#clojure log - Oct 21 2009

The Joy of Clojure
Main Clojure site
Google Group
IRC
List of all logged dates

0:08 * technomancy goes on a patch-testing spree

0:15 Calamitous: I have a newb question if anybody has a moment :)

0:16 I have a list of data [1 2 3]-- I'd like to send the elements of the list as params for a method (defn foo [& a] ... so that it comes out as though calling (foo 1 2 3)

0:16 but the method only acts as those I'm passing a single param (foo [1 2 3])

0:17 Is there a way to pass the list elements in as individual params?

0:17 or am I completely barking up the wrong tree?

0:19 mikehinchey: apply

0:20 ,(apply + [1 2 3])

0:20 clojurebot: 6

0:20 Calamitous: oh i gotcha

0:21 kewl, thanx! :D

1:05 tomoj: is there an idiomatic way to "loop" through a seq with indices?

1:12 slashus2: tomoj: I guess you could use doseq and use range. What are you trying to do?

1:13 tomoj: I don't see how doseq would do it

1:13 cgrand: tomoj: there's indexed somewher in contrib. But iterating over a seq with indices isn't idiomatic.

1:13 tomoj: right

1:13 hiredman: (map (fn [v i]) some-seq (iterate inc 0))

1:13 tomoj: sometimes I've needed to do it, though

1:13 hiredman: ah, yes :)

1:14 cgrand: tomoj: agreed, sometimes you need it, often for interop

1:16 slashus2: What do you all think of these macros? http://paste.lisp.org/display/89023

1:16 tomoj: now I can't remember why I wanted it before

1:17 slashus2: I have been pondering over the second macro.

1:18 Maybe it is my bad habits, but I sometimes find that I need to do side-effects inside transactions. This seems to do the trick. I would like to add the option to specify whether you want to do a send or send-off.

1:20 The first one I found useful for replacements of the while (n = something > 5) { interop thing

1:20 tomoj: aren't transactions exactly where you want to be doing side-effects?

1:20 oh, you mean like printing stuff

1:20 slashus2: You shouldn't do side effects in transactions because it may repeat.

1:20 But the transactions only sends to agents when the transaction is complete.

1:20 send*

1:21 tomoj: I count mutating state as a side effect. is it not?

1:21 hiredman: it is

1:21 cgrand: slashus2: Why do you introduce a pred# and use resolve?

1:21 hiredman: but you don't do that in a transaction

1:21 tomoj: you don't mutate state in a transaction?

1:22 I thought that's what transactions were for?

1:22 hiredman: you take a an immutable value, produce a new immutable value, and when the transaction succeeds the new value is swapped in for the old

1:22 slashus2: cgrand: Because I am not very good at writing macros... :-P It is when you actually take the pred argument as data. It needs to be resolved as a function. I think at least.

1:22 tomoj: yes.. that sounds like mutating state to me

1:22 hiredman: no

1:23 slashus2: It is actually an atomic operation of pointing to a new object with potentially shared components.

1:23 cgrand: You shouldn't do *unmanaged* side effects in transactions because it may repeat. (managed side effects being send/send-off and anything that change refs)

1:23 tomoj: so it's atomic and the values are immutable, you're still changing the state, right?

1:23 hiredman: not within the transaction

1:23 tomoj: oh, I see

1:24 slashus2: cgrand: Do you see where the when-commit macro could be useful? It even allows you to specify your own agent when needed.

1:26 cgrand: slashus2: haven't looked at it yet

1:26 slashus2: okay

1:27 technomancy: tomoj: I've wanted doseq + counter like three times in the last week

1:28 considering a patch to add it

1:28 I'm using indexed right now, but adding it as another keyword arg to doseq/for would reflect intentions much more clearly.

1:30 tomoj: like (for [x aseq :index i] ...)?

1:30 slashus2: Have you all ran into the case that my when-let macro solves a lot?

1:30 cgrand: slashus2: http://paste.lisp.org/display/89023#1

1:31 slashus2: Can't believe I missed that. Sometimes destructed binding skips my mind.

1:33 cgrand: Looks wonderful. Thank you.

1:36 cgrand: I don't think your second macro down there works properly.

1:36 neither one

1:43 cgrand: http://paste.lisp.org/display/89023#2 I think this fixes it.

1:44 Or you could do without the apply and your version would work.

1:45 cgrand: slashus2: right, apply here is plain wrong

1:46 but I can't spot what you changed

1:46 without apply http://paste.lisp.org/display/89023#3

1:46 slashus2: Instead of ~@, I did '~

1:46 Meaning it still was in the list form

1:46 so apply could work on its contents.

1:47 Is the when-let macro well constructed?

1:47 I mean

1:47 when-commit

1:51 I am going to try to get Clojure working on a SR4 robot tomorrow. Need to figure out how to get clojure on the robot's module loader classpath.

1:52 hiredman: clojurebot: how can slashus2 get clojure on to the robot's classpath?

1:52 cgrand: ah ok, don't do that ('~) :-) and remove apply instead. '~ (and the induced resolve you had earlier) wouldn't behave like you expect: try with your last while-let: (while-let [x 0 > (+ 0 0)] (println "hello"))

1:52 clojurebot: clojure is a language to use if you want to up your game

1:52 hiredman: hmmmmm

1:53 that was an awful lot of cpu time for that

1:53 slashus2: cgrand: That is true....

1:54 hiredman: Yes. I have to extend some DefaultModule class to be able to initialize the module. It sends the kernel's module loader to it.

1:55 Thought it would be cool to write a module using clojure.

1:57 hiredman: lisp is the language of ai

1:58 something in clojurebot's fuzzy lookup logic is burning more cycles than I would like

1:58 cgrand: slashus2: http://paste.lisp.org/display/89023#4 you tend to do at runtime (in the expansion) things you can do better at compile-time

1:58 slashus2: It is for my AI class. We are using XLISP in our class. I should have tried to convince the professor to use clojure.

1:59 hiredman: nice

1:59 slashus2: cgrand: I see that now.

2:00 hiredman: I want to get pleo, and wire a gumstix to the debug port, then run clojure on it

2:01 but that is well over my toy budget

2:01 slashus2: These robots run a stripped down version of debian on ARM11 processors.

2:02 They have sonar, wifi, temperature gauges, etc.

2:03 hiredman: I google it :P

2:03 looks like a beast

2:03 man sonar

2:03 wow

2:03 slashus2: Looks sort of like R2D2

2:05 I wish it had arms that could grab things.

2:10 hiredman: I mean I was able to start a clojure repl inside the robot's OS. I guess that was pretty cool.

2:11 Unfortunately, I couldn't control any of the robot's modules successfully during the time I tried.

2:12 hiredman: :(

2:13 need a clojure <-> xlisp bridge

2:14 slashus2: No, the robot runs java.

2:14 I was just commenting that we use xlisp for some of the smaller applications in the class.

2:15 hiredman: oh

2:17 slashus2: hiredman: I tried to use :gen-class to make its main module, but, like I said before, I couldn't get clojure on the classpath.

2:19 Thought maybe there was a slight chance I could write my robot project in clojure.

2:21 hiredman: you might try writing a java class that creates your own classloader and use that to try and load clojure

2:22 slashus2: Is that very difficult?

2:22 hiredman: http://java.sun.com/javase/6/docs/api/java/net/URLClassLoader.html

2:22 I dunno

2:22 URLClaassLoader looks simple enough

2:23 slashus2: I will look into that.

3:55 AWizzArd: ~seen kotarak

3:55 clojurebot: no, I have not seen kotarak

6:53 AWizzArd: ~max people

6:53 clojurebot: max people is 187

6:53 AWizzArd: cemerick: last was 185 :)

6:53 licoresse: the daily max people routine ;-)

6:54 cemerick: it's good, it's good

6:54 I *still* wouldn't worry about it, tho. :-)

6:55 I think the tipping point of (at least moderate) longevity is long since past.

6:56 AWizzArd: As long Rich is healthy and interested.

6:59 licoresse: :D

6:59 plenty of brilliant people here

7:08 ambient: are there any ready-made macros for fast array processing? so i wouldn't have manually to aset aget everything?

7:09 like array-let and (set-array [a b c] my-array)

7:39 cgrand: ambient: apart from amap and areduce (which are not what you want), I don't know

7:39 what are you trying to do?

8:04 ambient: cgrand a lot of stuff with arrays of doubles. audio digital signal processing, real time

8:09 AWizzArd: cgrand: in PersistentHashMap.java I see that count is of type int. Does that mean that hashmaps are limited to contain 4 milliard elements only?

8:11 cgrand: AWizzArd: on a 32 bit CPU it's a safe bet. Java collections have that limit too.

8:11 ambient: ok, thx

8:15 AWizzArd: I see. So it is something for Java 8 or so, when it will be typical for business machines to have 0.5 TB RAM.

8:22 cgrand: AWizzArd: and int is signed so 2^31-1 items max

8:30 AWizzArd: cgrand: yes true, so it's even less

8:30 fortunately we have no need for 2 milliard k/v pairs right now in one single map ;-)

9:12 dom1410: hello i need some help.

9:12 i try to write a funktion which initializes a java object from a java class

9:12 like this

9:13 http://pastebin.com/m6871a4b8 but there is a java.lang.Exception

9:14 chouser: bla?

9:14 The-Kenny: dom1410: Are you sure the problem is this line? I don't see the symbol "bla"

9:14 dom1410: ^^ i try to call the funktion like this "(new-model bla)"

9:14 chouser: ah

9:15 dom1410: sorry..

9:15 The-Kenny: dom1410: Use a macro. the repl tries to evalue bla before calling the function.

9:15 chouser: 'def' does not evaluate its first arg

9:15 dom1410: ah ok..

9:16 chouser: it always takes the symbol you give it as the literal name of the Var to create.

9:17 dom1410: ok i so it's better to use a makro?

9:17 chouser: dom1410: also, it's generally discouraged to create new named Vars on the fly at runtime.

9:17 are you sure it's what you want to be doing?

9:18 dom1410: hm.. we want to make our GUI programm running with a console using a repl

9:19 chouser: you might consider something like (def DGEModels (ref {}))

9:20 then new-model can store your DGEModel instances there with appropriate concurrency behavior.

9:20 and no need for macros. :-)

9:21 tmountain: hey guys, I have a general question regarding Java interop

9:21 dom1410: that sounds nice... i will try..;)

9:21 thank you very much

9:22 tmountain: what's the typical approach in Clojure when working with Java classes that have side effects or aren't designed in a thread-safe fashion?

9:27 AWizzArd: tmountain: you should make sure to not use those classes in parallel then.

9:28 tmountain: AWizzArd: it's a hypothetical question really. I generally write almost everything side-effect free in Clojure, but some co-workers are interested in Clojure and probing into the theoretical what-ifs

9:30 chouser: tmountain: One general approach that is sometimes appropriate is to only mutute the object via a particular agent, to serialize access to it. (though that won't prevent reading it in a possibly invalid state)

9:31 tmountain: even when using a thread-safe class in Java, there are instances where you might have a conditional - if (!someVar.isSet()) { someVar.set(val); }

9:31 chouser: tmountain: but in my experience so far the biggest problem isn't accidental mutation of the object so much as keeping the stateful approach from leaking out into the rest of your program.

9:31 tmountain: and wrap that in a synchronized block from outside to make sure the whole operation is atomic

9:32 chouser: when you say serialize access to it, are you referring to using a blocking-queue type mechanism to ensure that everything happens in sequence or something completely different?

9:32 chouser: tmountain: you can use 'locking'

9:33 tmountain: yes, what you said.

9:33 (def out (agent output-stream)) (send-off out #(.write % "hello world"))

9:33 tmountain: so, if you're stuck in that situation, you should basically develop some policies to avoid an ad-hoc spaghetti threading nightmare

9:34 chouser: tmountain: yes. or use a different java lib. :-)

9:34 tmountain: chouser: preferrably, yes ;-)

9:34 so in the conditional example I posted above, you would use a lock, or the same agent approach?

9:35 chouser: I guess it would depend on the object and/or your desired usage.

9:36 tmountain: if you were to use a lock, what locking mechanism does Clojure provide?

9:36 chouser: 'locking'

9:37 tmountain: ahh, aptly named

9:38 thanks for all the help guys. I've convinced at least one co-worker to use Clojure for his senior project and more are gaining interest, so that's a good thing

9:39 snowwhite07: tmountain, Wow! thats really good

9:39 tmountain: snowwhite07: early adopters over here. I work at a startup with a lot of young people, so that helps

9:40 snowwhite07: tmountain, nice

10:13 AWizzArd: Will contribs sql lib by default return cursers, so that doing a select that will return huge amount of data won't be a problem?

10:15 uh, cursor :)

11:22 Licenser: hmm with the enclojure plugin for netbeans, is there a way to let it automatically add the second half of brackets? like typing ( and it adds ) after the cursor?

11:23 RomanRoe: Licenser: Wouldnt this be annoying when you want to wrap an existing form?

11:24 Licenser: RomanRoe: hmm I am used to it from TextMate, if I want to do it I just select what I want to wrap and then type ( so it autowraps it too

11:24 Chousuke: now you wish you had paredit.el :)

11:25 RomanRoe: Licenser: or the editor could be smart and only insert the ) when no character is behind the cursor

11:25 Chousuke: silly editors, thinking only in terms of characters... ;P

11:25 Licenser: Chousuke: I wish TM would work nicely with Clojure :P

11:25 RomanRoe: Chousuke: lol

11:26 The-Kenny: M-x mark-sexp ftw :)

11:26 Licenser: Sex?

11:26 arsatiki: yeaah

11:26 mark-sexp is The Thing missing from the TM plugin

11:27 Chousuke: emacs+paredit is so good that even though I can't use emacs at all I think I'm stuck with it forever for lisp editing.

11:27 The-Kenny: Licenser: No, sexp. from one ( to the corresponding ).

11:27 Licenser: oh okay

11:27 Chousuke: though even with paredit I can only use barf, slurp and kill-sexp

11:27 but it's still better than any other editor :P

11:28 Licenser: I want a good TM plugin :(

11:28 Well for the start I

11:28 d

11:28 arsatiki: Oh wow, there is a select enclosing brackets in TM \:D/

11:28 Licenser: argh

11:29 arsatiki: which plugin do you use, it seems that all I tried are broken or even more broken

11:32 arsatiki: Licenser: This one, I believe: http://github.com/stephenroller/clojure-tmbundle

11:33 The-Kenny: hm.. I never understand what's so cool about textmate.

11:33 arsatiki: Honestly, neither do I anymore.

11:33 Chousuke: it's sleek and maclike I guess.

11:33 arsatiki: Plz some1 rewrite emacs with clojure-lisp :-P

11:34 Chousuke: that would be fun

11:34 too bad emacs is basically core + eleventy billion libraries

11:34 The-Kenny: Just write a elisp-interpretet in clojure :p

11:35 *an elisp-interpreter

11:35 sorry, typing in bed isn't very easy.

11:35 Chousuke: and while rewriting the core is probably not too bad, it's the libraries that make emacs useful :P

11:35 The-Kenny: I remember someone has wrote an elisp-interpreter in common lisp and got it working.

11:36 Licenser: Nothing is really cool about textmate, it just integrates very nicely on OS X which is a big + in my eyes. It feels native to use and often the TM bundels are damn good

11:36 Chousuke: but if you write just an elisp interpreter, that won't help. /:

11:36 it'll still be elisp

11:36 arsatiki: I recently retried emacs after a 5+ year break

11:36 Licenser: write a elisp -> clojure compiler

11:36 saml: hey wanna teach me clojure over irc?

11:36 The-Kenny: Licenser: That's what I miss in aquamacs.. a better integration.

11:37 (It doesn't support services, for example)

11:37 arsatiki: It's scary how well your fingers remember all the commands after all this time

11:37 Licenser: I admit I never got used to EMACS

11:38 And at least for Ruby I didn't found a better editor yet

11:38 (then TextMate)

11:38 while there are a few things that horribly annoy me :P

11:38 stuartsierra: Some people tried to re-implement Emacs in Common Lisp. They never finished.

11:40 Maddas: One doesn't even finish such a task :-)

11:41 Licenser: I tell you a elisp -> clojure compiler is the way to go >P

11:41 tmountain: considering the current version is 23.1, there'd be a lot of ground to cover

11:41 Licenser: let it work a few days and you've emojure or how you want to call it

11:41 arsatiki: emacj

11:41 tmountain: I'm one of the heretics that writes Lisp code in Vim

11:41 The-Kenny: clomacs.

11:42 Maddas: tmountain: I use viper-mode (VI emulation), that works for me. Just don't mention it on #emacs. Or #vim.

11:42 * Licenser laughs

11:43 Chousuke: heh

11:43 chouser: I've been disappointed by viper mode.

11:43 The-Kenny: tmountain: I tried that too.. it just didn't work for me.

11:43 Raynes: I am disappoint.

11:43 chouser: So I started my own repl+editor. It's disappointing me as well.

11:43 Maddas: Oh, chouser and Chousuke are two different people :-(

11:44 chouser: so true...

11:44 kunley: Hi.

11:45 chouser: kunley: welcome.

11:45 kunley: How to make Ctrl-c break current operation in a repl instead of leaving?

11:45 tmountain: probably have to trap SIGINT somehow

11:45 Licenser: okay time for me to eat see you later people

11:46 Chousuke: hmm, there's a java IDE for emacs

11:46 using CEDET

11:46 * Chousuke tries

11:46 Chousuke: the plain java mode is a bit too plain...

11:47 rhickey_: any proposals for a different name for 'implement' in protocols given :implements in datatypes is about interfaces?

11:47 The-Kenny: Chousuke: Just use slime and clojure :)

11:47 Chousuke: adopt?

11:47 chouser: kunley: there's a thing in repl-utils to do that

11:47 Chousuke: The-Kenny: I'm about to hack src/clj/clojure/lang/Compiler.java :P

11:48 The-Kenny: Chousuke: Heh, ok.

11:48 kunley: chouser: thx, will check.

11:49 * rhickey_ decided against array equality default in deftype, better and less surprising to just match = and hash

11:49 arsatiki: rhickey: :follow or :follows, perhaps.

11:51 Licenser: I like follows that makes sense for a protocol

11:52 chouser: protocol should use :implement because it means the same as the Java word?

11:52 rhickey_: we're looking for a word to go first here: (_____ AProtocol AClass ...)

11:53 Licenser: what about :supports

11:53 Maddas: :satisfies, perhaps?

11:54 rhickey_: chouser: could be same/similar (AType :implements [AnInterface] ...), (implement AProtocol AClass ...) - too confusing?

11:54 arsatiki: If the name of the protocol does not actually contain the word "protocol", then :protocol might work. But it does seem odd that it is not a verb.

11:54 danlarkin: fulfill?

11:54 rhickey_: arsatiki et al - not looking for a keyword here, but a function name

11:54 arsatiki: oh

11:54 chouser: will protocols sometimes have the same name as a matching Java interface?

11:55 rhickey_: chouser: I hope not, but ns issues have to be clarified for both datatypes and protocols

11:56 (___ Seqable String (seq [s] ...))

11:57 Licenser: extend

11:57 rhickey_: Licenser: extend has the same relationship to Java extends that implement has to Java implements

11:58 Licenser: Oh didn't knew java stole that word already :P my java is horribl. It just sounded good there

11:58 danlarkin: technomancy repeated this quote to me yesterday, "There are only two hard problems in computer science, naming and cache expiry."

11:58 rhickey_: and implement is probably more correct

11:58 Licenser: very odd idea >P

11:58 teach

11:58 arsatiki: I like danlarkin's "fulfill" after looking at the example

11:59 Licenser: (teach Sequable String (seq [s] ...))

12:01 chouser: apply

12:01 rhickey_: I guess a primary question is, is :implements/implement too confusing. If not, then that's it

12:01 chouser: :-P

12:02 danlarkin: yeah I think they're too similar

12:02 liebke: I don't think :implements/implement is to confusing

12:02 Licenser: I am confused what would be confusing, but likely because I'm not sure what to confuse them with

12:02 and I

12:02 m hungry, so see you later peopel

12:03 cgrand: rename :implements to :interfaces?

12:03 rhickey_: conform

12:03 * Licenser tosses in http://thesaurus.reference.com/browse/implement

12:03 arsatiki: conform-to

12:03 chouser: implement/:implement is less confusing that map/map

12:04 arsatiki: Personally I don't think I'd be confused by implement

12:04 kunley: chouser: yeah, add-break-thread! does the trick. Your responsiveness is awesome btw

12:04 rhickey_: cgrand: interfaces are a host thing, if people ask, how do I implement an interface and you say :implements, no surprises

12:05 chouser: kunley: glad it'll work for you.

12:05 arsatiki: Too bad implement-protocol is probably too long.

12:05 chouser: hm. with-protocol

12:06 kunley: The-Kenny: I left comments on your clojure-couchdb commits. Basically I was wondering if vi & emacs clojure modes give different indents, or just it was coincidence that we had them different

12:06 rhickey_: chouser: mismatch given general (with-this do that)

12:06 cgrand: rhickey: true and the interface concept may not exist on the host (eg javascript)

12:06 chouser: using-protocol

12:06 apply-protocol

12:07 danlarkin: fulfill-protocol :)

12:08 The-Kenny: kunley: I've seen the comments and added some comments too. I don't know if there's a difference... The only thing I could do is telling my emacs to not use tabs for indent :)

12:08 arsatiki: Thesaurus suggested "effectuate" =)

12:10 chouser: realize

12:22 rhickey_: unlike interfaces, you will be able to 'unimplement' a protocol, might help in thinking about it. When implementing a protocol you are both saying how to do it, and adapting/extending the dispatch to the new type. 'include' et al then come to mind

12:23 cgrand: specify

12:23 rhickey_: cgrand: the defprotocol itself seems more like the spec

12:28 Luyt: Hello Rich, I've seen a few of your clojure talks videos, I like the way you *don't* say "eh...." dozens of times in every sentence, like so many people do ;-)

12:29 arsatiki: Haskell uses the term "instance declaration", but I can't get anything useful out of that

12:29 (except declare, but...)

12:30 rhickey_: Luyt: :)

12:32 adapt?

12:33 Luyt: Yes, and I'm now reading Stuarts book. I dabbled a bit with CL but I think clojure is more worthwhile learning - although I havent used it yet in any serious app

12:34 BTW, the java interop works nicely, it's easy to call java objects from Clojure, but does it work the other way round too?

12:42 chouser: Luyt: you can create classes and interfaces in Clojure that can be used in Java without the Java user even knowing they're implemented in Clojure.

12:42 Luyt: chouser: Excellent. Thanks.

12:43 All thanks to the JVM, I suppose?

12:43 tmountain: Luyt: you can also invoke Clojure from Java - http://en.wikibooks.org/wiki/Clojure_Programming/Tutorials_and_Tips#Invoking_Clojure_from_Java

12:43 chouser: well that certainly helps. :-) To make it transparent to Java users, you do have to do a bit of work in Clojure that you otherwise wouldn't bother with (gen-class and gen-interface to be specific)

12:44 tmountain: right, that's the other way to do it -- specify nothing special in the clojure code and require Java users to make foreign-looking calls.

12:46 tmountain: chouser: yeah, probably not the best way

12:47 but I suppose it's nice to have options at least

12:54 eno: where can I find explanation of the '->' macro?

12:55 i don't quite understand (doc ->)

12:55 The-Kenny: eno: It's really simple. Just play around with it a bit. You could also look in the source

12:59 tmountain: (-> 1 inc inc inc) would give you 4

12:59 just feeds the result of the last function in the the next

13:01 eno: tmountain: thx for the example

13:02 so the parameters after the 1st one should all be functions, right?

13:05 kinda like comp, just different order

13:06 technomancy: comp returns a function without calling its function arguments; this actually calls the functions

13:07 eno: right

13:10 cgrand: eno: parameteres after the 1st one they can be function + trailing args

13:10 ,(-> 1 (+ 1) (+ 1) (+ 1))

13:10 clojurebot: 4

13:11 cgrand: ,(-> [] (conj 1) (conj 2) (conj 3))

13:11 clojurebot: [1 2 3]

13:12 eno: so (+ 1) is like a partially applied function, except that it is only a form, not a function

13:13 good to know this idiom

13:13 Chousuke: it's most useful with java interop I think

13:13 cgrand: yeah and that only the first arg is missing

13:13 useful with nested data structures too

13:14 Chousuke: ,(-> "foobar" (subs 3) .toUpperCase); can intermix java methods and clojure functions, unlike ..

13:14 clojurebot: "BAR"

13:14 cgrand: ,(-> {:a {:b {:c 42}}} :a :b :c)

13:14 clojurebot: 42

13:15 eno: cool

13:15 Chousuke: there's also ->> which is a variant of this :)

13:15 chouser: ,(->> (range 20) (remove #(< 5 15)) (filter even?))

13:15 clojurebot: java.lang.RuntimeException: java.lang.IllegalArgumentException: Wrong number of args passed to: sandbox$eval--4434$fn

13:15 chouser: ,(->> (range 20) (remove #(< 5 % 15)) (filter even?))

13:15 clojurebot: (0 2 4 16 18)

13:15 Chousuke: it's fairly recent though.

13:16 cgrand: ,(-> "hello" java.io.StringReader. java.io.BufferedReader.)

13:16 clojurebot: #<BufferedReader java.io.BufferedReader@14ec8d6>

13:16 Chousuke: what fun. I'm trying to compile emacs and it fails with a bus error...

13:43 konr: How do the clojure XML libraries work? Do they offer something similar to Xpath?

13:47 cgrand: konr: they turn the XML into a tree, tree which can be traversed using clojure.zip. zip-filter (in contrib) is XPath-like

13:48 jasapp: Someone just was working on xpath, weren't they?

13:48 http://github.com/kyleburton/clj-xpath

13:49 cgrand: and (shameless plug) if your XML isn't namespace heavy enlive can allow you to query it with CSS-like selectors

13:58 Licenser: so how will it be called?

14:47 lpetit: rhickey: still searching a new name for implementation .

14:47 ?

14:50 rhickey_: lpetit: got one?

14:51 lpetit: rhickey_: not brilliant ideas, but thought I ought to share them now, and not complain later :) : defimpl, or defimplementation (but too long this last I guess)

14:51 rhickey_: as defmulti / defmethod

14:52 rhickey_: others have pointed out defmethod probably shouldn't be def- becasue it doesn't introduce any names, same with implement

14:53 lpetit: rhickey_ : I remember, but defmethod defines a method for the protocol, and has already opened the can of worms :-p

14:53 rhickey_: time to close it

14:54 lpetit: rhickey_: want to rename defmethod ?

14:54 rhickey_: ok. At least I tried :)

14:55 rhickey_: also, I've reconsidered the array and field equality stuff from yesterday. Today I think default should just be = (so identity for arrays), when you want anything else, define equals

14:55 cemerick: was there another alternative?

14:55 lpetit: rhickey_: I myself am not to annoyed by :implements / 'implementation . After all, for people not dealing with host code, it's almost not an issue. And for people dealing with host code, it will just resemble the cascading extension capabilities of interfaces

14:57 rhickey_: cemerick: lpetit had a neat suggestion for per-field flags - see bottom of: https://www.assembla.com/wiki/show/clojure/Datatypes

14:58 lpetit: rhickey_: I don't want to misunderstand you : it's just a change concerning array fields, or do you want to go back to a simple implementation with =, and specific equals for anything else (thus throwing away :equals metadata) ?

14:59 rhickey_: in the end I think anything that encourages people away from compatibility with = and maps is something they have to take on themselves

14:59 lpetit: right, no more :equals metadata

14:59 I slept on it and it didn't survive the night

14:59 cemerick: Good idea -- seems like something a macro could take care of on top of deftype

15:00 mikehinchey: rhickey_: however, defmethod does modify the application code like other defs, rather than being function or state-changing, so defmethod seems consistent to me

15:00 lpetit: rhickey_ : grmlml :-)

15:00 mikehinchey: rhickey_: function -> functional

15:01 liebke: rhickey_: I don't think :implements/implement is too confusing, but what about bind, bind-protocol, or bind-impl?

15:02 rhickey_: liebke: if it weren't for existing bind, maybe

15:02 liebke: fair enough

15:03 rhickey_: I agree :implements/implement is probably not too bad, I wondered over lunch if people might try to stick protocols in the :implements list

15:03 and if that would be wrong or not

15:04 liebke: interesting, that might make sense

15:54 danlarkin: I have a question about nested atoms:

15:54 is this safe? (swap! (@state account) assoc folder {})

15:55 state is an atom for a map keyed on account, and the value is another atom keyed on folder

15:55 hiredman: you are interacting with two atoms there

15:56 danlarkin: right

15:56 hiredman: atoms are uncoordinated

15:57 danlarkin: I'm asking if I've done that wrong, I guess

15:57 lpetit: danlarkin: why ask if it's safe ?

15:57 hiredman: I think you meant to use refs :P

15:57 danlarkin: oh perhaps you're right about that

16:00 yes, you're right, I just have my nested maps in a ref and then all operations will be in a transaction and I don't have to worry aboot it

16:00 about it, that is

16:02 chouser: right, otherwise you might swap the outer atom while still retrying the inner one -- when the inner one succeeds its value would have already been lost

16:03 danlarkin: thanks for the brain-check guys :)

16:04 sometimes I think about what an awful mess this would make in another language

16:25 mikehinchey: With clojure.test, is there a way to make it stop testing? If my database is down, I don't want tons of repeated errors, just 1 error from a :once fixture.

16:26 technomancy: perform the check in your fixture?

16:26 stuartsierra: mikehinchey: you could (System/exit 1) in the fixture

16:26 * technomancy recommends against System/exit since it precludes interactive test running with a persistent JVM

16:27 stuartsierra: yeah

16:27 technomancy: just conditionally execute the argument to your fixture

16:27 stuartsierra: Yeah, that would work.

16:28 * technomancy makes a mental note to modify some of his tests to do this

16:28 mikehinchey: the fixture is throwing the connection exception

16:29 wait, that doesn't sound right, let me check more closely

16:32 you're right, my connection isn't really happening until later, so I need a check in the fixture. thanks.

16:37 Licenser: hrm after working a bit with clojure in netbeans i get the feeling it tries to foce people into the java way of doing stuff, I already have like 3 packages and all that o.O meh

16:40 chouser: rhickey: so no per-field :equals flag in deftype? will always use = which is 'identical?' for arrays?

16:42 solussd: why does something like (range 100000) cause a outofmemory exception?

16:43 chouser: ,(last (range 100000))

16:43 clojurebot: 99999

16:43 chouser: solussd: ?

16:43 ,(reduce + (range 1e6))

16:43 clojurebot: 499999500000

16:43 mwoelker: ,(range 1000000)

16:43 clojurebot: Eval-in-box threw an exception:java.lang.OutOfMemoryError: Java heap space

16:44 mwoelker: don't use the whole seq at once i guess

16:44 chouser: well, you didn't want all those values printed anyway, I bet. :-)

16:44 solussd: seems to only throw the OutOfMemoryError when it is printed

16:44 mwoelker: probably not ;-)

16:45 chouser: set *print-length* or restrict the output in some other way.

16:45 Chousuke: solussd: that's because printing it holds the entire sequence in memory.

16:45 solussd: but I did... :) b/c I'm playing

16:45 mwoelker: it gets boring after the first three pages

16:45 technomancy: mwoelker: sssh; no spoilers

16:45 solussd: shutup! I'm looking for the pattern in the output of range

16:45 mwoelker: it was the butler...

16:47 solussd: the python interpreter will print ridiculously long lists generated by range for me. :) Maybe clojure is trying to stop me from wasting my time.

16:47 chouser: well, you can try starting your JVM with a bigger heap

16:48 ,(pr-str (range 1e5))

16:48 clojurebot: "(0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 1

16:48 chouser: I actually don't get an exception here.

16:49 solussd: try 1e6

16:49 i think the clojurebot has a bit more heap than i do

16:49 chouser: bah. takes 10 times longer. ;-)

16:49 solussd: ,(pr-str (range 1e6))

16:50 chouser: clojurebot: ping?

16:50 clojurebot: PONG!

16:51 chouser: I guess he didn't want to do your pr-str for you.

16:53 cemerick: Aside from immutable data structures being preferable in general, is there any reason why putting mutable stuff in an agent is a Bad Thing?

17:22 arohner: If I want to read from multiple refs at the same time, and I want them to be consistent, will derefing them in a dosync do what I want?

17:23 i.e. (dosync [@foo @bar])

17:26 eno: arohner: isn't it writer's responsibility to make sure the refs are consistent?

17:27 as long as all writers dosync, reader does not need to worry about consistency?

17:27 arohner: hrm. In my case, the refs are two different views of essentially the same data

17:28 eno: then when will it be inconsistent?

17:28 arohner: each ref is consistent, but I want to read them both at at point when they are consistent with each other

17:28 if I do [@foo @bar] without a dosync, a transaction could run between the @foo and the @bar

17:29 maybe my data model is wrong.

17:29 eno: if there is one single ref

17:30 holding a container of the two

17:31 arohner: yeah, I may have to do that

17:31 previously I didn't need to look at both of them at the same time

17:31 hiredman: arohner: you should get consistent information from (dosync [@a @b])

17:32 arohner: hiredman: thanks

17:33 hiredman: http://clojure.org/refs

17:35 konr: Is there a function that prints a tree with indentation, or represents it visually, so I can get a clue on what it is?

17:37 eno: hiredman: thx, "a consistent snapshot ..."

17:42 dnolen: konr: you mean like pretty print?

19:14 gregoryg: /leave

19:22 regularfry: Evening all... anyone got clojure-project working on emacs-snapshot-gtk on jaunty? It's not working for me, and I'm a little stumped

19:23 technomancy: regularfry: what are you seeing?

19:24 regularfry: technomancy: Exception in thread "main" java.lang.NoClassDefFoundError: clojure/main in *inferior-lisp*

19:24 clojure-src-root is correctly set

19:25 and M-x slime works. Just not M-x clojure-project.

19:25 technomancy: regularfry: clojure-project is meant to work with self-contained projects

19:25 so all the dependencies should exist in the project root

19:25 either as jars in lib/ or unpacked in target/dependency

19:26 regularfry: Ah, maybe I've misunderstood something then - I thought it created/copied/managed that process.

19:26 technomancy: that should be better-documented I suppose... I forget that a lot of people don't use dependency management. =(

19:27 regularfry: No worries. I think I just assumed that since all the *other* tools manage everything for me, that must as well :-)

19:28 technomancy: yeah, doing M-x slime without a project is mostly for exploration etc. and for hacking on clojure/contrib itself

19:29 regularfry: That's what I figured. It just seemed odd that one worked but not the other. As long as it's not a weird version-interop problem (and I haven't screwed anything major up) then I'm happy as larry

19:29 technomancy: great. (assuming larry is happy)

19:55 Licenser: hrm the API documentation is a horror :/

20:00 * technomancy believes the preferred term is "opportunity to contribute"

20:01 Licenser: technomancy: well you could put it that way

20:01 I get the feeling you

20:01 ... you've to know what you are looking for to find help in the API

20:02 dnolen__: Licenser: yeah, fortunately there's not much to know. Have you looked at the cheat sheet? It's much more helpful

20:02 technomancy: s/to find help in the API//

20:03 Licenser: they are great to refresh ones memory but I find the documents often very unhelpful if you want to figure out what function to use especiually when function parameters have so very expressive names like m s re i or something like that

20:03 ataggart: does something like this (admittedly trivial) fn exist?

20:03 (defn assocf [map key f]

20:03 (assoc map key (f (map key))))

20:04 technomancy: ataggart: maybe update-in?

20:04 ataggart: I *knew* it had to be there

20:04 thx, I'm blind today

20:06 * Licenser is trying to tokenize a sting and finds it horrible hard o.O

20:14 ataggart: (doseq [s (.split "hello world" " ")] (println s))

20:14 or use re-split from contrib

20:23 Licenser: ataggart: yes that works great if you have 1 rule, it gets quite complicated if you've more then 1 rule

20:24 there isn't a scanner feature, at least I didn't find one

20:25 ataggart: too complicated for a regex?

20:26 Licenser: I want to tokenize sourcecode

20:26 doing that in one single regexp is suicide

20:26 Tried it, is the wrong way :P

20:27 chouser: most tokenizers use a regex for each token, don't they?

20:29 Licenser: chouser: yes but not one for all tokens

20:31 chouser: oh, I see.

20:34 Licenser: there we go, re-scis born!

20:35 ugly uygly hack but it works

20:51 hah it's working!

21:06 hmm any hints about how I make a regexp multiline in clojure/

21:23 ah tricky found it

21:42 regularfry: technomancy: I'm not much cop at emacs lisp (yet... hopefully) but this does what I expect: http://gist.github.com/215647. Don't know if it'd be handy, consider it available equally for taking or ignoring

22:58 jlilly: how do you guys manage your jars? Do you guys all use that sort of run script I've seen about?

23:03 hiredman: jdk1.6 and beyond lets you put wildcards in the classpath

23:03 so I just put jars in ~/.jars/ and set my CLASSPATH to "$HOME/.jars/*"

23:11 jlilly: I think that's my issue. looks to be v1.5

23:15 wlr: hiredman: would putting symlinks in ~/.jars/ pointing to actual installed locations be a workable alternative?

23:20 jlilly: anyone know if openjdk6 translates into jdk1.6?

23:25 * jlilly hopes macports does him well :-/

23:44 yangsx: wlr: it's workable (I'm using Linux system).

23:47 wlr: yangsx: thanks. i might give it a try.

23:58 jlilly: having some classpath issues. hoping someone might be able to give me a hand. http://dpaste.de/Eu3R/

23:59 that's a traceback showing my CLASSPATH, and the commands I'm attempting.

Logging service provided by n01se.net