	or, at least I think it isn't.
17:58	sproingie	not that i don't love clojure, but if you have a clos thing you want to port to java, perhaps you should consider ABCL?
17:58	sproingie	or one of the many scheme implementations where there's a tinyclos port
17:58	Chousuke	,(let [a {:a {:deep :map}} b {:b 1} ] [(assoc b :foo (:a a)) b a])
17:58	clojurebot	[{:foo {:deep :map}, :b 1} {:b 1} {:a {:deep :map}}]
18:00	Chousuke	the :foo value and :a value are identical, but there is no connection between the "parent" maps
18:02	Chousuke	manic12__: is that what you're worried about, or something else? :/
18:03	Chousuke	manic12__: the submaps can freely be the same object, because it's not like anything can change them :)
18:23	spuz	Hello, I've having a really straing problem accessing external files from clojure
18:23	spuz	first of all, I cannot load any files using load-file
18:25	hiredman	have you read the load-file docs?
18:25	spuz	also, I cannot seem to 'see' files on my hard disk. Running (.exists (java.io.File. "C:/pe5.clj")) returns false even though that file exists
18:26	spuz	hiredman: yes, I can use relative or absolute paths apparently, but neither seem to work
18:26	hiredman	have you tried "C:\pe5.clj"
18:26	spuz	yes
18:27	spuz	at least "C:\\pe5.clj"
18:27	manic12__	Chousuke: those sub objects may be immutable, but if they are the wrong ones then that is a problem
18:27	spuz	however "C:/Windows" returns true
18:27	spuz	it's very strange
18:28	hiredman	spuz: does the pe5.clj file exist?
18:28	spuz	hiredman: yes!
18:28	Chousuke	manic12__: how would they be wrong?
18:29	spuz	I've tried putting it in the current dir and using "pe5.clj" as well
18:29	hiredman	spuz: what does dir c:\pe5.clj say?
18:29	Chousuke	manic12__: or is there some situation where the deep copy actually "changes" some of the inner map? :/
18:29	spuz	hiredman: hmm file not found!
18:30	hiredman	so it doesn't exist
18:30	hiredman	problem solved
18:30	spuz	hiredman: uh well, not exactly, typing that path into explorer finds the file, but clearly it's not a clojure problem
18:32	spuz	hiredman: ah, how annoying, I have extensions turned off on this machine and the file was actually called "pe5.clj.txt"!
18:32	spuz	yes I'm a retard
18:34	sproingie	blame notepad
18:35	akhudek	While playing around building a small svg-like vector library using Java2d, I ran into an interesting problem dealing with mutable trees.
18:35	akhudek	I decided to represent a scene graph as nested vectors inside a ref.
18:35	hiredman	notepad++
18:36	akhudek	Since the scene graph is a shared mutable resource that theoretically multiple threads may modify.
18:36	hiredman	I actually did a few euler problems in notepad++
18:36	akhudek	Now, a common operation may be to frequently change one specific node in the graph.
18:36	akhudek	Say for a text label during an animation.
18:37	hiredman	have you seen zippers?
18:37	akhudek	yes, but you can't have multiple cusors
18:37	akhudek	A simple animation would likely change a label node in one place, and say another object somewhere else.
18:37	akhudek	Traversing the tree between both positions each frame seems ineffecient.
18:38	hiredman	(require '[clojure.zip :as zip])
18:38	hiredman	,(require '[clojure.zip :as zip])
18:38	clojurebot	nil
18:39	akhudek	The only decent solution seems to be to put frequently mutated nodes into references.
18:39	akhudek	And have the rendering code automatically deref any references it encounters.
18:40	hiredman	:(
18:40	hiredman	why do you want multiple cursors?
18:40	hiredman	how deeply nested are the vectors?
18:41	akhudek	So that you can refer to more than one object in the scene.
18:41	hiredman	eh?
18:41	hiredman	,(-> '[:a [:b :c]] zip/vector-zip zip/next zip/next zip/next (zip/replace :x) zip/next (zip/replace :y) zip/root)
18:41	clojurebot	[:a [:x :y]]
18:41	akhudek	Right, but in a graphics context where you modify more than one node each frame.
18:42	akhudek	You'd need to move between all nodes you need to change
18:42	hiredman	that modifies two nodes
18:43	akhudek	my complaint is that having to move between the two nodes may be a lot of work
18:43	sproingie	could collect a list of changes and sort it
18:43	sproingie	sort of like elevator queueing
18:43	hiredman	you could use something besides vectors
18:43	sproingie	that too
18:44	akhudek	such as?
18:44	hiredman	you could flatten nest vectors by wrapping them into a map
18:44	akhudek	and have nested maps?
18:44	sproingie	you normally have to sort scenegraph updates anyway
18:44	hiredman	{[0 1] :x [0 0] :a} => [[:a :x]]
18:44	hiredman	no, just one map
18:45	akhudek	hmm, maybe the problem is in my representation of the graph then
18:45	hiredman	{[0 1] :x [0 0] :a [1] :p} => [[:a :x] :p]
18:46	akhudek	I was translating svg elements into vectors like [:group :width 10 :height 10 [:rect :x 1 :y 1] [:rect :x 20 :y 20]]
18:46	akhudek	my idea was to render these during paint
18:47	akhudek	by traversing the tree and carrying the state as needed by the svg rendering model
18:47	hiredman	akhudek: have you looked at how clojure.xml/parse represents xml?
18:47	akhudek	I should probably do that.
18:47	akhudek	It doesn't use zippers?
18:48	hiredman	no, it uses a structmap, but you can traverse it via zippers
18:48	akhudek	ok, I'll take a look, thanks
18:49	hiredman	I am rather taken with {[0 1] :x [0 0] :a [1] :p} => [[:a :x] :p] myself
18:59	danlarkin	,(+ nil 1)
18:59	clojurebot	java.lang.NullPointerException
18:59	danlarkin	suggestions other than try/catch?
18:59	hiredman	,(or nil 0)
18:59	clojurebot	0
19:00	hiredman	,(-?> nil (partial + 1))
19:00	clojurebot	nil
19:00	hiredman	,(when nil (+ nil 1))
19:00	clojurebot	nil
19:01	akhudek	ah, my current code is actually nearly the same as clojure.xml in terms of representation
19:01	danlarkin	I think the first suggestion is best so far
19:01	akhudek	I simplified to nested vectors here, but really I use structs where a content field contains a vector of structs
19:02	akhudek	it's great for reading the data, but not so great for rapidly changing several deeply nested nodes
19:03	akhudek	hiredman: you're representation is interseting
19:04	hiredman	"I find your representation intriguing, and I would like to subscribe to your newsletter"
19:05	Chousuke	danlarkin: the maybe monad :P
19:05	danlarkin	Chousuke: oh jeez!
19:06	Chousuke	hey, monads are sometimes actually useful :)
19:06	Chousuke	not just for pretending IO is mathematically pure.
19:19	akhudek	hmm, your rep. with using a sorted map with an appropriate sort function defining dfs order would work. I'll work with that for now, thanks. :)
20:52	bpattison	is there anyway in clojure to create/define a hash value through a macro? something like (defmacro m [a] `(func :~a)) ?
20:55	arohner	bpattison: can you give an example of how you want it to work?
20:56	arohner	macros transform s-exprs, so it's much easier to figure out what you want if you know the start and end points
20:56	arohner	:-)
20:57	bpattison	here's the macro I'm trying to get working
20:57	bpattison	(defmacro def-keyword [word]
20:57	bpattison	`(def k-~word
20:57	bpattison	(constant-semantics
20:57	bpattison	(lit-conc-seq "~word" nb-char-lit)
20:57	bpattison	(make-node-keyword :~word))))
20:57	bpattison	sorry about all the newlines
20:58	Makoryu	bpattison: http://gist.github.com
20:58	Makoryu	:p
20:59	arohner	right, but you haven't explained where you're starting and where you're ending up, and I have to guess at what this code is supposed to do
20:59	arohner	give me an example of working code without the macro, and what you want the result of the macro to be
20:59	bpattison	http://gist.github.com/198618
21:01	bpattison	okay, I want it to create a def named "k-~word" that makes use of a hash named ":~word" so the macro would produce "k-new" with hash value ":new" when the macro is invoked with word=new
21:02	bpattison	and so :new would used in other parts of the code
21:03	arohner	so calling (m {:word "new"}) would call (def k-new) ?
21:04	arohner	or (m {:word :new}) would call (def k-word :new) ?
21:06	bpattison	hmm, its more like creating a short-cut for a larger expression and naming it k-new where that larger expressions uses a hash key of :new as well
21:08	bpattison	(def-keyword new) would produce (def k-new .... :new ...)
21:10	bpattison	but I get a syntax error on :~word in the macro
21:19	arohner	:word is a literal
21:20	arohner	if you want the keyword-ized version of an argument, use (keyword word)
21:20	arohner	where word is a symbol or a string
21:20	arohner	sorry, only a string
21:20	arohner	i.e.
21:20	arohner	,(keyword "foo")
21:20	clojurebot	:foo
21:20	hiredman	,(keyword (name 'foo))
21:20	clojurebot	:foo
21:28	bpattison	arohner: excellent! thanks
21:46	Makoryu	, (+ 1 "food")
21:46	clojurebot	java.lang.ClassCastException: java.lang.String cannot be cast to java.lang.Number
22:01	kevin__	, (+ 1 2 3) ; curious
22:01	clojurebot	6
22:02	kevin__	nice, i can write a distributed computing app
22:23	hiredman	~suddenly
22:23	clojurebot	CLABANGO!