#clojure – 2009-06-09

2009-06-09

00:51	replaca	wow, I just hacked slime to use pretty print and it's really nice :-)
00:55	replaca	but not very fast -> more motivation to improve performance!
01:06	mrsolo	pp feels slow
01:10	replaca	mrsolo: it is pretty slow -> about 150 lines/sec on my machine
01:10	replaca	i'm working on making it faster
01:11	mrsolo	nice.. i use it a lot :-)
03:35	cgrand	~seen quidnunc
03:35	clojurebot	quidnunc was last seen quiting IRC, 1959 minutes ago
03:40	frodef	I'm learning clojure and trying to figure out the zip/zipper function, but can't seem to get it to work as expected ..
03:41	hiredman	what are you trying to do?
03:41	frodef	If I apply zipper to e.g. (zip/seq-zip [:a :b [:c :d] :e]), it seems to me that the branch? function isn't ever called.
03:42	hiredman	well
03:42	hiredman	that is a vector
03:42	hiredman	and you are calling seq-zip
03:42	hiredman	,(doc clojure.zip/seq-zip)
03:42	clojurebot	"([root]); Returns a zipper for nested sequences, given a root sequence"
03:42	frodef	ah, so a vector isn't a squence then..?
03:42	hiredman	,(doc clojure.zip/vec-zip)
03:42	clojurebot	"/;nil; "
03:43	hiredman	frodef: nope
03:43	frodef	thanks
03:43	hiredman	you can turn a vector into a seq easily, and a lot of functions do it implicitly
03:45	frodef	hiredman: maybe even seq-zip? I seem to get the exact same behavior with vector-zip.
03:45	hiredman	frodef: how exactly are you calling it?
03:46	hiredman	,(-> [:a [:b :c] :d] clojure.zip/vec-zip)
03:46	clojurebot	java.lang.Exception: No such var: clojure.zip/vec-zip
03:46	hiredman	,(-> [:a [:b :c] :d] clojure.zipper/vec-zip)
03:46	clojurebot	java.lang.ClassNotFoundException: clojure.zipper
03:46	frodef	(zip/zipper (fn [x] (println "hello world") true) ... ... (zip/vector-zip [:a :b [:c :d] :e]))
03:46	frodef	vector-zip, no?
03:47	hiredman	no
03:47	hiredman	maybe
03:47	hiredman	yeah, vector-zip
03:47	hiredman	but you don't call zip/zipper
03:47	frodef	naming could be more consistent.. :)
03:48	hiredman	zipper is if you want to make your own zipper creating function
03:48	hiredman	vector-zip is pre-made
03:48	hiredman	,(-> [:a [:b :c] :d] clojure.zip/vector-zip)
03:48	clojurebot	[[:a [:b :c] :d] nil]
03:48	hiredman	,(require '[clojure.zip :as zip])
03:48	clojurebot	nil
03:49	frodef	hiredman: I'm just using vector-zip as an easy way to get a zip object to pass to zip/zipper.
03:49	hiredman	,(-> [:a [:b :c] :d] zip/vector-zip zip/next zip/next)
03:49	clojurebot	[[:b :c] {:l [:a], :pnodes [[:a [:b :c] :d]], :ppath nil, :r (:d)}]
03:49	hiredman	frodef: you don't pass zip objects to zip/zipper
03:49	hiredman	vector-zip returns a zipper
03:49	hiredman	,(-> [:a [:b :c] :d] zip/vector-zip zip/next zip/next zip/node)
03:49	clojurebot	[:b :c]
03:50	frodef	ah, of course..
03:50	frodef	I somehow had the idea that zipper was for mapping one zip to another.
03:51	frodef	hiredman: thank you for helping.
03:51	hiredman	a zipper is a nifty method for tree traversal and editing
03:53	hiredman	,(pl (↕map range $ 3 inc)))
03:53	clojurebot	(1 2 3)
03:53	hiredman	pl uses a zipper to edit s-expressions
03:53	hiredman	,(macroexpand '(pl (↕map range $ 3 inc))))
03:53	clojurebot	(do ((flip map) (range 3) inc))
09:57	rhickey	anyone thinking about how to use GWT with Clojure?
09:58	Chouser	I tried to run src/jvm through it once. That didn't turn out so well.
09:59	Chouser	Since it consumes java sources when producing javascript, I've pretty much just written it off.
09:59	rhickey	c-in-c -> java source -> gwt ...?
10:00	Chouser	seems likely.
10:00	asbjxrn	c-in-c ?
10:01	rhickey	clj-in-clj
10:01	Chouser	clojure's current src/jvm uses a bunch of classes that are off-limits for gwt, but that sort of thing ought to be easier to fix via c-in-c than manually fixing up hand-written .java.
10:01	rhickey	clinj?
10:02	Chouser	as in "that clinj's it!"?
10:03	Chouser	asbjxrn: when most of the clojure compiler is written in clojure (instead of in java as it is now), it'll be easier to add new target languages.
10:03	asbjxrn	Got it.
10:04	Chouser	I suppose you ought to be able to try gwt's "hosted mode" from clojure, but I certainly haven't done that.
10:05	rhickey	ick - how will we generate this:
10:05	rhickey	private native void setPayload(String val) /*-{
10:05	rhickey	this.payload = val;
10:05	rhickey	}-*/;
10:06	Chouser	hehe. that's not so much "c-in-c -> java source -> gwt" as it is just "c-in-c -> gwt source"
10:07	Chouser	it'd be so much easier if I just loved writing java code.
10:08	rhickey	right - clinj->gwt would be bypassing the advantages they wanted to get in GWT by using Java - refactoring, type checking etc
10:10	Chouser	oh, I didn't mean trying to generate the gwt .js directly -- that stuff is scary. I just meant the gwt-java target might not have all that much in common with the regular java target.
10:11	Chouser	I don't know if you'd get much benefit out of refactoring support, but that'd give you hosted mode with type checking.
10:11	aravind	rhickey: I looked at it, gwt, but was scared away by all the java in there. my quest for the perfect web framework is still unfulfilled. I started looking for ways to port weblocks into clojure, but that depends on so many cl libraries, it seemed like a daunting task.
10:11	Chouser	still not good debugging support, at least not all the way back to the .clj code.
10:12	rhickey	Chouser: I understand, just meant we only get some of the GWT benefits, but there is lots of desirable tech in there
10:13	aravind	it just upsets me that anyone should have to write any code to generate boiler plate html. GWT seems like its a subset of the ideas in weblocks from what I understood.
10:13	Chouser	yeah, the deployed apps look awfully nice, and I trust Google's support of all the browsers.
10:14	Chouser	pre-clojure I was completely uninterested in compiling anything to javascript because I just didn't see anything (java included) as being enough better than js to be worth the trouble (debugging generated code, etc.)
10:15	rhickey	aravind: GWT does a huge amount of very practical and important things, abstracting away JS, CSS, browser differences, testability, debugging, communications, minimization, image spriting, browser modularization etc etc
10:16	rhickey	It was nice to see they actually used it themselves for the wave client
10:18	Chouser	yeah, I was waiting a bit for that -- they weren't using for gmail or their other "office" apps, as far as I could tell.
10:19	aravind	rhickey: yeah, the base of it seems to be that you operate on widgets and provide callbacks that act on these widgets. I feel that even tha is not going far enough. Why should I have to provide those callbacks. Afterall the whole point of a webapp is to display and manipulate data. CL's weblocks takes that widget idea to its fulfillment. It is based on widgets as well, but you define those widgets in CL, and it depends on having getter setter methods
10:21	aravind	I guess, I am just mad that I can't have weblocks in clojure :(
10:21	rhickey	aravind: widgets are the smallest part of the problem, IMO
10:22	aravind	I probably don't see the whole picture then.
10:23	asbjxrn	When writing a wrapper for a java api, would you use the enums defined in the API, or translate to keywords? I find keywords feel more clojury myself.
10:25	asbjxrn	While performance is not an issue in my current project, for things like the opengl wrappers it could be. Would a keyword lookup be measurable at all?
10:27	Chouser	asbjxrn: if you have to look up a keyword in a map to get the enum value at runtime, that'll cost something. Maybe not enough to matter, but something.
10:27	Chouser	I sometimes use macros to do that kind of lookup at compile time instead.
10:30	asbjxrn	Yea, in my current case I have the enums in a hashmap, with the enums as they key and the keyword as the value.
10:31	asbjxrn	So obviously a hash lookup would be made.
10:32	triddell	rhickey: With GWT, I've used clojure/compojure to focus on the server-side piece initially (which I know may be obvious and wasn't the topic of your earlier discussion on GWT.) Writing basic JSON/REST services with compojure and persisting data in the app engine data store took an amazingly small amount of completely generic clojure code though.
10:32	Chouser	I've got a macro in one project that does a lookup at compile time and throws an exception if nothing is found. This helps catch typos early.
10:33	rhickey	triddell: but you used GWT for the front-end?
10:35	triddell	rhickey: yes... and it was just a proof-of-concept in this case. I mostly was interested in the clojure server-side.
10:36	asbjxrn	chouser, That would be for use within the library itself? In my case I'm getting midi events and translating things like ShortMessage/CONTINUE to :continue so the user of the wrapper would see that as the midi event type. Not sure if a macro is relevant there?
10:37	Chouser	asbjxrn: ah, sure -- that's got to be done at runtime. I was talking about passing in keywords.
10:37	Chouser	(.fictionalMethod obj (flags :foo :bar))
10:48	Chouser	rhickey: is there any reason a transaction couldn't be shared by multiple threads?
10:58	rhickey	Chouser: when would it end?
10:59	Chouser	when told to, I suppose.
11:00	rhickey	it would be a big change
11:00	Chouser	Oh, how could it possibly be retried!? nm.
11:01	rhickey	I think it is cleanest to have the transactional thread do all ref interop, but you can spawn threads for calculation
11:01	Chouser	I was thinking of passing around a snapshot view of the ref universe.
11:01	Drakeson`	,1010b
11:01	clojurebot	Invalid number: 1010b
11:01	Chouser	,2r1010
11:01	clojurebot	10
11:01	rhickey	Chouser: the beauty of refs is that there are multiple simultaneous universes
11:02	Drakeson`	thanks
11:31	ctdean	What are people using for simple distributed programming? Wondering if there is a consensus.
11:32	stuartsierra	"simple distributed programming" is an oxymoron
11:33	ctdean	ok, lol. How about small scale distributed systems. on the order of 10 nodes
11:33	ctdean	I can always roll my own with http, thrift, etc. That's what I've done in the past
11:33	stuartsierra	ctdean: I use Hadoop, which is very good but only suitable for certain tasks.
11:35	drewr	stuartsierra: Curious, do you use straight HDFS or an abstraction on top of it?
11:36	stuartsierra	HDFS
11:36	drewr	I've been experimenting and I'm about to load a butt-ton of data into it.
11:37	drewr	I think hbase is too immature for me.
11:37	stuhood	i'd definitely recommend trying pig/hive/cascading first, before you work directly with the hadoop api
11:38	drewr	I've wrapped the bare api with clojure and it's still pretty nasty.
11:46	Drakeson`	what does this mean: No matching method found: toString for class java.lang.Integer
11:46	duck1123	is it possible to define a method for a multi-method from within a different namspace?
11:46	Drakeson`	trying to: (.toString 9 2)
11:47	drewr	Drakeson`: It means there is no method named toString on Integer that takes an Integer as an argument.
11:47	duck1123	i think it's because that's like calling 9.toString(2)
11:47	drewr	,(.toString 9)
11:47	clojurebot	"9"
11:47	ctdean	,(str 9)
11:47	clojurebot	"9"
11:48	drewr	,(str 9 2)
11:48	clojurebot	"92"
11:48	Drakeson`	is 9.toString(2) different than toString(9,2) ?
11:48	drewr	What are you trying to accomplish?
11:48	drewr	(Yes, to answer your q.)
11:49	duck1123	,(map #(.toString %) '(9 2))
11:49	clojurebot	("9" "2")
11:49	Drakeson`	How does one call http://java.sun.com/j2se/1.4.2/docs/api/java/lang/Integer.html#toString%28int,%20int%29 in clojure?
11:50	replaca	,(Integer/toString 9 2)
11:50	clojurebot	"1001"
11:50	drewr	Ah, static method.
11:50	replaca	yup
11:51	drewr	Couldn't figure out where he was going with that.
11:52	ctdean	I was just reading about the clojure meetup after javaone. The author mentioned Swarmli. anyone know what that is?
11:52	drewr	ctdean: http://olabini.com/blog/2009/06/the-clojure-meetup-and-general-geekiness/
11:53	Drakeson`	replaca: thanks.
11:54	Drakeson`	is there any circumstance that someMethod(x,y) is equivalent to x.someMethod(y), in java? (like python)
11:54	stuartsierra	no
11:55	drewr	ctdean: It hasn't shown up on IRC before, FWICT. http://www.google.com/cse?cx=004621955734675372103%3A0oxuogeollc&ie=UTF-8&q=swarmli&sa=Search
11:55	duck1123	it would be possible, but the author would have too write it that way
11:55	Drakeson`	stuartsierra: thanks.
11:56	ctdean	drewr: yep, thanks. just wondering about it
11:59	integrity	is there any clojure goodness wrapping java reflection api for dealing with java classes?
11:59	clojurebot	clojure is like life: you make trade-offs
12:02	Chouser	integrity: http://paste.lisp.org/display/67182
12:02	Chouser	something like that?
12:07	replaca	ctdean: the right name is Swarmiji
12:08	ctdean	replaca: aha! thanks
12:08	replaca	ctdean: it's a project being done by Amit Rathore and Tim Dysinger and Technomancy (Phil Hegelberger (sp?)) are working with him to make it open source
12:09	replaca	ctdean: it might be what you're looking for - very easy to distribute compute jobs through function-like interface, monitor them and retrieve results
12:09	replaca	no implicit data model (like map-reduce)
12:10	drewr	I remember that now from Amit's blog.
12:12	drewr	http://www.google.com/cse?cx=004621955734675372103%3A0oxuogeollc&ie=UTF-8&q=swarmiji&sa=Search
12:12	stuartsierra	For my presentation tonight: where would you put Vars in the mutable-types diagram? (coordinated/independent and asynchronous/synchronous)
12:13	stuartsierra	This one: http://tinyurl.com/mllzc4
12:14	rhickey	stuartsierra: var bindings are not shared, you'd need another dimension for that
12:14	stuartsierra	that's what I thought
12:15	rhickey	synchronous/unshared, implying no coordination needed
12:15	stuartsierra	ok, thanks
12:15	stuartsierra	cool
12:20	duck1123	ok, I think I'm missing something. How do I test the type of a struct-map. I thought it would be isa?, but I can't seem to get it to work.
12:20	stuartsierra	duck1123: struct-maps aren't types. You can add a key or metadata to indicate a type.
12:22	duck1123	stuartsierra: that sucks. I was hoping that info would've been stored somewhere
12:57	cemerick	oh, my -- slurp doesn't accept an encoding, and uses a FileReader (!!!!!)
12:58	stuartsierra	c.c.duck-streams/slurp* uses a default-encoding
12:58	cemerick	using default encoding ==> fail
12:58	stuartsierra	I mean default-encoding is a var that you can rebind.
12:59	cemerick	right -- but having slurp sit there as it is is really bad.
12:59	Chouser	what encoding is worth mentioning besides utf-8?
13:00	stuartsierra	A ton of stuff is still ISO-8859-*
13:00	mrsolo	yes my copy of clojure book is here
13:00	mrsolo	man it smells bad
13:00	cemerick	besides that....I'll introduce you to a couple of my Japanese friends and the old cobol system integrators that need to work with ebcdic
13:05	technomancy	(doc as-str)
13:05	clojurebot	"clojure.contrib.java-utils/as-str;[[x]]; Returns the name or string representation of x"
13:15	rhickey	Did anyone see the LWN article on Clojure?
13:16	stuartsierra	no!
13:16	rhickey	http://lwn.net/Articles/335966/
13:16	rhickey	I'm not a subscriber
13:16	stuartsierra	Bah, me neither.
13:17	Chouser	two more days
13:18	Chouser	"Vala supports modern language features as the following... Foreach"
13:20	Chousuke	:P
13:20	Chousuke	http://lwn.net/SubscriberLink/335966/6e9d1e30bb6a0b16/ here
13:21	Chousuke	LWN's subscriberlink feature is really neat.
13:21	Hun	yeah, it gives you an actually working link.
13:21	Chousuke	They just trust their subscribers not to rip off their content.
13:21	stuartsierra	nice
13:23	Chousuke	no artificial DRM or anything. :)
13:23	Hun	if i can read it, i can rip it off.
13:23	Hun	so drm would be useless
13:24	Chousuke	yeah
13:25	Chousuke	They actually allow you to post subscriber links on mailing lists etc. as well. They only ask you not to do that with every article. :)
13:27	cemerick	rhickey: hope you don't mind the random bug report on slurp. Encoding issues are just near and dear to my heart.
13:32	mrsolo	hmm httpagent barfs when i tried to compile
13:32	stuartsierra	mrsolo: it"s not finished yet
13:33	mrsolo	ok
13:36	stuartsierra	not sure why it's doing that, tho
13:37	stuartsierra	oh, found it
13:38	stuartsierra	fixed, thanks
13:38	mrsolo	nice
13:39	technomancy	usually when people have unfinished libs they don't put them straight into contrib. =)
13:39	stuartsierra	Eh, it works (now), wanted to let people see it.
13:40	technomancy	yeah, it just contributes to the "contrib is just a bucket of code; anything goes" notion.
13:40	mrsolo	hmm
13:41	stuartsierra	Yeah, I kind of like that notion.
13:41	mrsolo	at least compiles i hope...
13:41	duncanm	any SLIME users around?
13:41	stuartsierra	yes
13:41	technomancy	duncanm: sure
13:41	stuartsierra	technomancy: I suppose we could branch contrib, make a formal release.
13:42	mrsolo	i like that noting those to be of more cough quality release
13:42	duncanm	technomancy: i'm using an app that has a custom Clojure REPL built-in to it - i'd like to be able to connect to that app using Swank, are there any guides out there for doing that?
13:42	technomancy	stuartsierra: I think that's a good idea, provided it's a quality release and works with clojure 1.0.
13:43	technomancy	duncanm: you should be able to load swank and start a listener from within the app, but I don't know of any docs explaining how to do that.
13:43	duck1123	contrib probably won't get good releases until it's broken up. (like apache commons)
13:44	stuartsierra	Well, still needs some clean-up, maybe leave out a few less-developed libs.
13:44	duck1123	duncanm: I run a swank server from within my compojure app
13:44	duck1123	I've posted instructions on how to get it running to the compojure ml
13:46	duncanm	duck1123: is there a URL?
13:46	duck1123	one sec
13:46	stuartsierra	technomancy: I don't have time to lead the charge right now; maybe in a few weeks.
13:47	technomancy	stuartsierra: any specific area you'd like help with?
13:48	stuartsierra	General organization, naming, apis.
13:49	duck1123	duncanm: It's not thee greatest, an I think I should be using an agent to do it right, but since it was only supposed to protect against re-evaluation of my source, I wasn't too concerned.
13:49	duck1123	duncanm: http://groups.google.com/group/compojure/browse_thread/thread/a4708c9bdb780cbc/2f63267744aa8731
13:57	replaca	stuartsierra, technomancy: also maybe new code should be developed outside contrib (say on github or whereever) and added to contrib once it's reached a certain level of maturity. That way, folks can see it and comment participate while it's under development, but they can count on the things in contrib to some extent
13:57	technomancy	replaca: I quite agree.
13:57	stuartsierra	replaca: I'd rather just separate contrib into "stable" and "sandbox" areas.
13:57	replaca	stuartsierra: why?
13:58	stuartsierra	One less thing to keep track of.
13:58	replaca	stuartsierra: shouldn't you have your own sandbox for stuff under development?
13:58	duck1123	or a clojure.extras, where any code that has a dependency other than clojure's dependencies would go
13:58	stuartsierra	why? Lots of open-source repos have a sandbox.
13:59	stuartsierra	Back in the day, before github, nobody had their own repository.
13:59	technomancy	I don't understand why the sandbox should be centralized. Seems like a throwback to ye olde days.
13:59	replaca	stuartsierra: yeah, that was back in the day
13:59	Chousuke	technomancy: it'd get a lot more attention that way
14:00	stuartsierra	Basically, I've already got a github acct. that I use for work.
14:00	clojurebot	github is git://github.com/kevinoneill/clojure.git
14:00	duck1123	"when I was your age, we had to connect to the server just to see what we changed."
14:00	technomancy	Chousuke: once it achieves some level of stability/quality and needs attention, it can be promoted.
14:00	technomancy	my random experiments have little if anything to do with yours
14:01	Chousuke	technomancy: the sandbox could be a good place for new additions to contrib that haven't seen much real use yet.
14:01	stuartsierra	yes
14:01	replaca	I'm with technomancy on this one. contrib is (currently) at least a very visible part of the system
14:01	Chousuke	then eventually things would migrate to the actual contrib namespace
14:02	technomancy	look, a sandbox => http://github.com/languages/Clojure/updated
14:02	technomancy	=)
14:02	Chousuke	the thing is, when you import a clojure.contrib.sandbox.foo, you will know it's somewhat experimental, but that it will appear in contrib someday
14:03	Chousuke	Like this, the sandbox wouldn't be free-for-all. That's what github is for :)
14:03	replaca	I wonder if we should decide what will appear in contrib while it's still experimental
14:03	technomancy	Chousuke: I agree it's better than keeping experimental libs in contrib itself, no question there.
14:04	replaca	Chousuke: it would only be a free for all for committers!
14:04	technomancy	yeah, that's a big issue; it creates stratification between those blessed with commit rights and the rest of the rabble.
14:04	replaca	but I can't object to that too much as long as it's clearly separated
14:04	Chousuke	Maybe putting your lib into .sandbox could be taken as a "promise" that you will make it into something workable
14:04	replaca	technomancy: again, I'm with you
14:05	replaca	Chousuke: yeah, but why?
14:05	replaca	there are plenty of ways to share code that's "not ready for prime time"
14:05	duck1123	the bottom line is contrib should always at the very least compile
14:06	replaca	duck1123: no, stuff should work
14:06	Chousuke	replaca: that's not what .sandbox would be for. It'd be more like "I think this is fine, but it needs more usage testing"
14:06	stuartsierra	yes
14:06	replaca	duck1123: otherwise newbies (or just regular users) get frustrated
14:06	mrsolo	the entire contrib is pretty small right now i don't mind to take all of it
14:06	mrsolo	not sure this will stay this way
14:06	Chousuke	replaca: and .sanbox would be a convenient "incoming" box for contrib; people would see new libraries appear in it and possibly test them.
14:06	replaca	Chousuke: that's really different than "experimental"
14:07	mrsolo	also yes for us newbies there is no quality measurent of 'oh this is dead..not to use this etc'
14:07	replaca	or "just beginning development"
14:07	mrsolo	but i don't like committe either.. will get politicial sooner or later but hey it should at least compile :-)
14:07	Chousuke	replaca: if you're just starting with the lib, then I think it should stay outside contrib completely
14:09	replaca	I agree with both Chousuke and technomancy
14:09	replaca	I'm not opposed to a sandbox for things that are "almost ready"
14:09	Chousuke	regarding the committer issue, I don't think there's a way around that.
14:10	Chousuke	since you can't allow just anyone to put stuff in contrib. people need at least a CA :/
14:10	mrsolo	ability to comment toward libray on wiki page will be nice
14:10	duncanm	duck1123: and which files/jars do I need to have in the classpath?
14:10	replaca	Chousuke: agreed but that doesn't mean that committers should treat contrib completely as their private source code control
14:11	Chousuke	true.
14:11	Chouser	last time I asked about where to put work in progress, everyone told me contrib was fine -- gotta put it somewhere...
14:11	replaca	Chousuke: so committers should treat their "status" with some delicacy
14:11	Chousuke	SVN makes this more of an issue than it should be ;P
14:12	replaca	Chousuke: well, with git we could have "private" branches to play with
14:12	replaca	that would be nice just for evolving stuff that's alredy in
14:12	duck1123	duncanm: you need to have the swank-clojure lib on your classpath
14:12	replaca	*already
14:12	Chousuke	with any DVCS everyone has their own repo, so "committers" wouldn't be tempted to just check in their work in the contrib repo when they think it's ready.
14:13	replaca	(sorry, didn't mean to be git-specific :-))
14:13	Chousuke	I personally don't mind, I'm a fan of git. :)
14:13	Chousuke	but there are some mercurial-people around I think :P
14:13	replaca	in any case, gotta run. see you all later
14:38	cemerick	duncanm: FWIW, we use enclojure's REPL library extensively, and it works very well. I know you're using slime, but I thought I'd toss that out there.
14:46	duncanm	cemerick: i'm using fiji, and fiji has its own repl implementation
14:46	duncanm	cemerick: so are you saying enclojure is yet another implementation of a REPL?
14:46	duncanm	cemerick: i want to use SLIME because i'm typing everything else in emacs ;-P
14:47	Hun	you could also use comint-run with a wrapper script if connecting to slime is too hard
14:47	Hun	though comint is far less cool
14:48	duncanm	so i first need to build a jar file for swank
14:48	duncanm	i don't see a build script in the github page
14:48	technomancy	you can just jar up the swank-clojure dir; it doesn't need AOT
14:48	cemerick	duncanm: enclojure adds clojure support to NetBeans, and yes, it provides its own REPL impl.
14:49	cemerick	I just thought I'd mention it in case you weren't tied to emacs :-)
14:53	duck1123	duncam: I have a fork of swank that uses maven to produce a jar
14:54	duck1123	I haven't pushed it out yet, but I could if you're interested
15:05	duck1123	I'm experiencing an issue where I've modified a library (clj-record) that uses a macro to define functions in the current namespace. I'm trying to modify one of the defn's so it wraps the return value with metadata. If I start it up, I don't get any metadata, but when I re-eval the macro call, I get the metadata attached. Any idea what's going on?
15:07	duncanm	technomancy: to jar up swank-clojure, i should just take the swank directory as the base, right?
15:10	duck1123	duncanm: yes
15:12	technomancy	duncanm: no, I think you need the project root as the base
15:13	technomancy	since swank/core.clj provides the namespace swank.core
15:16	duncanm	duck1123: where is IGNORE-PROTOCOL-VERSION from?
15:17	duck1123_	swank.swank
15:17	duncanm	hmm
15:18	duncanm	how do i view the classpath again?
15:18	duck1123_	http://gist.github.com/126735
15:20	technomancy	,(System/getProperty "java.class.path")
15:20	clojurebot	java.security.AccessControlException: access denied (java.util.PropertyPermission java.class.path read)
15:20	technomancy	oh snap!
15:21	duncanm	sigh, it returns a string?
15:21	technomancy	there's something in contrib for getting it more sanely
15:22	duncanm	i don't think i have access to contrib with this REPL
15:22	duncanm	hmm, so swank is in the classpath
15:22	duncanm	(.contains cp "swank") => true
15:23	duncanm	so now i need to load it in?
15:23	duncanm	do i use LOAD?
15:24	stuartsierra	duncanm: maybe require
15:24	duncanm	aha
15:24	duck1123_	duncanm: if you use the gist I just posted, you can (net.mycyclopedia.swank/init)
15:26	duncanm	Caused by: java.lang.Exception: Unable to resolve symbol: with-bindings in this context
15:26	duncanm	but i did (require '(clojure.main))
15:27	duck1123	then you would have to use clojure.main/with-bindings
15:28	duncanm	wait, so i need to fully-qualified all the names?
15:28	duncanm	aha
15:28	duncanm	now my problem is, on Windows, there's no /dev/null
15:28	duck1123	clojure.main isn't automatically refered, that's why I use it the way I do
15:29	duncanm	i tried that NS form, but it didn't work
15:31	duncanm	aha
15:31	duck1123	I suppose you could replace "/dev/null" with the location of a file. (nil might work, but I haven't tried that)
15:33	duncanm	duck1123: according to Google, "NUL" works
15:33	duncanm	so nice, i got it working now
15:33	duncanm	the only nitpick is that my slime versions are mismatched
15:33	stuartsierra	Apache Commons IO defines a NullWriter
15:34	duck1123	I have that same problem, I haven't figured out how to get around that one yet
15:52	duck1123	wrt my macro question, I cleaned then re-built and now it's working
16:02	duncanm	this is very nice, having the REPL to my app be available inside emacs
16:02	duncanm	duck1123: do you have any code for tearing down the swank connection?
16:03	duck1123	duncanm: I just kill " cl-connection"
16:04	duck1123	any other way will kill your app
16:04	dnolen	slime-kill-repl works as well
16:05	duncanm	duck1123: i mean, doing it from the app-side, in Java, say
16:05	duncanm	so right now, i made it pop up a dialog, but it's not actually tied to anything
16:06	duck1123	duncanm: no, I couldn't find anything to do that.
16:32	cemerick	hrm, can't use #^chars to type hint character arrays...though #^"[C" does work.
16:35	stuartsierra	Yes, chars isn't a type
16:35	opqdonut	gah, i didn't know about the #^" type of type hints
16:36	hiredman	stuartsierra: but most primitives have something like "bytes" you can use to hint an array
16:36	Chouser	I keep trying to forget, and generally succeeding.
16:36	opqdonut	hiredman: actually, there is no bytes
16:36	opqdonut	at least in 1.0
16:36	cemerick	yeah, #^chars really should be included, as well as #^bytes
16:36	hiredman	I was pretty sure there is a bytes
16:36	cemerick	I think these nicer array hints were added alongside the unchecked math stuff, so those omissions make sense.
16:36	mrsolo	memoize rocks :-)
16:38	hiredman	,#^bytes (into-array Byte/TYPE (map byte (range 10)))
16:38	clojurebot	#<byte[] [B@14b1dbf>
16:38	Chouser	chars and bytes both work for me in 1.0
16:38	Chouser	,(expression-info '(let [x #^bytes (make-array Byte/TYPE 0)] x))
16:38	clojurebot	{:class [B, :primitive? false}
16:38	Chouser	,(expression-info '(let [x #^chars (make-array Byte/TYPE 0)] x))
16:38	clojurebot	{:class [C, :primitive? false}
16:39	hiredman	~def c.l.Compiler
16:39	Chouser	cemerick: what makes you think #^chars isn't working?
16:39	hiredman	http://code.google.com/p/clojure/source/browse/trunk/src/jvm/clojure/lang/Compiler.java?r=1382#880
16:40	cemerick	Chouser: I was trying to eliminate reflection on a particular call, and it didn't immediately work with #^chars, but eventually I got it to work with the string variant. I'll double-check.
16:47	jonafan	what do you guys think about grand central dispatch
16:49	hiredman	jonafan: I think it is mac only
16:50	jonafan	not really, it's just a thread pool with closure support in objective c
16:51	hiredman	I think SEDA(http://www.eecs.harvard.edu/~mdw/proj/seda/) is similar
16:51	technomancy	in practice nobody uses objective C on free platforms though.
16:52	hiredman	and SEDA is old news, there was a usenix paper in '03 comparing seda queue&event architecture to using threads
16:53	jonafan	okay but it could be easily implemented in any language
16:53	hiredman	(using threads directly)
16:53	technomancy	if someone went back in time to assassinate bjarne stroustrup, things could be different, but given the current state of things I find it difficult to care.
16:54	hiredman	http://www.usenix.org/events/hotos03/tech/full_papers/vonbehren/vonbehren_html/index.html
16:54	hiredman	"Why Events Are A Bad Idea (for high-concurrency servers)"
16:55	jonafan	so, not impressed?
16:56	Chouser	If I had to write high-level apps in Objective C, I imagine I'd be very excited about it.
16:56	Chouser	Or even in C++, a similar solution would be better than none.
16:57	Chouser	But Java by itself already gives you most of those features, doesn't it, if only at a per-process level: blocking queues, thread pools, etc.
16:58	jonafan	so you're saying that apple is playing catch up as far as these features are concerned
16:58	dnolen	jonafan: I do some Objective-C/Cocoa programming, i think it's great for that.
16:58	hiredman	I think apple is making a lot of hype for something not very hype worthy
16:58	technomancy	jonafan: it's more that it's totally irrelevant to me. I care about as much as I care about new advances in yachting technology.
16:58	dnolen	it's good to see other solutions to lockless concurrent programming
16:58	Chouser	you slap some Clojure closures on that and you've pretty much arrived. Then sprinkle on the agent abstraction, refs, transactions, etc...
16:59	hiredman	I mean, it's neat, but it is not the completely revolutionary game changer that apple declares everything to be
16:59	jonafan	i agree with hiredman
16:59	dnolen	hiredman: no, the marketing speak is ridiculous of course
16:59	jonafan	i honestly don't think it solves much and it's certainly not new
16:59	hiredman	I wonder if apple has filed for any patents on GCD
16:59	jonafan	and the terming is nuts
16:59	Chouser	hiredman: it may be hype-worthy to the extent that they've re-written all the main Mac apps to use this and actually take advantage of their multicore machines.
16:59	dnolen	hiredman: but bringing lockless concurrency to Cocoa programming is good stuff.
17:00	Chouser	it's only lockless if you don't have to coordinate changes on multiple pieces of state at the same time, right?
17:00	drewr	Multicore Finder is interesting to me.
17:00	hiredman	is it actually lockless concurrency?
17:00	hiredman	I doubt it
17:00	dnolen	Chouser: private queues handle shared datastructures.
17:00	Chouser	I'd like to see them write ants
17:01	hiredman	it is more about parallism then concurrency
17:01	dnolen	if i understand all the blocks on private queue get execute as a atomic op
17:01	Chouser	hiredman: of course there are locks inside, but the idea is apps don't have to explicitly deal with locks. Like clojure.
17:01	dnolen	could be wrong though, that's what the paper sounded like.
17:01	Chouser	dnolen: yeah, but think of the clojure ants example. One queue per ant? or one per cell?
17:02	jonafan	man, i trolled. sorry guys
17:02	Chouser	can't be one per ant, as they're all changing cell state.
17:02	dnolen	jonafan: no, I was interested in this as well, I brought it up earlier.
17:02	dnolen	concurrency is a huge interest to Clojurians :)
17:03	jonafan	i figured you would be good people to ask
17:03	dnolen	Chouser: I agree, implementing ants on it would be a great case study to see how good it is.
17:03	Chouser	so one queue per cell -- an ant queues it's "update" action on a queue for that cell ... but it needs to check nearby state.
17:03	hiredman	then main issue is it is still C so you still have direct references to mutable data
17:04	dnolen	hiredman: blocks provide an interesting solution to that, can functional programming be achieved in C by dealing with blocks
17:04	dnolen	blocks copy the stack
17:04	hiredman	oh
17:04	hiredman	but if what is on the stack is a point
17:04	hiredman	er
17:05	dnolen	i think there are issues of course, you can act at a distance, but I wonder how cheap blocks are.
17:05	dnolen	if they are cheap, it could be interesting.
17:06	dnolen	i am no C expert, but here's a description of how blocks will work
17:06	dnolen	http://lists.cs.uiuc.edu/pipermail/cfe-dev/2008-August/002670.html
17:09	hiredman	I've been watching http://www.youtube.com/view_play_list?p=9D558D49CA734A02 to improve my C
17:09	Chouser	closures with mutable locals variables.
17:09	hiredman	yeah
17:09	hiredman	seems like apple is in need of a blessed functional language for their stack
17:10	hiredman	Objective-F?
17:10	danlarkin	that sounds like a profanity
17:10	dnolen	Chouser: the closure looks like it supports const locals as well.
17:14	Chouser	my vague impression of Objective C's runtime model is that it would play nicely with a Clojure-for-OC
17:15	hiredman	slightly offtopic, toward the end of the those youtube videos, he switches from C to scheme and they use kawa
17:15	dnolen	Chouser: the only real missing bit it seems to me is there is no concept of final, doesn't Clojure get quite a perf boost from everything being final?
17:15	clojurebot	Who??
17:18	technomancy	is it just me, or are java users really bad at IRC?
17:19	hiredman	uh
17:19	danlarkin	there is already a lisp dialect for the objc runtime, Nu... might be worth checking out
17:21	mrsolo	hmm is there a more sophisticated version of memoize that works with limited space?
17:58	Drakeson	,(bit-or 3 5)
17:58	clojurebot	7
17:58	Drakeson	,(bit-or 3)
17:58	clojurebot	java.lang.IllegalArgumentException: Wrong number of args passed to: core$fn
17:59	Drakeson	clojurebot: how am I supposed to do (reduce bit-or some-coll) then?
17:59	clojurebot	0
17:59	Drakeson	botspank
18:00	Cark	,(reduce bit-or 0 '(1 3 5))
18:00	clojurebot	7
18:00	Cark	hum
18:02	Drakeson	Cark: thanks. I meant to be a bit sarcastic to get someone to fix (bit-or single-arg)
18:03	Cark	i don't know if that would be a good idea though
18:03	Cark	usually when using bit arithmetic, you want good performances
18:04	Cark	and it is well known that it's faster to use (+ 3 (+ 1 4)) than (+ 3 1 4)
18:04	Cark	for some reason =P
18:04	Cark	so i guess this would translate to bit arithmetic
18:05	durka421	,(meta (var bit-or))
18:05	clojurebot	{:ns #<Namespace clojure.core>, :name bit-or, :file "clojure/core.clj", :line 776, :arglists ([x y]), :inline #<core$fn__4129 clojure.core$fn__4129@1272e52>, :doc "Bitwise or"}
18:05	durka421	,(meta (var +))
18:05	clojurebot	{:ns #<Namespace clojure.core>, :name +, :file "clojure/core.clj", :line 549, :arglists ([] [x] [x y] [x y & more]), :inline-arities #{2}, :inline #<core$fn__3946 clojure.core$fn__3946@1b9ab6c>, :doc "Returns the sum of nums. (+) returns 0."}
18:05	replaca	Cark: it is also known that that behavior of + is a bug and will be improved at some point
18:06	Cark	right
18:07	Cark	though i wonder how to do this without sequences, and moreover without loosing the neat lazy parameters we have now
18:08	Cark	then again, i'm no language designer ...
18:30	Drakeson	,(cl-format nil "~e" 1/10)
18:30	clojurebot	java.lang.Exception: Unable to resolve symbol: cl-format in this context
18:30	Drakeson	,(clojure-contrib.pprint/cl-format nil "~e" 1/10)
18:30	clojurebot	java.lang.ClassNotFoundException: clojure-contrib.pprint
18:31	Drakeson	(requir 'clojure.contrib.pprint)
18:31	Drakeson	,(require 'clojure.contrib.pprint)
18:31	clojurebot	nil
18:31	Drakeson	,(clojure-contrib.pprint/cl-format nil "~e" 1/10)
18:31	clojurebot	java.lang.ClassNotFoundException: clojure-contrib.pprint
18:32	Drakeson	,(progn (require 'clojure.contrib.pprint) (clojure-contrib.pprint/cl-format nil "~e" 1/10))
18:32	clojurebot	java.lang.RuntimeException: java.lang.ClassNotFoundException: clojure-contrib.pprint
18:32	Drakeson	,(progn (require 'clojure.contrib.pprint) (clojure.contrib.pprint/cl-format nil "~e" 1/10))
18:32	clojurebot	java.lang.Exception: Unable to resolve symbol: progn in this context
18:32	Drakeson	,(clojure.contrib.pprint/cl-format nil "~e" 1/10)
18:32	clojurebot	"1./1E+3"
18:32	durka42	i think we call progn "do" ... but clojurebot will execute multiple statements anyway
18:33	Drakeson	,(clojure.contrib.pprint/cl-format nil "~e" 1/10)
18:33	clojurebot	"1./1E+3"
18:33	Drakeson	is this correct ^^^^^^
18:33	durka42	looks strange to me...
18:35	Drakeson	,(map #(clojure.contrib.pprint/cl-format nil "~e" %) (list 1/10 1/5 1/2 11/13 1/1 13/11))
18:35	clojurebot	("1./1E+3" "1./5E+2" "1./2E+2" "1.1/13E+4" "1.0E+0" "1.3/11E+4")
18:36	durka42	,(clojure.contrib.pprint/cl-format nil "~e" (.doubleValue 1/10))
18:36	clojurebot	"1.0E-1"
18:37	Drakeson	so this is a bug, right?
18:40	hiredman	looks like it
18:45	mrsolo	is there something simliar to range that work with something other than integer?
18:46	hiredman	,(map char (map (partial + 96) (range 1 26)))
18:46	clojurebot	(\a \b \c \d \e \f \g \h \i \j \k \l \m \n \o \p \q \r \s \t \u \v \w \x \y)
18:48	mrsolo	hmm don't see it something like (frange start stop step-function)
18:49	mrsolo	so (frange jan may next-month) gets you (jan feb ... may) etc
18:50	hiredman	iterate
18:51	hiredman	or even just use a map
18:51	mrsolo	iterate will do i think
18:53	hiredman	,(map [:jan :feb :march :april: :may :june :july :august :sept :oct :nov :dec] (map #(mod % 12) (range 15)))
18:53	clojurebot	Invalid token: :april:
18:53	hiredman	,(map [:jan :feb :march :april :may :june :july :august :sept :oct :nov :dec] (map #(mod % 12) (range 15)))
18:53	clojurebot	(:jan :feb :march :april :may :june :july :august :sept :oct :nov :dec :jan :feb :march)
18:54	mrsolo	month is just a example... specifally that i want to do is to get X sample points out of date-time range
19:33	replaca	Drakeson: looks like cl-format ~e isn't handling rationals right. That's the problem you're thinking about, right?
20:05	replaca	I don't think anyon'es looking at this anymore, but I just looked at the cl-format code and ~e for rationals is on the list. I'll push it up.