#clojure log - Jan 23 2012

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

0:06 ibdknox: Alright folks, I'm outta here

0:08 johnmn3: peace

0:28 randomly thought the /join #sopa. 358 people in there!

0:29 zellio: o.O

2:28 cemerick: I can't believe I hadn't noticed before now that special forms aren't hintable. :-|

2:31 muhoo: hmm that seesaw is interesting. would be cool to have a hiccup/enlive style clojure native syntax for java gui stuff

2:58 LauJensen: if I have an fn like (defn x [] (if (= "x" (System/getProperty "user.name")) x y) will the compiler optimize that so that the if is not re-run unless the environment changes or is the user.name var polled on every call?

2:59 johnmn3: I'd guess it stays as the user that started the java process

3:04 jowag: I think it won't optimize your fn.

3:28 amalloy: LauJensen: i would be quite surprised if the compiler optimized that, but you can trivially do it yourself: (let [result (if (= "x" (System/getProperty "user.name")) x y)] (def foo (constantly result)))

3:29 Blkt: good morning everyone

3:42 AWizzArd: amalloy: though I think that Lau wants it to get evaled if the Properties change.

3:43 Basically (if (and (System/propertiesUpdated) (= "x" (System/getProperty "user.name"))) …), but such a static method doesn't exist.

3:57 LauJensen: amalloy: Right - I was just curious if the compiler would work it out on its own - Would have been sweet

4:22 clj_newb: I can't find anything interesting on google for "clojure real time audio." Does Java's GC basically make this mpossible?

4:23 AWizzArd: clj_newb: did you see http://mad.emotionull.com/ or https://github.com/overtone/overtone ? They might deliver what you are looking for.

4:24 clj_newb: this is awesome

4:32 Kototama: hi, is it possible to use compojure with clojure 1.3?

4:33 Vinzent: yes

4:34 augustl: couchdb-lucene seems to include lucene itself, is that correct?

4:34 err, wrong channel

4:39 Kototama: how can I use clojure 1.3 with compojure then?

4:39 i'm already using it but the default dependency is clojure 1.2

4:42 AWizzArd: Kototama: you could import the Compojure dependency with an exclusion on Clojure, and manually add Clojure 1.3 as a dependency.

4:43 Vinzent: it works without excluding too

4:43 AWizzArd: Kototama: For example, in one project I include the Apache XML-RPC Client as a dependency. Unfortunately it comes with JUnit, which I don't want, so I did: [org.apache.xmlrpc/xmlrpc-client "3.1.3" :exclusions [junit]]

4:49 Kototama: ok thx i'll try that

5:04 when alternatives are written as [org.clojure/clojure "[1.2.1],[1.3.0]"], how can I specify which version from Clojure I want?

6:20 tsdh: Why is there no documentation for definterface in the Clojure API docs?

6:54 jaley: does anyone know where clojure.contrib.mock was moved to in 1.3? Can't find the info on the wiki...

7:11 AWizzArd: jaley: sorry, I don’t know that. But one association I have with “mock” is that Midje works in that style very well: https://github.com/marick/Midje

7:12 jaley: AWizzArd: ok cool. I was starting to look at Midje actually, I guess I'll carry on down that path

8:34 mr_rm: is nrepl considered to be the best way to have a remote repl connection in clojure 1.2 and 1.3?

8:35 that is: https://github.com/clojure/tools.nrepl

8:36 krunaldo: mr_rm: uh, normal nailgun repl or slime repl should be enough

8:36 mr_rm: also protecting it with a ssh/ssl/vpn/whatever tunnel is recommended

8:37 mr_rm: sorry, i should have said this is not from emacs. my IDE is eclipse ccw and i'd also like to connect from command line

8:37 krunaldo: but good point about the ssl tunnel - understood on that point

8:38 krunaldo: unsure how to proceed actually.

8:39 mr_rm: Do you need to have the repl running inside eclipse?

8:39 mr_rm: krunaldo: that is not a requirement.

8:40 krunaldo: basically i just want the running clojure program to listen on a port that i can connect to for a repl

8:40 krunaldo: hmm, nrepl seems to be the best one for that

8:48 kral: hi pals

8:59 compj: hi, does someone know why a aliased macro with (def ^#{:macro true) name #'other.ns/mac) produces a warning when called after it were aot compiled?

9:00 AWizzArd: compj: what is the warning?

9:00 compj: warning is: ArityException Wrong number of args (X) passed to <other.ns/mac>

9:01 can be tested with the seesaw project https://github.com/daveray/seesaw/issues/73

9:04 AWizzArd: compj: they suggest to delete the compiled class files. Can you try a “lein clean” first and then retry your example?

9:05 compj: yes that is working but I'm wondering, why can't you aot compile such a (def...)?

9:05 isn't the ^#{:macro true} meta data not intended for this, or is it a bug?

9:09 pandeiro: anyone use domina (clojurescript dom lib used by cljs one)? is there an easy way to iterate over a collection and output list items that I am missing?

9:16 compj: here, mnimal example of the bug (?) http://pastebin.com/n83uWJgu

9:18 AWizzArd: compj: can you try to clean and then aot the code and then try your test?

9:19 compj: the example I posted works outside a project, just put it in a empty dir, compile it and you can see the warning

9:27 kij: Hey, whats the state of the videos from Clojure/Conj11 ?

9:28 bhenry: kij: i saw them listed on confreaks.com but they are not available yet.

9:35 phil_: why doesnt this work?

9:35 (deftype Bar [a b c d e])

9:35 (def b (Bar 1 2 3 4 5))

9:35 #<CompilerException java.lang.RuntimeException: Expecting var, but Bar is mapped to class hs.base.view.Bar, compiling:(REPL:4)>

9:35 raek: phil_: you are missing a dot: (def b (Bar. 1 2 3 4 5))

9:36 phil_: raek: ah, did the syntax change?

9:37 raek: phil_: not since the release of 1.2, at least

9:37 kij: bhenry, Thanks.

9:37 phil_: raek: i see, then im looking at outdated examples :) thx!

9:37 raek: Bar is a jvm class, so you use the constructor interop syntax

9:51 lucian: i have a seq of maps, i'd like to split it into several seq based on one value in the map. is there hof i could use?

9:52 joegallo: group-by, probably

9:52 lucian: joegallo: that's it, thanks

9:52 joegallo: it's not lazy, though (i mean, how could it be?)

9:53 lucian: yeah

9:54 TimMc: joegallo: It could be partially lazy.

9:54 lucian: well, i guess it could be

9:54 TimMc: The first time you iterated all the way through an inner seq, the entire source seq would be realized.

9:55 joegallo: yes

9:55 you could do that

9:55 quite right

9:55 TimMc: Wait, the result is a map of seqs... no,

9:55 but if it is a seq of seqs, yes.

9:55 lucian: it'd work in something like haskell for sure, i think

9:56 semantically it's not necessarily fully eager

9:59 TimMc: You could write a lazy associative data structure in Clojure. But the first 'get on it for a key that wasn't already known would also cause a full realization.

10:00 phil_: is it possible to pr-str deftypes?

10:00 in the same way that i can pr-str maps for example?

10:01 pr-str only returns the java reference, but not the contents

10:03 raek: phil_: not yet, I think

10:04 something like that is in the works (at least for records)

10:04 phil_: ok, do deftypes then offer any advantages to self-made polymorphism?

10:05 i guess that returning closures for every newly created type may be inefficient :/

10:14 stuartsierra: pr-str works on defrecords in 1.3

10:15 But not deftype.

10:17 phil_: there are no defrecords in cjs though, correct?

10:29 wiseen: How do I call Java method overload that takes a vararg of byte[] ?

10:29 usually to call overload that takes vararg it's (Foo/bar (into-array type [...])

10:30 but how do I get byte[] as type ?

10:32 AWizzArd: wiseen: the Class of it is "[B"

10:32 You can type-hint such objects too: (SomeClass/method ^"[B" my-byte-array)

10:32 wiseen: (class (byte-array 0))

10:33 wiseen: AWizzArd, so (into-array [B [(byte-array ...) ..])

10:33 AWizzArd, so (into-array ^"[B" [(byte-array ...) ..])

10:33 ?

10:34 AWizzArd: byte-array takes a value-sequence.

10:34 wiseen: it takes varargs of byte-arrays not a byte-array

10:34 AWizzArd: ,(doc byte-array)

10:34 clojurebot: "([size-or-seq] [size init-val-or-seq]); Creates an array of bytes"

10:34 AWizzArd: ,(long-array [10 20 30])

10:34 clojurebot: #<long[] [J@134afa4>

10:34 wiseen: AWizzArd, yeah I know, but I need an array of byte-arrays

10:34 Java function has : wrappedBuffer(byte[]... arrays)

10:35 that's one overload

10:35 AWizzArd: Ah okay, then make-array

10:35 ,(make-array Byte/TYPE 10 20) ; a 10x20 array

10:35 clojurebot: #<byte[][] [[B@f6f949>

10:35 AWizzArd: ,(class (make-array Byte/TYPE 10 20) ; a 10x20 array)

10:36 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.RuntimeException: EOF while reading>

10:36 AWizzArd: ,(class (make-array Byte/TYPE 10 20) ; a 10x20 array))

10:36 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.RuntimeException: EOF while reading>

10:36 AWizzArd: ,(class (make-array Byte/TYPE 10 20)) ; a 10x20 array

10:36 clojurebot: [[B

10:36 Kototama: isn't it more clear that a *variable-like-this* is global? Why did that get deprecated? It's harder to see the difference between local and global variables now

10:37 AWizzArd: wiseen: note that (make-array Byte 10 20) is the wrapper Byte[][], while (make-array Byte/TYPE 10 20) returns you a native byte[][]

10:37 Kototama: I would also prefer if one could say: (def ^{:dynamic false} *my-var* 10)

10:37 wiseen: AWizzArd, so Byte/TYPE is a way to access value type byte

10:37 packed

10:38 AWizzArd: Kototama: But this way it is very easy to see which are bindable (and thus “slow”) vars.

10:38 wiseen: *unpacked type

10:38 AWizzArd: wiseen: yes

10:39 TimMc: phil_: There's a multimethod you can overload for printing your type, I forget what it is.

10:40 phil_: TimMc: but no multimethods in cljsc right?

10:40 but defrecords work, just tested

10:42 pandeiro: phil_: there weren't but the branch of cljs used by clojurescript one has them

10:42 phil_: pandeiro: multimethods or defrecords?

10:42 pandeiro: multimethods

10:43 phil_: and defrecords?

10:43 pandeiro: not sure about that

10:43 phil_: they work, but the docs say they shouldnt

10:44 is the branch used in cljs one gonna be merged back into the main repo btw?

10:44 pandeiro: phil_: not sure, probably have to ask the crew

10:44 phil_: pandeiro: allright, thx!

10:57 bhenry: can someone point me to a good futures resource? or tell me how to check the status of a future without blocking by deref?

11:01 TimMc: bhenry: realized? I think

11:02 ,(realized? (future 5))

11:02 clojurebot: false

11:03 jkkramer: ,(apropos "future-")

11:03 clojurebot: (future-done? future-call future-cancel future-cancelled?)

11:04 bhenry: future-done? thanks!

11:05 AWizzArd: Hmm, is the difference between realized? and future-done? more of a subtle nature?

11:05 It seems realized? works for more than just futures.

11:05 jkkramer: yes, it's more general. also, it's 1.3+ only

11:06 LauJensen: AWizzArd: Hi buddy, its been a while, is your DB out yet?

11:09 stuartsierra: There's also a non-blocking version of `deref` now that takes a timeout argument.

11:10 exupero: @joe http://en.wikipedia.org/wiki/1967_NFL_Championship_Game

11:10 AWizzArd: LauJensen: in use, but getting more stable.

11:10 exupero: sorry, wrong irc

11:10 LauJensen: Anything on Github or something I can try out?

11:10 AWizzArd: LauJensen: on BitBucket, but in a private repo :)

11:11 LauJensen: Oh okay - You're really drawing out the suspense on this one :)

11:11 In other News, ClojureQL is 1.0.3 and nearly bug free :)

11:12 AWizzArd: LauJensen: you are right, I am very slow on this. And Grats to the new version of CQL.

11:12 LauJensen: Thanks :)

11:12 AWizzArd: gtg, l8rs

11:18 langmartin: I have a namespace in program called platform.util, if I load a file that depends on it (slime + lein), I get the error that the namespace isn't found in the file.

11:19 If I open the file and load it, the namespace can be found

11:19 if I change the namespace to platform.toolbox, the load works as expected

11:19 us there an issue with a namespace called "util"?

11:20 Vinzent: are you sure the file is called "platform/util.clj"? if yes, I don't know

11:20 langmartin: yes, I am sure.

11:21 sritchie: technomancy: is there a way to exclude a file from a jar, but keep it around in the uberjar?

11:35 Dotan_: hi, is there an ubuntu package for clojure for a quick start?

11:37 pjstadig: Dotan_: I think there is, but i'm not sure how recent it is

11:37 may be 1.2.1

11:37 Dotan_: i see, ok, i'll build locally.

11:37 Vinzent: Dotan_, anyway install leiningen first

11:38 jsabeaudry: Dotan_, Closest thing to q quickstart that I know of is certainly leiningen: https://github.com/technomancy/leiningen

11:38 Dotan_: oh. i suspected installing lein first might be better but dismissed that thought

11:39 jsabeaudry: Dotan_, depending on what you intend to do, tryclj.com might be even faster

11:39 Dotan_: yep thanks, i always like to fiddle locally despite of try* sites :)

11:48 Kototama: is there a release for math.combinatorics (formerly clojure.contrib.combinatorics) ?

11:51 babilen: Kototama: http://search.maven.org/#artifactdetails%7Corg.clojure%7Cmath.combinatorics%7C0.0.2%7Cjar comes up

11:56 Kototama: hum okay

11:56 TimMc: Dotan_: In general, Clojure is not reliably available through package managers.

11:56 Kototama: i looked on clojars

11:56 thanks

11:57 Dotan_: TimMc, thanks. i went the lein way, very cool.

11:58 benares_98: :TimMc why isn't clojure reliably available through package managers?

12:00 TimMc: benares_98: Lack of maintainers, first of all -- but I think there are deeper problems with that distribution model vis-à-vis Clojure.

12:00 (and that was supposed to be an a-with-grave-accent, but my terminal is fucked)

12:01 benares_98: lol thanks for the insight

12:01 I was just curious, someone else said not to get clojure through package managers but I didn't ask why at the time.

12:02 TimMc: The reason is the unreliability. :-)

12:03 stuartsierra: There's a mismatch between the Java distribution model (every app has its own classpath of JARs) and most OS package managers.

12:04 babilen: What would need to be done to change that? What is wrong with clojure as currently packaged? Which libraries/applications need to be packaged most urgently?

12:07 dnolen: babilen: communicate to users that they should install lein via their package manager not clojure

12:09 babilen: technomancy: ^^^ Could you push a change to leiningen's README to mention Debian/Ubuntu packages? (Arch has packages as well, but I had the impression that they mostly just rely on lein's autoinstall)

12:09 wjlroe: I seem to have a value of byte[] coming from a sql query - how can I test for that - there's not bytes? method...

12:10 babilen: dnolen: But I am serious -- I am currently working on packaging most of https://github.com/clojure/ + other highly ranked libraries (compojure, ...) and am always open for suggestions.

12:11 dnolen: My long-term plan is to have everything managed by clojure/core in Debian in the newest version + backports to stable, but this is some work and any comments from the clojure community are welcomed.

12:11 algernon: babilen: midje & marginalia would be lovely to have, too.

12:12 babilen: algernon: Yeah, those are highly ranked on whatever website was mentioned on clj-user recently that provided a ranking based on dependencies.

12:13 stuartsierra: babilen: I don't want to denigrate your efforts, but I would not expect OS packages for Clojure libraries to get much use.

12:13 babilen: algernon: It's a bit hard to choose and I guess that a lot of developers will just rely on installing libraries via leiningen anyway. My dream would be that most of what is now on "Getting Started with ..." has a corresponding package in Debian/Ubuntu.

12:14 stuartsierra: Since the "packages" are just source code, there's little benefit to having them managed by an OS-level tool.

12:14 Also, Clojure is a young language and the landscape of libraries and tools is likely to change rapidly for the foreseeable futuer.

12:15 babilen: stuartsierra: Which is why for now only leiningen and clojure is packaged and I am debating with myself how to proceed. I guess that the real need for packages arise as soon as users really want to install applications written in Clojure.

12:16 algernon: babilen: *nod*

12:16 stuartsierra: In the unlikely event that desktop applications are written in Clojure, they would probably be distributed as monolithic JARs containing all their dependencies, as many Java applications are toay.

12:16 *today

12:17 hiredman: stuartsierra: but debian's policy is to take those monolithic jars and break them up again

12:17 stuartsierra: eh, whatever

12:17 hiredman: much like I think the debian package of clojure has(or had) asm stripped out

12:18 stuartsierra: nice trick since it was namespaced under the "clojure." package

12:18 babilen: yeah, but those monolithic JARs can and shouldn't be distributed by distributions -- I am currently trying to get most of clojure/core and other things into Debian and will merely package dependencies thereafter. I think, for example, that being able to "apt-get install clooj" and similar things would ease the life for people.

12:18 stuartsierra: go for it

12:18 babilen: hiredman: It is stripped out

12:19 hiredman: *eyeroll*

12:19 babilen: stuartsierra: Oh, I will. I just fear that this might end up in a similar situation as the Debian/Ruby disaster (developers being discouraged to use distribution packages) and would like to avoid that.

12:19 stuartsierra: too late

12:20 babilen: stuartsierra: Why?

12:20 stuartsierra: We already tell developers to use Lein.

12:20 babilen: hiredman: Huh? asm *is* packaged in Debian and used by clojure -- It is just not part of the Clojure package.

12:22 hiredman: babilen: sure, but why?

12:22 babilen: hiredman: Because there is no need to have the same library 10 times on a system.

12:23 stuartsierra: Which really makes sense right now. Just like it makes sense to target quite new and unstable libraries at this point. But clojure will (hopefully) settle down a bit eventually. As I said earlier: This will really become important once users want to use user-level applications in clojure.

12:23 hiredman: babilen: sure there is, clojure uses a specific version of asm, other tools may use other versions

12:23 stuartsierra: One dream at a time. 'Later dudes.

12:24 TimMc: It hink the question at the moment is: Does distribution via uberjars lock us in later?

12:25 And I'm not sure what the answer is.

12:25 babilen: hiredman: Those versions will have to be packaged independently in versioned packages if they are not compatible.

12:25 TimMc: babilen: Like Java?

12:25 babilen: TimMc: For example

12:26 hiredman: babilen: for what gain?

12:26 TimMc: I could see that.

12:26 hiredman: (over just taking the existing clojure jar with asm included)

12:27 TimMc: babilen: WOuld you have clojure-1.2 or clojure-1.2.1?

12:27 Never mind, that's irrelevant.

12:27 Clojure 1.2 breaks API, so it is clojure-1.2 v1.2 and v1.2.1

12:28 babilen: hiredman: There are a couple of reasons why it is undesirable. I guess the most important one is that it eases security support as you only need to update a single library and all applications that use that library are "patched" as well. If we would ship the same library in multiple packages all these packages would need to be updated.

12:28 TimMc: outside of that, you can handle it with OR statements.

12:29 babilen: TimMc: Right now we have clojure1.2 and clojure1.3 -- I expect a clojure1.4 to emerge after the release.

12:29 s/the/its

12:30 hiredman: babilen: but any security update is a new a version which may or may not work ith clojure, while the asm distributed with it is always going to work

12:30 babilen: If there would be a set of stable libraries for 1.3 and 1.4 that are individually maintained you could expect those to be packaged as something like libclojure1.2-foo -- But those are not versioned right now.

12:32 hiredman: In that case the dependencies have not been declared correctly -- software should clearly state that it works with specific versions. (say 1.5 - 1.6 or so) and security updates to a library shouldn't break backwards compatibility. (e.g. 1.6.1 shouldn't break applications written against 1.6)

12:33 hiredman: babilen: "shouldn't"

12:33 babilen: indeed

12:33 hiredman: babilen: they do all the time

12:34 babilen: clojure has no "dependency" on an external package, because the "external" package is not external anymore

12:34 TimMc: hiredman: That takes away the entire point of a security update.

12:34 babilen: hiredman: Sure there are dependencies (you write those down in your project.clj (for example) all the time)

12:34 TimMc: Not that you can read this.

12:35 hiredman: babilen: the whole thing is such a knee-jerk stiff-necked forcing of an antiquated C view of the world

12:35 babilen: clojure doesn't have any

12:36 (actually clojure doesn't build with lein so there is no project.clj, and even if you have a fork that does build with lein, which I happen to have, it doesn't have any dependencies)

12:36 babilen: hiredman: I am referring to Clojure-the-ecosystem rather than specifically clojure itself.

12:37 hiredman: babilen: if you want apt to work with the ecosystem the best thing to do would just by to make apt front maven

12:38 but that will not work well because apt with apt the default is an assumed single version of a package installed at one time, where as maven/the jvm just don't care

12:39 babilen: hiredman: We have things like that to ease packaging maven projects -- Please don't get me wrong here as my sole intent is to make it easier for Debian(-based) users/developers to use and enjoy Clojure. I do, however, strongly believe that a more stable development cycle + support for "old" versions will be a good thing in the long run.

12:40 hiredman: babilen: it is already very easy, you bootstrap yourself out of the apt nonsense and just use maven repos

12:42 babilen: the ruby community is seeing issues with the apt/C style global global namespace of libraries and has started using tools which avoid that (rvm, bundler, recommending local installs instead of global, etc)

12:45 babilen: hiredman: Indeed, whereas it works pretty well for Python even though *developers* still use tools such as pip/virtualenv/... during their development.

13:10 phil_: what is the function to generate an object from a string? im trying (read-string (pr-str object)) but i get an "unreadable form" error message

13:11 joegallo: you might need to play some of the ear-muffed variables that affect printing of objects

13:12 phil_: like atoms?

13:13 joegallo: i'm thinking like http://clojure.github.com/clojure/clojure.core-api.html#clojure.core/*print-level*

13:13 for instance

13:13 depending on what you're printing

13:13 phil_: yep (read-str (pr-str (atom 6))) bombs out :/

13:13 joegallo: i don't think you want to print and read atoms

13:13 just my opinion

13:13 you might want to print and read the value in the atom

13:14 but not the atom itself

13:14 phil_: well i was hoping that by reading the atom a new atom will be generated with the same value

13:15 joegallo: i don't think that atoms are designed to have a readabl toString, so that doesn't work

13:15 you could probably sneak around it if you really really want to do it that way

13:16 phil_: yea i guess, but it gets complicated... is there a reason why atoms are deemed unfit for "serialization"?

13:17 joegallo: ours not to reason why, ours not to make reply

13:18 Vinzent: phil_, there was discussion about that on the group somewhere... Why don't you save the value of your atom?

13:27 TimMc: I'm not sure what the point is of cloning transactional objects.

13:46 technomancy: has anyone had issues generating incanter charts from jdk7?

13:56 * jkkramer wishes docstrings could go after arguments

13:56 * compj too

13:57 the-kenny: +1

14:12 dnolen: jkkramer: compj: the-kenny: i see no patch in jira ;)

14:13 the-kenny: hmm shouldn't be that hard. Do I still need a contributor agreement?

14:15 technomancy: amalloy_: what's the latest on the data.xml release? still blocked on maven?

14:16 dnolen: the-kenny: yep

14:16 the-kenny: dnolen: Gnah. I should finally bring myself to do that :)

14:22 compj: can someone tell whether this is a bug? see comment: http://pastebin.com/raw.php?i=2GFGY2NW

14:23 TimMc: compj: I believe macro fns take some extra args, maybe &env and &form

14:23 Bronsa: compj: defmacro puts &env and &form as first args

14:23 hiredman: compj: it is not

14:24 Bronsa: TimMc: you were faster :)

14:24 TimMc: I'm on my work computer, which has less lag.

14:24 compj: but it works as exspected when I dont compile

14:24 jkkramer: dnolen: i'll write a patch. think i should bring it up on clojure-dev or just create a jira case?

14:24 compj: but if the module is aot compiled this warning is produced

14:25 dnolen: jkkramer: mention it in on clojure-dev just for record keeping

14:25 jkkramer: k, will do

14:25 dnolen: jkkramer: it always hard to tell how quickly people will move on such tickets since it's minor enhancement - so don't hold your breath

14:25 jkkramer: Clojure/core only really works on CLJ on Fridays

14:26 hiredman: dnolen: really, because I am pretty sure relevence is the one with an open source friday policy, and clojure/core claims not to be relevence

14:27 dnolen: hiredman: seeing as Clojure/core is mostly relevance + rhickey, the distinction is a minor one

14:27 hiredman: well, I guess that is not true, the website says "Clojure/core is a specialized technical practice brought to you by Relevance, Inc."

14:28 dnolen: right, I just know if you suggest they are the same on the ml you get immediately "corrected"

14:30 compj: it seems the compiler does not set the macro flag for vars defined using (def #^{:macro true})

14:31 hiredman: compj: what version of clojure are you using? #^ has been deprecated for two releases now

14:32 nrichards: Does anyone maintain a list of clojure users groups / meetups? I can't seem to find one

14:32 technomancy: nrichards: there's one on jira

14:32 compj: I just started using clojure, I''m on version 1.3.0 but learned from older tutorials

14:32 jkkramer: dnolen: yeah, i know. i figure i shouldn't complain unless i'm willing to actually write a patch, though :)

14:32 technomancy: which is not renowned for making things easy to find

14:32 nrichards: The Austin Clojure meetup is starting up, and I was hoping to get some inspiration from what other groups are doing

14:33 jira? OK. I'll check. thanks

14:33 di-csuehs: Austin Clojure?

14:33 nrichards: yes

14:33 dnolen: nrichards: sweet! I used to live in ATX

14:33 technomancy: nrichards: sorry, confluence

14:34 nrichards: http://www.meetup.com/Austin-Clojure-Meetup/

14:34 hiredman: compj: metadata on the var is the only "macro flag" there is

14:34 but you also need to stop using single segment namespaces

14:34 ~namespace

14:34 clojurebot: It's greek to me.

14:34 hiredman: clojurebot: jerk

14:34 clojurebot: you cut me deep, man.

14:35 compj: the single namespace was just for the demo

14:35 hiredman: ~namespaces

14:35 clojurebot: namespaces are (more or less, Chouser) java packages. they look like foo.bar; and corresponde to a directory foo/ containg a file bar.clj in your classpath. the namespace declaration in bar.clj would like like (ns foo.bar). Do not try to use single segment namespaces. a single segment namespace is a namespace without a period in it

14:35 compj: so how do I add the :macro true to a var in 1.3.0?

14:36 hiredman: compj: so what makes you think the compiler is not

14:36 compj: so what makes you think the compiler is not seeing it as a macro?

14:36 compj: see the example http://pastebin.com/raw.php?i=2GFGY2NW

14:36 hiredman: I see the example, and I don't see anything there that would lead to that conclusion

14:37 compj: i expect no warning when using (mac-alias) since it is marked to be a macro

14:37 and if i do (.setMacro #'var) it works

14:37 so my assumptions was the compiler does somehow drop this information

14:37 while it works when you don't aot compile it

14:37 Raynes: compj: Not that it matters, but refheap.com and gist.github.com (and a couple of other ones) are pastebins that have Clojure-specific highlighting. The first one is even written in Clojure by a couple of really awesome dudes. One whose name starts with an R.

14:38 Bronsa: oh, i recall there was a problem like this for ^:dynamic

14:38 compj: raynes: tanks :) I was looking for a clojure paste service but did not find one

14:38 hiredman: compj: have you looked at what setMacro does?

14:38 TimMc: Raynes: Get on that SEO!

14:39 Raynes: TimMc: Indeed.

14:41 compj: no, I did not look what setMacro does but I think it sets a var so that the 2 args passed to macro funs are handled in a way the user expects it

14:41 hiredman: compj: nope

14:41 Raynes: TimMc: Looks like "clojure pastebin" is covered in pastebin.com. and thus destroys us.

14:41 djh_: Can anyone tell me what the significance is of surrounding var declarations wtih *s, e.g. *database-connection*?

14:42 di-csuehs: ugh...why does it require a photo

14:42 Raynes: Those usually indicate dynamic vars that will be rebound at some point with ##(doc binding)

14:42 lazybot: ⇒ "Macro ([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. The new bindings are made i... https://refheap.com/paste/386

14:42 pjstadig: djh_: it generally conveys the fact that your var is meant to be bound dynamically

14:42 though in 1.3 you have to specifically add the dynamic metadata

14:43 Raynes: di-csuehs: Why does what require a photo? :o

14:43 compj: hiredman: but you agree the behaviour is not as expected?

14:43 djh_: pjstadig: thanks, that's all I needed to know :)

14:43 compj: since the compiled and non-compiled version differ?

14:43 hiredman: compj: I dunno, I haven't checked, and since you are new to clojure I think it is more likely you are making a mistake somewhere

14:44 di-csuehs: nrichards Austin Clojure meetup

14:44 nrichards: di-csuehs: the group? I think that's a meetup.com thing.

14:44 I didn't setup the group - but I can ask Sam to look at the configuration. I know he wrote the signup questions

14:45 di-csuehs: I found one...glad I had my laptop handy. See you there!

14:46 compj: yes, of course I thought of this option. but I could not find any documentation for this so I think it's possible that this might be a bug

14:46 the call to setMacro was just a workaround I picked from the defmacro code

14:47 nrichards: di-csuehs: awesome!

14:47 di-csuehs: assuming the kids have been good ;)

14:57 Bronsa: chops: fwiw with this patch http://sprunge.us/KZbH it works

14:58 chops: Bronsa: I think you might be confusing me with someone else

14:58 Bronsa: yeah

14:58 tabfail

14:58 chops: lol :)

14:58 Bronsa: *compj

15:00 compj: Bronsa: so it's a bug?

15:00 Bronsa: it's a missing feature i guess

15:00 compj: :)

15:01 Bronsa: if somebody with a CA can submit it, it'd be great

15:01 compj: +1

15:05 TimMc: Bronsa: You'd want to reformat with tabs. (yuck)

15:05 Bronsa: oh :(

15:16 TimMc: Bronsa: The clojure.core formatting guide is pretty heinous.

15:16 Bronsa: then i guess http://sprunge.us/bWcc this should be good

15:17 Raynes: TimMc: Have you actually looked at clojure.core's formatting?

15:17 TimMc: It looks like it was edited by 20 different editors.

15:18 llasram: Raynes: How so?

15:19 TimMc: like someone unleashed a whitespace-randomizer on it, I know

15:20 Raynes: llasram: Well, Emacs indents non-defun-style things with one space if you put a newline before the first argument. `cond` is an example. Vim doesn't bother and puts two spaces. This sort of inconsistency is all over clojure.core.

15:20 I don't actually care about 1 vs 2 spaces. It's just worth noting.

15:21 llasram: Raynes: Ah, I see. Interesting. I had not noticed that. And also thought you were talking about the clojure.lang Java code for some reason

15:21 Raynes: I wouldn't dare look at that.

15:22 llasram: I was put off by it at first, but after poking around, it's not bad. The formatting is unconventional, but largely consistent. There aren't many comments, but most of the code (other than the reader) is pretty clear

15:27 TimMc: Raynes: Ah, I was talking about the Java stuff.

15:31 dnolen: TimMc: blame it on IntelliJ

15:35 TimMc: dnolen: Why? Isn't it configurable?

15:38 Raynes: TimMc: Yes. It has settings "Ugly, Hideous, and Unreadable."

15:38 dnolen: Personally I have few problems with the Java source, weirdly formatted - yes. Unreadable hardly.

15:41 llasram: I wonder where rhickey (I assume?) got that formatting style from. It's very much like the formatting style McConnell recommends for C-ish code in /Code Complete/ (for pretty concrete reasons, although I still don't see it very often or use it myself)

15:43 jkkramer: http://en.wikipedia.org/wiki/Indent_style#Whitesmiths_style

15:44 llasram: Interesting.

15:44 hiredman: I like to imagine it was just an easy style to generate from a set of cl macros

15:45 llasram: heh

15:45 hiredman: the only joking part of that sentence is "cl macros" it would of course be generated by cl functions

16:12 grim_radical: is there a way to add a docstring via "defmethod"?

16:13 i see how to do it for defmulti, but i'd like to add some method-specific docs

16:13 phil_: is there a way in paredit.vim to swap to expressions (or just move an expression to the right / left)?

16:13 two*

16:13 Vinzent: grim_radical, I don't think it's possible

16:14 grim_radical: Vinzent: bummer. :(

16:16 raek: grim_radical: in defmulti/defmethod there is only one var involved (and therefore only one place for the metadata to go), namely the one that is defined by defmulti

16:16 grim_radical: raek: ah, ok

16:17 TimMc: raek: Is there a way to ask a multimethod for the implementation it would use given an argument list?

16:17 brehaut: grim_radical: if you used https://gist.github.com/432752 to install the methods so you could define the functions independently so that they each have a var to hang a doc on

16:17 grim_radical: brehaut: neat! i'll definitely take a look at that

16:17 raek: TimMc: I'm not aware of any public interface for that

16:18 brehaut: grim_radical: you wont get much advantage over just using a ;; comment

16:18 grim_radical: brehaut: yeah, that's my current fallback

16:20 raek: the defmulti docs or the ns docs where the defmethod resides are alternative locations

16:21 jsabeaudry: Is there a way to comment out "the next s-exp" ?

16:21 brehaut: ,#_ (prn "a") :foo

16:21 clojurebot: :foo

16:21 brehaut: jsabeaudry: ^

16:22 jsabeaudry: brehaut, Excellent! Thanks!

16:22 brehaut: jsabeaudry: use it carefully. its a great way to cause you pain ;)

16:23 jsabeaudry: not all syntax highlighters highlight it and the form thats removed as a comment

16:23 jsabeaudry: brehaut, Oh ya, clojure-mode doesn't color it :(

16:23 brehaut: indeed

16:24 im pretty sure the JS SyntaxHighlighter brush does

16:26 (not that that is useful when writing code)

16:32 jondot: hey guys, i've got a couple of questions while evaluating clojure

16:32 brehaut: jondot: just ask them

16:32 ~anyone

16:32 clojurebot: Just a heads up, you're more likely to get some help if you ask the question you really want the answer to, instead of "does anyone ..."

16:32 jondot: first, is clj good for handling low level binary structure? something like erlangs pattern matching

16:33 sorry my g/f interupted

16:34 hiredman: match (the pattern matching library) has some facilities for matching binary data, but I don't know how widespread that is in practice (I haven't used them)

16:34 brehaut: jondot: clj inherited some pain from jvm: theres no unsigned primative types which can make unpacking some stuff difficult. but beyond that binary data is fine

16:35 compj: TimMc: something happenend to the :macro pathc?


16:35 *patch

16:35 hiredman: https://github.com/clojure/core.match

16:36 jondot: i see. well i planned my first project (just to get a feel for the language) to be an implementation of this, for ring: https://github.com/jondot/rack-ping

16:36 brehaut: hiredman: dnolan was suggesting yesterday that people probably shouldnt build ontop of match till it gets to beta because of some gnarly issues still being resolved

16:36 jondot: i'm wondering, if that is a good fit for a first project - it would mostly be HTTP pipework, regex and imperative style workflow

16:37 tmciver: jondot: not sure about pattern matching but gloss is supposed to be a good binary data lib: https://github.com/ztellman/gloss

16:37 hiredman: brehaut: I see

16:38 TimMc: compj: Is it deleted? Bronsa created a new one. http://sprunge.us/bWcc

16:39 compj: I mean will it be added

16:39 jondot: tmciver: thanks, that looks interesting

16:40 TimMc: compj: Oh, no idea. I'm not on the Clojure/core team (though I do have a CA), I'm not invested in the issue, and it has an easy workaround (I think).

16:40 brehaut: jondot: what does rack_ping do that needs binary eating?

16:41 jondot: brehaut: sorry, i was afraid someone would make the connection. it is unrelated, just me asking question #2.

16:41 brehaut: oh right.

16:41 in that case, yeah ring level stuff is a fine place to get started in clojure

16:42 jondot: i dont know rack well enough to be able to make sense of the rack ping code though

16:43 jondot: well i've also implemented the same code in javascript (node) and c# (.net) any of that make sense to you?

16:43 brehaut: i understand all the langs, but not the libraries ;)

16:44 jondot: ah! well, my question i guess is along the lines of - there is a bit of grunt work there, like opening a HTTP pipe, regex'ing the content, and 3-4 decision points.

16:44 brehaut: jondot: is it just setting up a url that will attempt to test a variety of http resources?

16:44 jondot: yes.

16:44 brehaut: jondot: thats all trivial!

16:44 clj-http takes care of making http connections easily

16:45 think of it as the dual of ring github.com/dakrone/clj-http/

16:45 jondot: yes, it is very trivial, but my closest background to lisp is scheme (i've never even opened a socket in scheme) :)

16:46 brehaut: ive never opened a socket in clojure ;) not directly but ive written some xmlrpc crap which is similar to what you are doing

16:46 just with more mess

16:46 jondot: well, great then, i'll give it a try for fun.

16:46 brehaut: jondot: https://github.com/brehaut/necessary-evil/blob/master/src/necessary_evil/core.clj

16:47 Raynes: I've opened a socket.

16:47 I didn't like what was inside.

16:47 jondot: hehe

16:47 Raynes: https://github.com/Raynes/irclj has some socket stuff, I think.

16:47 compj: CA is quite a barrier if you just want to make a small contribution

16:48 Raynes: I agree, but it has been discussed until the end of time and they aren't going to change it.

16:48 compj: very sad but they surely have their reasons

16:48 brehaut: jondot: i dont claim that that lib is good code, but its doing the mechanical stuff you want

16:49 jondot: good, i think both of those project make a great reading for newcomers (or at least myself)

16:49 compj: isn't there a simple way to just open a ticket or point to an issue/patch?

16:49 jondot: ah!! and the last question - clj does have tail call right? courtesy of JVM?

16:49 Raynes: They use JIRA which… Sorry, I actually have no clue what they use that abomination.

16:49 jondot: It doesn't. But it has 'recur' which works around that.

16:50 brehaut: jondot: nope. it has loop / recur to simulate self calls in the tail position

16:50 Raynes: Well, you don't need a loop for a recur.

16:50 brehaut: of course. my bad

16:50 jondot: hmm is that JVM's fault? i remember one platform which didn't have TCO that felt very weird to me.

16:50 Raynes: https://refheap.com/paste/388

16:50 brehaut: if you want mutual recursion to be tail recursive you need a trampoline

16:51 Raynes: The JVM doesn't have anything for tail calls.

16:51 TimMc: jondot: JVM might add TCO some day, there was talk of it.

16:51 brehaut: yeah the JVM is to blame. clojure could just do its own tialcalls (like scala does) but that causes problems for interop

16:51 TimMc: Didn't make it into 7.

16:51 Raynes: jondot: See my paste above for an example of recur.

16:51 jondot: oh, well nice to know that scala does it. thanks Rayens - checking it

16:51 Raynes: jondot: You can basically just replace a tail call with recur and it magically works.

16:52 jondot: i see, well from my narrow experience, looks like good enough

16:53 thanks guys, i've learnt a lot, off to start fiddling around with ring!

16:54 technomancy: avian has TCO

16:54 Raynes: technomancy: And flu.

16:54 ~rimshot

16:54 clojurebot: Badum, *tish*

16:54 brehaut: avian?

16:54 technomancy: Raynes: how badly do you want it?

16:55 Raynes: That sounded dirty.

16:55 technomancy: clojurebot: avian is an alternate JVM implementation that's capable of running some subset of Clojure with TCO and continuations: http://oss.readytalk.com/avian/index.html

16:55 it also can AOT-compile straight to machine code

16:55 clojurebot: where did you go?

16:55 clojurebot: Pardon?

16:55 brehaut: huh thats really interesting

16:56 Raynes: Yeah, it's totally stellar, dood.

16:56 technomancy: clojurebot: avian is an alternate JVM implementation that's capable of running some subset of Clojure with TCO and continuations: http://oss.readytalk.com/avian/index.html

16:56 clojurebot: Roger.

16:56 brehaut: hows its performance compared to oracle/open jdks?

16:57 clojurebot: Alles klar

16:57 technomancy: brehaut: not very good

16:57 brehaut: i guess that shouldnt be surprising

16:57 technomancy: it's intended for embedding in larger C programs a la guile IIUC

16:58 brehaut: oh right. thats interesting in its own right

16:59 arkh: is there a why to specify classpath for clojure-jack-in ?

17:00 brehaut: arkh: err, i would guess that you do so in the project.clj for the project you are jacking in to

17:00 raek: arkh: clojure-jack-in uses whatever classpath leiningen provides it.

17:01 arkh: I'm specifically running into a problem with clojurescript one and it's dependences that aren't in ./lib

17:02 maybe I just need to look at leiningen but clojurescript one has a shell script that grabs its sundry dependencies

17:09 brehaut: holy crap. international shipping on oreilly books is prohibitively expensive 30$

17:09 kedoodek: ebooks are the future

17:09 jodaro: and the present

17:11 brehaut: no kidding

17:14 arkh: how does one add to what lein builds for a classpath?

17:15 technomancy: arkh: I think you should try the lein-cljsbuild plugin

17:15 schleyfox: hi, I'm having a bit of trouble with proxy. I'm using a lib that requires anonymous classes derived from a class, so I'm proxying it. It's two functions are the same arity and name but different types

17:16 I'm type hinting in my proxy, but I can't seem to get the right method called

17:16 phil_: technomancy: ive been using cljs-watch for now but this looks great

17:20 arkh: brehaut, raek, technomancy: thank you

17:24 schleyfox: so any ideas on why proxy is ignoring my type hints and proxying the wrong function?

17:24 the-kenny: phil_: cljsbuild is awesome. I used cljs-watch some weeks ago too, cljsbuild is much better :)

17:24 (though not perfect)

17:25 brehaut: schleyfox: you'll probably want to paste up an example somewhere (gist, refheap etc)

17:26 emezeske_: the-kenny: taking suggestions on how to make cljsbuild perfect. ^_^

17:26 the-kenny: emezeske_: haha :D

17:26 emezeske_: I'll have to try the newest version first, I'll come back to you :)

17:26 muhoo: didn't proxy get replaced by something else? reify maybe?

17:26 emezeske_: the-kenny: cool, thanks

17:27 technomancy: emezeske_: do you have a version that uses the canonical clojurescript artifacts yet?

17:27 brehaut: muhoo: not exactly. proxy fills a similar but different role

17:27 emezeske_: technomancy: yep, the latest version depends on them

17:27 technomancy: great

17:27 brehaut: muhoo: in particular proxy's are fully dynamic (and can be monkey patched!) and allow you to proxy to a class, where reify is only protocols and interfaces

17:27 the-kenny: emezeske_: One thing which I found slightly annoying was the copying of the .clj files to the cljs dir in order to make crossover compilation possible.

17:28 emezeske: Also, lein cljsbuild auto didn't watch the crossover source-files.

17:29 emezeske: the-kenny: regarding point #2, when you try the latest version, if it doesn't do that, could you open an issue? I thought that worked.

17:29 the-kenny: regarding point #1, would you prefer they were copied elsewhere? or just used in-place?

17:30 schleyfox: brehaut: https://gist.github.com/1665869

17:30 muhoo: also reify works on protocols and interfaces

17:31 I need concrete class stuff

17:31 hiredman: schleyfox: type hints like that do nothing for proxies

17:32 they work by name+arity

17:32 the-kenny: emezeske: In-Place. It's too easy to confuse the copy with the original .clj and make changes on the wrong file

17:32 schleyfox: hiredman: damn, any way to get that behavior?

17:32 the-kenny: emezeske: I'll open that ticket when it fails with the newest version :)

17:32 hiredman: so you'll need to check if ther argument is an instance of something else and forward via proxy-super

17:33 (if (instance? … …) do-stuff (proxy-super thisMethod …))

17:33 schleyfox: hiredman: but from the docs: "they have no other access to protected members, nor to super, as these capabilities cannot be proxied."

17:33 hiredman: ,(doc proxy-super)

17:33 clojurebot: "([meth & args]); Use to call a superclass method in the body of a proxy method. Note, expansion captures 'this"

17:34 schleyfox: yeah, I just did

17:34 kmicu: emezeske_:cljsbuild support for ":define "goog.userAgent.ASSUME_MOBILE_WEBKIT"

17:34 " in project.clj would be great! :)

17:34 schleyfox: cool, thanks

17:34 the-kenny: emezeske: Sorry, have to go. University is so annoying :(

17:35 Night!

17:36 MenTaLguY: hi, so I was reading http://clojure.org/lazy, and I'm not sure I fully understand why the "step" helper function is necessary to avoid a leak

17:37 hiredman: that is old stuff

17:37 streams are currently off the table

17:37 MenTaLguY: eh?

17:37 is lazy-seq being dropped from future versions of clojure then?

17:37 hiredman: oh, sorry that is not the streams page, forget I said anything

17:37 MenTaLguY: oh

17:37 hiredman: streams are something totally different

17:38 nickmbailey: oh i saw the streams page the other day and thought it looked interesting, not happening?

17:39 broquaint: I'm wonder if I'm getting lazy when I consider a lein plugin to add deps to project.clj e.g lein add-dep noir ...

17:41 hiredman: https://gist.github.com/a1f737bfe0d70a385d4c

17:41 emezeske: kmicu: is that a cljs compiler option?

17:42 kmicu: emezeske: yes java -jar compiler.jar --help

17:42 emezeske: more info in Google Closure Definitive Guide

17:43 emezeske: kmicu: do you know if there's anyway to pass that option via the normal 'cljsc' command?

17:43 kmicu: emezeske: i try to run ClojureScript One on Android and now I've done several tricks like that option manually :)

17:44 emezeske: kmicu: I see. Right now, I think the options that can be passed through the clojurescript compiler into the google closure compiler are kind of limited

17:45 kmicu: if you wouldn't mind opening an issue at https://github.com/emezeske/lein-cljsbuild/issues?sort=updated&direction=desc&state=closed , I'd appreciate it

17:45 MenTaLguY: hmm

17:45 kmicu: emezeske: no problem

17:45 MenTaLguY: so is the helper function, "step", really necessary to avoid a leak in the example on clojure.org/lazy?

17:45 emezeske: kmicu: thanks.

17:45 schleyfox: hiredman: ok, so the issue now is that my proxied method only gets called from within the method that I proxy-super to, so my proxy is removed while it executes

17:45 MenTaLguY: or is that there for some other reason

17:46 schleyfox: hiredman: I assume this is now the time to break out my diddling gloves and get real close with RT and java

17:46 broquaint: Thanks, hiredman :)

17:47 Aha, it's elisp, time to write me a keybinding.

17:50 hiredman: schleyfox: you'll need to gen-class most likely

17:51 broquaint: I think that gist may be missing bits like the favorite-libraries and get′

17:52 broquaint: https://gist.github.com/e6a92b792d43775ad615

17:57 schleyfox: hiredman: thanks for the tips, I think I'm just going to extend the class in java to take a provided IFn and invoke it

17:57 TimMc: MenTaLguY: It does seem kind of weird, doesn't it?

17:57 MenTaLguY: yeah, I'm actually trying the example now without the helper fn, and it seems to work fine without any extra memory usage

17:58 TimMc: MenTaLguY: Actually... how do you write it without the helper?

17:58 broquaint: Thanks again, hiredman :)

17:58 MenTaLguY: (defn myfilter [pred coll]

17:58 (lazy-seq

17:58 (if (pred (first coll))

17:58 (cons (first coll) (filter pred (rest coll)))

17:58 (filter pred (rest coll)))))

17:59 ideally you'd really want to call seq on coll and then save that in a local

17:59 TimMc: Ah, right. I've seen that style all over the place.

18:01 lpetit: MenTaLguY: when does the iteration stop in your definition ? ^^ :)

18:01 phil_: is assoc-in copy-on-write in cljs?

18:02 aclj: hi =D

18:02 phil_: lpetit: i assume (pred) returns nil when (first coll) returns nil

18:02 MenTaLguY: lpetit: oh, whoops, I forgot a (when)

18:02 yeah, but pred returning nil won't terminate iteration

18:02 dnolen: phil_: everything except cons'ing on a seq is copy-on-write

18:02 in cljs

18:03 lpetit: MenTaLguY: a when-let will also store (first coll)

18:03 phil_: MenTaLguY: if(nil) is false i think?

18:03 aclj: anyone tried to create libs in clojure and import them in android ? how is the performance ?

18:03 phil_: dnolen: i mean "real" copy-on-write, like vectors etc, i.e. without data reuse

18:04 dnolen: phil_: maps, sets, vectors all copy-on-write

18:05 technomancy: aclj: not very good

18:05 phil_: defrecords? or are they just maps?

18:05 lpetit: phil_: yes, any data structure provided by clojure, by default (that is, unless explicitly specified otherwise in the doc)

18:06 alg: technomancy: i mean, not develop the entire app in clojure, but only a set of lib to import in the java code of the app

18:06 technomancy: alg: you can't really trim clojure code down below a certain level; it brings the whole runtime with it

18:07 phil_: lpetit: im not talking about persistence, the problem is that cljs copies the WHOLE vector / map etc whenever something changes, as opposed to the normal clojure data structures where only a small part is created from scratch and everything else is reused

18:07 lpetit: phil_: oh indeed, (blush)

18:08 phil_: :)

18:08 so im outta luck unless i patch em myself :(

18:09 TimMc: phil_: a.k.a. contribute a patch

18:09 MenTaLguY: yeah, I seem to be having a really hard time provoking the space leak that clojure.org/lazy warns about, even without using a helper fn

18:10 (myfilter #(= % 20) (map inc (range 10000000000))) has been running for a couple minutes now, and the resident size and heap size are still flatlined

18:10 phil_: TimMc: yea, im planning on doing so anyway, just not yet... i heard there is some contributor agreement or something in place?

18:10 MenTaLguY: (more or less)

18:11 alg: teachnomancy: that's a pity, thanks

18:11 technomancy: hah; teachnomancy... nice.

18:15 TimMc: technomancy: I suppose resource-constrained platforms like Android are an argument against the uberjar distribution model.

18:16 technomancy: TimMc: I dunno, Android is already absolute crap at code-sharing

18:16 TimMc: Although... maybe two Clojure apps *wouldn't* be able to share a Clojure instance. I don't know enough about JVMs and classloaders.

18:16 Ah, OK.

18:16 technomancy: dead-code elimination will likely be much more beneficial

18:16 TimMc: Interesting.

18:17 klutometis: Anyone have any idea how I'd quote caret? '^ doesn't work: the reader thinks I'm talking about metadata.

18:18 Chousuke: caret? as in the character?

18:18 \^

18:19 klutometis: Chousuke: I'm trying to use it as a symbol, not a character.

18:19 Chousuke: you can't.

18:19 brehaut: (symbol "^")

18:19 klutometis: I guess (symbol "^") suffices, but it's ugly.

18:19 brehaut: Thanks.

18:19 Chousuke: it's an illegal symbol though

18:19 TimMc: klutometis: Ugly things shold look ugly...

18:20 Chousuke: so you really shouldn't :P

18:20 klutometis: TimMc: Oh, Zeus; really? In Scheme, no character is ugly.

18:21 TimMc: klutometis: Shceme uses symbols differently.

18:21 Chousuke: what do you need a caret symbol for anyway?

18:21 klutometis: I'm defining a regex-like DSL, by the way, where ^ signifies ordered subsets; it's a beatiful pun, I think.

18:21 TimMc: klutometis: (^ 4 5) is gonna fail.

18:21 Chousuke: yeah, you can't use ^ for your own purposes

18:21 it's a reader macro

18:22 klutometis: TimMc: You're right; that sucks. Ok: maybe `!', then. Though I've got to find something else for `not'. Maybe `~'.

18:22 Thanks.

18:22 amalloy: ~ won't make you very happy either

18:22 clojurebot: Huh?

18:23 amalloy: for similar reasons

18:23 TimMc: klutometis: Use some non-ASCII. :-P

18:23 Chousuke: well, ~ is not a reader macro

18:23 though I'm not sure if it works outside `

18:23 ,~

18:23 TimMc: &(let [~ 4] ~)

18:23 lazybot: java.lang.RuntimeException: Unmatched delimiter: )

18:23 hiredman: well it is, but it isn't

18:23 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.RuntimeException: EOF while reading character>

18:23 Chousuke: hmm, I suppose

18:23 hiredman: like '

18:24 brehaut: how about (💩 4 5)

18:24 Chousuke: so ~ is a reader macro outside syntax-quote but doesn't work :P

18:25 TimMc: &(let [~~ 4] ~~)

18:25 lazybot: java.lang.RuntimeException: Unmatched delimiter: )

18:25 amalloy: jeez, brehaut, you should hang out with us on the BMP. my font doesn't have a glyph for that one :P

18:25 brehaut: amalloy: BMP ?

18:25 TimMc: Oh, I thought he was suggesting question-mark.

18:25 Basic Multilingual Plane

18:25 brehaut: nope: unicode pile of poo

18:26 sorry

18:26 TimMc: Pick stuff from the Emoticons block: http://www.fileformat.info/info/unicode/block/emoticons/images.htm

18:27 amalloy: i only vaguely know what the basic multilingual plane actually means, but since my client rendered that as a three-byte pile of soggy mess, i think it must be outside the two-byte BMP area

18:28 TimMc: http://www.fileformat.info/info/unicode/char/1f4a9/index.htm

18:28 Haha, the spec actually says "poo".

18:28 * TimMc is an eigth-grade boy

18:28 brehaut: frighteningly on the mac its rendered with a little colored image and it has eyes

18:30 TimMc: haha

18:30 It's from a block that is mostly (all?) from Japanese charsets

18:30 brehaut: the emoji set?

18:30 TimMc: Hmm, perhaps not...

18:31 "U+1F46F WOMAN WITH BUNNY EARS" <-- well, maybe

18:32 brehaut: haha

18:32 TimMc: There's stuff in there like "blue heart", "green heart"... is this supposed to include color? o.O

18:34 amalloy: TimMc: you didn't know? unicode uses three bytes to encode RGB color, so the BMP is just all the different colors for "squarish blob"

18:38 TimMc: haha

19:08 solussd: long shot, but does anyone have any idea why my program would blow up when it tries to add a string containing a ≪ to a collection? It works fine in the repl but the same code blows up when executed from a jar

19:17 stevesj76: debating on the use of futures vs agents when we don't care about the return value. any advice?

19:17 (sending email in the background)

19:21 brehaut: futures would be my pick because its simpler

19:21 they both use the thread pool under the hood i believe

19:21 stevesj76: does that mean that the thread pool works like send for agents?

19:22 brehaut: yeah i believe so

19:23 hiredman: no, send uses a bounded threadpool

19:23 stevesj76: more like send-off then?

19:23 hiredman: send-off and futures use a theoretically unbounded pool

19:23 stevesj76: awesome thanks

19:23 hiredman: yes, future uses the same threadpool as send-off

19:32 dgrnbrg: Are records redefinable within a repl, or are they not conducive to repl debugging?

19:32 brehaut: you can redefine them

19:35 technomancy: you can redefine them, but not retroactively IIRC

19:35 brehaut: existing instances stay with the old implementation right?

19:35 technomancy: I think so; instances of the old record definition won't pick up the changes even though they will claim to be of the same class

19:35 or rather, they will claim to be of a class which is visually indistinguishable from the current one, but actually different

19:36 brehaut: sure. that makes sense

19:37 thats going to result in potential funny interactions with protocols right?

19:37 dgrnbrg: brehaut, technomancy: thanks

19:37 I'm currently using protocols and records and I'm getting lots of class already defined exceptions from the repl

19:38 I just want to develop rapidly, and get away from the need to constantly restart the jvm

19:39 man, there are weird rules for what goes inside (comment) blocks

19:39 No trailing colon, no number followed by 2 asterisks

19:40 hiredman: dgrnbrg: it has to be a readable form

19:40 dgrnbrg: Are those bugs or quirks?

19:40 hiredman: (comment ...) is a just a macro

19:40 so the body has to pass through the reader

19:40 dgrnbrg: Is there a way I can make it shut up?

19:40 hiredman: use ; or #_

19:40 dgrnbrg: just totally disable reader input

19:40 clojurebot: Titim gan éirí ort.

19:40 dgrnbrg: ; would be as effective at that?

19:41 awesome, I can prefix (comment...) lines with ; that have special characters

19:41 brehaut: dgrnbrg: comment is not magic: its roughly just (defmacro comment [& body] nil)

19:42 ; is just like a line terminating comment in any other language

19:42 dgrnbrg: so ; is handled by the reader, while (comment) isn't?

19:42 comment being a macro

19:43 brehaut: correct

19:43 #_ is also handled by the reader, but it behaves different (its still form oriented)

19:43 dgrnbrg: brehaut, when do I want #_ vs. ;?

19:44 brehaut: where comment results in a nil, #_ skips the form entirely

19:44 dgrnbrg: ah, i see

19:44 brehaut: ,(if #_true false :a :b)

19:44 dgrnbrg: so comment is an argument to a function potentially?

19:44 clojurebot: :b

19:44 brehaut: ,(if true false :a :b)

19:44 clojurebot: #<CompilerException java.lang.RuntimeException: Too many arguments to if, compiling:(NO_SOURCE_PATH:0)>

19:44 brehaut: ,(if (comment true) false :a :b)

19:44 clojurebot: #<CompilerException java.lang.RuntimeException: Too many arguments to if, compiling:(NO_SOURCE_PATH:0)>

19:44 brehaut: exactly

19:44 dgrnbrg: ahah

19:45 very interesting

19:45 I really like clojure

19:45 I'm using it to write an fpga programming language

19:51 brehaut: dgrnbrg: clojure is a likable language ;)

19:53 TimMc: solussd: Can you give me a minimal testcase with the string?

19:58 I'd like to compare simple version numbers (3, 1.2.0, etc.) -- anyone have a thought on that?

20:00 brehaut: ,(compare [3] [1 2 0])

20:00 clojurebot: -1

20:00 TimMc: ooh

20:00 brehaut: :)

20:00 TimMc: brehaut: How do I drop trailing zeroes?

20:01 I want 1.2 == 1.2.0

20:01 I guess I could preprocess with a regex. -.-

20:01 brehaut: ,(take-while (complement zero?) [1 2 0])

20:01 clojurebot: (1 2)

20:01 technomancy: TimMc: there's a simplistic implementation in leiningen.core.main/version-satisfies?

20:01 TimMc: ,(take-while (complement zero?) [0 0 3 0 1 2 0])

20:01 clojurebot: ()

20:01 brehaut: rats

20:01 TimMc: It can't be done in the standard way.

20:02 I mean, in a take/drop way

20:02 With two reverses it becomes simple.

20:02 brehaut: ah yes.

20:02 theres a name for that sort of operation

20:02 but i cant remember it

20:03 TimMc: drop-last is close...

20:04 technomancy: Oh, is that on master?

20:04 brehaut: TimMc: i'd just do the reverse trice

20:05 technomancy: TimMc: aye, in the leiningen-core directory

20:05 TimMc: brehaut: 3 times seems a bit excessive, no? :-P

20:05 brehaut: TimMc: conjugate is the name of that op

20:06 solussd: TimMc: I've figured out that calling out to libc/stat with a file-path that contains non-ascii characters doesn't work. Unrelated to clojure in every way. thanks though. :)

20:07 TimMc: solussd: Heh, OK.

20:10 technomancy: so... should throwing an exception be treated as a side-effect (something you should do before your return value in a when) or placed inside an if next to the return value?

20:10 I've always done the former but haven't thought about it much

20:11 TimMc: "when" screams "side-effect". It also simplifies the visual flow of control.

20:12 brehaut: ,((fn [a b] (let [s->v (fn [vs] (vec (reverse (drop-while zero? (reverse (map #(Integer. %) (.split vs "[.]")))))))] (compare (s->v a) (s->v b))) "" "0.1.3")

20:12 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.RuntimeException: EOF while reading>

20:13 brehaut: ,((fn [a b] (let [s->v (fn [vs] (vec (reverse (drop-while zero? (reverse (map #(Integer. %) (.split vs "[.]")))))))] (compare (s->v a) (s->v b)))) "0.1.3" "")

20:13 clojurebot: 0

20:14 brehaut: technomancy: i'd treat it as a side effect i think?

20:15 TimMc: technomancy: version-satisfies? uses #"\d" instead of #"\d+" -- probably a bug.

20:15 technomancy: TimMc: oy; yes

20:15 brehaut: yeah, I'm starting to think the only people who'd consider the alternative reasonable would be folks who have been damaged by a runtime that doesn't support exceptions. =)

20:15 brehaut: technomancy: lol :)

20:16 is there a runtime that doesnt support exceptions?

20:16 TimMc: brehaut: Ah, I like [.] instead of \\.

20:16 brehaut: the robot eye operator

20:16 technomancy: brehaut: I mean returning integer error codes

20:16 C

20:16 brehaut: technomancy: aha right

20:17 TimMc: brehaut: I'll end up using \d+ after seeing technomancy's impl.

20:17 brehaut: technomancy: some haskell programmers would suggest that exceptions form an implicit sum type and that you can use monads to control the flow without having to do a violent stack unwind

20:17 technomancy: I believe them; you can do anything with monads apparently.

20:17 brehaut: (eg using an Either type)

20:18 hiredman: except the stack unwind is faster

20:18 (on the jvm)

20:18 brehaut: hiredman: sshh!

20:18 schleyfox: that just sounds painful and academic

20:18 technomancy: didn't realize they could be used to curb violence though

20:18 hiredman: technomancy: in the same way any weapon can be used to curb violence

20:18 brehaut: technomancy: only on the stack

20:18 tavis: brehaut: something's not Right

20:18 dgrnbrg: How do I set the hierarchy that a multimethod uses? I tried writing (defmulti docstring dispatch-fn :hierarchy myhier) but it failed with PersistantArrayMap cannot be cast to IRef

20:19 but if I remove ":hierarchy myhier" it works

20:19 brehaut: tavis: wait. i thought right was the error path

20:19 tavis: depends, are you left handed?

20:20 technomancy: TimMc: well here's your chance for a sticker

20:20 tavis: Left is usually the error

20:20 brehaut: tavis: so it is; apparently ive just always visuallized the type backwards in my head :P

20:21 TimMc: technomancy: I think I already qualify.

20:21 technomancy: oh, nice

20:22 tavis: brehaut: well, things are upside down in your part of the world ;)

20:22 brehaut: thats true enough

20:22 tavis: I lived in Dunedin for 5 years, btw

20:22 brehaut: oh right

20:22 tavis: was in Hamilton last year

20:22 brehaut: university?

20:22 tavis: yep

20:22 brehaut: huh

20:22 im in hamilton this year

20:23 tavis: figured

20:23 technomancy: wow, NZ is really milking this LOTR thing for all it's worth; naming a city after the Dunedain.

20:23 brehaut: lol

20:23 amalloy: technomancy: fwiw, i prefer (... (if (broken?) (throw...) (work))) to (... (when (broken?) (throw...)) (work))

20:23 technomancy: amalloy: exprain?

20:25 amalloy: it feels to me like two directions the computation could go: throw an exception, or work. the when-variant reads more like: "i'm going to do work, but first i may need to do this other thing as a side effect to make it work"

20:25 technomancy: you mean like ... pre-conditions? =)

20:26 actually I should just use real preconditions for this

20:26 derp

20:26 solussd: hmm. shouldn't I be able to create a NativeString like this: (com.sun.jna.NativeString. "myString" true), I get an exception: [Thrown class java.lang.IllegalAccessError] any ideas?

20:26 TimMc: technomancy: I'll send a pull request, mostly because I need the practice.

20:26 * amalloy notches another win for his approach of "argue about unimportant things until other party solves problem independently"

20:27 hiredman: solussd: my guess is nativestring's constructor is private

20:27 solussd: hiredman: hmm. docs say public: http://jna.java.net/javadoc/com/sun/jna/NativeString.html

20:28 hiredman: what does the rest of the stacktrace say?

20:29 TimMc: technomancy: Umm... should I be able to run lein 2 using lein 1.6.2?

20:29 technomancy: TimMc: you should be able to operate on leiningen-core using lein1

20:29 for the top-level project use the bin/lein script

20:29 TimMc: got it, thanks

20:31 technomancy: Do you prefer people create feature branches, or just work on the target branch?

20:31 solussd: hiredman: ah ha! com.sun.jna.Native/toCharArray is what I needed. some reason, despite the docs, NativeString is not public

20:32 technomancy: TimMc: depends on the work. your regex tweak can go on master. =)

20:33 TimMc: Conditional on complexity, got it.

20:38 dgrnbrg: Can anyone show me code that uses defmulti with the :hierarchy optional argument?

20:38 I tried (defmulti name #(stuff %) :hierarchy my-hierarchy) but it doesn't compile

20:40 amalloy: you need to give it a ref of a hierarchy, so that it can be modified when you (derive) something

20:41 &@#'clojure.core/global-hierarchy

20:41 lazybot: ⇒ {:parents {}, :descendants {}, :ancestors {}}

20:41 dgrnbrg: amalloy, could I do #'my-hierarchy if I def my hierarchy as a global var?

20:42 technomancy: amalloy: language!

20:42 amalloy: probably. i dunno, try it

20:42 dgrnbrg: so that's not a standard pattern?

20:42 I tried it, it works

20:43 amalloy: setting up hierarchies for multimethods is not very common

20:43 TimMc: It's not deprecated, though.

20:44 dgrnbrg: ok, i see

20:44 what's the function like use for namespaces that are already loaded?

20:45 TimMc: dgrnbrg: refer

20:45 dgrnbrg: thanks

20:45 technomancy: you can just keep doing use though

20:46 dgrnbrg: Is a hierarchy mutable or immutable?

20:46 Oh, I see

20:46 I'm supposed to def an atom for the hierarchy

20:46 maybe?

20:46 does derive have side effects?

20:47 derive doesn't have side effects

20:47 it seems like that should be documented

20:53 TimMc: dgrnbrg: Of course it does, it changes the hierarchy!

20:54 tavis: amalloy: I think lazybot just insulted your parents, your ancestors and descendants as a retort to &@#'

20:59 TimMc: Interesting design decision there.

20:59 They could have written it so that you had to swap! and explicitly refer to a hierarchy.

21:00 I suppose derive and underive are safely repeatable, though.

21:25 TheBusby: Q: Anyone have any experience effeciently serializing seq's of Clojure data structures?

21:26 brehaut: yes someone is bound to

21:26 TheBusby: currently using a combination of pr-str and java.util.zip , but having issues with custom dictionaries for dealing with longs seq's of maps

21:26 brehaut: why?

21:26 clojurebot: Why is the ram gone is <reply>I blame UTF-16. http://www.tumblr.com/tagged/but-why-is-the-ram-gone

21:27 brehaut: ~botsnack

21:27 clojurebot: thanks; that was delicious. (nom nom nom)

21:27 TheBusby: working with rather large amounts of data, like 100M records like {:name "bob" :age 44}

21:27 brehaut: TheBusby: building strings rather than using a string builder (or printing to *out*) is inefficient

21:27 TheBusby: it all gets converted to byte[] in the end

21:28 brehaut: TheBusby: what do you mean by " issues with custom dictionaries"

21:29 TheBusby: java.util.zip uses huffman encoding to reduce the space needed for common symbols

21:30 if I compress all 100M records at once, everything works fine

21:30 but if I want to compress a record at a time, you need to provide a "dictionary" (java API term)

21:30 so the compression routine knows what is common and what isn't

21:31 same even if ":name" only appears once in each string, the fact that it appears in 100M strings means it's a good candidate for reduction

21:31 brehaut: TheBusby: what is the exact class you are using?

21:31 TheBusby: anyways, this was just the road I was currently going down. Curious if anyone else had encountered similar issues and had other solutions

21:31 brehaut: class I'm using for compression? java.util.zip.deflate

21:32 brehaut: should I be looking at a different compression library?

21:32 brehaut: TheBusby: if you use an output stream i think you will have more luck

21:33 TheBusby: brehaut: that would be compressing all of the records at once though wouldn't it?

21:33 brehaut: TheBusby: i havent done any file IO in a while, but i recall that output streams can be stacked. *out* is an OutputStreamWriter of some description

21:33 TheBusby: pass

21:33 TheBusby: brehaut: unfortunately I need 100M independent compressed byte[] :(

21:34 brehaut: does it have to be defalte ?

21:35 TheBusby: brehaut: no another library would work, but I need fast random access of the 100M records/elements

21:35 brehaut: TheBusby: right. zip will let you have random access (each map becomes a file)

21:36 TheBusby: and gzip uses adaptive compression so it doesnt need to know the entire contents at once

21:36 TheBusby: brehaut: Just FYI, each map becomes a byte[], which is stored in RAM

21:37 isn't gzip ~= java.util.zip.Deflator?

21:37 brehaut: hmm. it might be?

21:38 im out of my depth sorry

21:39 TheBusby: brehaut: any other ideas of how to store a large number of maps in RAM?

21:39 brehaut: TheBusby: i'd store them in a DB of some kind, perhaps with a cache in front of it

21:40 from the sales pitch, redis sounds like it might be a good choice

21:40 TheBusby: most DB's have a significant overheard for each record

21:40 brehaut: they do

21:41 TheBusby: I looked a redis for something else a year or two back and it had an overhead of 10-20B per record

21:41 thank you for the advice though, I'll double check to see if they've improved things! ;)

21:41 schleyfox: TheBusby: to be fair, so do most languages

21:42 brehaut: TheBusby: the reason i would use a DB is that they have already put the hard work into managing memory and caching etc

21:42 TheBusby: if you store the whole data set in memory though, typical (non-CPU) caching isn't effective

21:43 brehaut: if you have to store the whole dataset in memory and are concerned about cpu though, compressing each record is going to eat a lot of CPU too though

21:43 also, if its 100MB and i knew it wasnt going to grow fast, i'd just put the whole lot into a map

21:43 map of maps that is

21:43 TheBusby: decompression is fairly fast though

21:43 brehaut: and eat the ram

21:43 clojurebot: No entiendo

21:44 TheBusby: faster than RAM access in some cases

21:44 brehaut: TheBusby: my preference is always for a simple solution first though

21:45 TheBusby: brehaut: I'm a firm believer in KISS as well. Unfortunately that isn't really suitable here. KISS in my case would probably be a Hadoop cluster... :(

21:45 brehaut: TheBusby: out of curiosity, what are you trying to solve?

21:47 TheBusby: brehaut: Building graphs to describe entities and then match them together from multiple data sets

21:47 brehaut: TheBusby: do the entities have consistent keys and change infrequently?

21:48 TheBusby: brehaut: no fairly static

21:48 brehaut: TheBusby: this is hypothetical, but you might find that records use less ram than maps

21:48 TheBusby: brehaut: an example would be taking data sets about people and matching them together (not what I'm actually doing)

21:49 brehaut: records or structs would be great, but unfortunately this is fairly schema-less :(

21:49 brehaut: oh :(

21:49 TheBusby: brehaut: using the people example, everyone has a name but not everyone has a dog

21:57 brehaut: thanks for the input, I'll see if I can put up a gist or a brief blog when I get something going. ;)

21:57 brehaut: TheBusby: id be interested to see that

21:59 schleyfox: TheBusby: I know nathanmarz is writing a book basically just about that

22:00 TheBusby: schleyfox: I hope to catch him on IRC or at the Conj then. ;)

22:01 appropriate link to book: http://www.manning.com/marz/

22:01 schleyfox: I believe that's the one

22:02 TheBusby: schleyfox: thank you for pointing me to that. I may have to get the MEAP

22:03 schleyfox: it's still pretty early stage, but the second chapter does layout a graph type structure that may be along the lines of what you're thinking

22:03 TheBusby: at the moment though, I'll see if I can put together a little library for compressing a large seq

22:03 schleyfox: very hadoop/storm oriented though

22:03 nice

22:04 TheBusby: schleyfox: if my data set was 10x larger I'd be doing that, since I can fit everything into RAM though (< 100GB) I've found it's faster to keep everything on one box since in many cases the data sets are have links to each other everywhere

22:04 schleyfox: yeah, that makes sense

22:05 TheBusby: clojure's sequence library makes so many operations so convenient. I'd love to stick a compression/serialization library in the middle somewhere just to save on memory usage

22:06 TimMc: Man, is there any reason why ###{1 1 1} should throw?

22:06 schleyfox: that would be awesome

22:06 TimMc: &#{1 1 1}

22:06 lazybot: java.lang.IllegalArgumentException: Duplicate key: 1

22:06 dgrnbrg: TheBusby, fyi, you can add keys to records that you didn't initially declare on them and they'll go into an augmentation map thingy

22:07 TheBusby, so if you have some common data, they can sit in records, but you can use it just like a map otherwise

22:07 TheBusby: Ooooh that sounds very valuable!

22:07 phil_: TimMc: a hash set with duplicate keys doesnt make sense :/ or am i misunderstanding something?

22:08 TheBusby: dgrnbrg: Wouldn't I still need to serialize and compress though if I have many strings?

22:08 dgrnbrg: Yes, but for RAM utilization, you wouldn't need MapEntries for the common data

22:08 it depends on whether each record has 4 fixed keys and 100 variable keys

22:08 brehaut: TheBusby: is it possible to have keywords instead of strings? keywords are already interned for you so much lower ram usage

22:09 dgrnbrg: or 4 fixed keys and a handful of non-fixed

22:09 TheBusby: sorry, I meant strings as values not keys

22:09 dgrnbrg: strings are also interned, often

22:10 you can call String.intern() to get the interned version for certain

22:10 TimMc: phil_: Sets don't have instance count semantics, so it shouldn't matter if you specify something twice.

22:10 tmciver: ,(set [1 1 1])

22:10 TheBusby: going off to read about String.intern() be back in a bit. Thanks all!

22:10 TimMc: (element count, whatever -- I'm making up a phrase)

22:10 clojurebot: #{1}

22:10 apwalk: ,(into #{} [1 1])

22:10 clojurebot: #{1}

22:11 TimMc: ,(hash-set 1 1)

22:11 clojurebot: #<IllegalArgumentException java.lang.IllegalArgumentException: Duplicate key: 1>

22:11 phil_: ,(into #{1} [1])

22:11 clojurebot: #{1}

22:11 TimMc: (let [x ___] #{1 2 5 x 9}) ; might fail, who knows?

22:12 jeremyheiler: TimMc: looks like insertion is different that instantiation.

22:12 phil_: TimMc: yea, but (int #{} [1 2 5 x 9]) won't

22:12 into*

22:13 TimMc: Right, there are easy workarounds, but it shouldn't even be a thing you have to think of.

22:13 phil_: but it shouldnt bomb out when making a new instance

22:13 its unexpected

22:13 yea

22:13 jeremyheiler: it's interesting

22:14 apwalk: is the expected way to handle it, to wrap it in a try?

22:14 TimMc: ,(import clojure.lang.PersistentHashSet)

22:14 clojurebot: clojure.lang.PersistentHashSet

22:14 TimMc: apwalk: Nah, just create it a different way.

22:14 phil_: apwalk: that would be insane (if it was expected)

22:14 TimMc: ,(PersistentHashSet/create 1 1)

22:14 clojurebot: #<CompilerException java.lang.IllegalArgumentException: No matching method: create, compiling:(NO_SOURCE_PATH:0)>

22:15 TimMc: ah, that's a varargs...

22:15 ,(PersistentHashSet/create (to-array [1 1]))

22:15 clojurebot: #{1}

22:16 TimMc: ,(PersistentHashSet/createWithCheck (to-array [1 1]))

22:16 clojurebot: #<IllegalArgumentException java.lang.IllegalArgumentException: Duplicate key: 1>

22:16 phil_: the reader bombs out, not the class itself

22:16 (i think)

22:16 ah

22:16 TimMc: Nah, the reader is calling createWithCheck

22:16 phil_: ok :D

22:16 TimMc: but why?

22:17 phil_: maybe "fail fast" philosophy?

22:18 TimMc: But it's a set!

22:19 phil_: yea, somebody thought adding the same thing twice to a set might be an indication of a bug

22:19 i guess

22:19 brehaut: certainly for literals

22:20 although more so for maps than sets

22:20 dnolen: ,{1 2 1 2}

22:20 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.IllegalArgumentException: Duplicate key: 1>

22:20 TimMc: The semantics of (apply hash-set [x y z]) should be precisely identical to (into (hash-set) [x y z]).

22:20 jeremyheiler: phil_: the exception only goes off if the element failed to be added

22:20 brehaut: dnolen: i think (1 2, 1 3) is a better example

22:21 phil_: are defrecords hashmaps?

22:21 TimMc: My first thought was, "Oh, someone is trying to be clever and use the provided collection directly, and is checking it first to make sure it is safe." But that's not true, it's all conses under the surface.

22:22 dnolen: phil_: not quite, you have optimized access of declared keys, an internal extension map for those that aren't

22:23 brehaut: phil_: map? says yes

22:23 well, it says they are maps, but not necessarily hash maps

22:23 tmciver: ,#{[1 2] (1 2)}

22:23 clojurebot: #<ExecutionException java.util.concurrent.ExecutionException: java.lang.IllegalArgumentException: Duplicate key: (1 2)>

22:24 tmciver: Hmm, page 94 of JoC shows that to work.

22:24 phil_: any cljs specific differences?

22:24 brehaut: tmciver: that are in the same equality partition, so it shouldnt

22:25 tmciver: but you are right that it does show that

22:25 tmciver: "Given two elements evaluating as equal, a set will contain only one, independent of concrete types."

22:25 brehaut: it shouldn't what, fail?

22:25 brehaut: tmciver: joy of clojure is wrong

22:26 tmciver: But it's right here in black and white!

22:26 brehaut: it probably slipped through from an older clojure where the reader didnt check keys

22:26 TimMc: ~guards

22:26 clojurebot: SEIZE HIM!

22:26 TimMc: argh, bad timing

22:26 brehaut: 1.1 i believe

22:26 tmciver: slow connection? ;)

22:26 * brehaut has much lag

22:27 tmciver: I agree with TimMc, it seems like it shouldn't throw.

22:27 (to me)

22:27 brehaut: sets at runtime shouldnt, but set literals, the ones that are baked into your source should IMO

22:27 imagine you have

22:28 #{1 2 3 4 5 6 7 8 9 … 1}

22:28 later you realise you shouldnt have 1

22:28 so you remove it from the end of the set

22:28 not realising its there at that start too

22:28 thats just become an annoying bug that could have been avoided

22:29 dnolen: phil_: not beyond the slower performance of extending with non-specified keys

22:30 dgrnbrg: Where can I find a tutorial explaining how to use tests with leiningen? I seem to be doing things wrong in a very basic way.

22:30 phil_: dnolen: so the extension map is a hashmap and theres some other, more optimized map for declared keys?

22:31 amalloy: "should" is a pretty strong word in this context, brehaut. i'd phrase it more like "given my cursory exploration of the topic, i'd be happier if ___". if sets behave the other way, the bugs that sneak in aren't eliminated; they're just replaced by different ones

22:31 dnolen: phil_: no map, the host type

22:32 amalloy: phil_: conceptually it's a more optimized map, yes: a java class with declared instance fields

22:32 phil_: dnolen: but there are no host types in cljs

22:33 dnolen: phil_: think a bit more about that :)

22:34 phil_: dnolen: well, everything is a hashmap in js

22:34 tmciver: brehaut: I think I agree with you. literals are called that for a reason, after all. The behavior that I was expecting is achieved with the set function.

22:34 phil_: dnolen: or do you mean the "optimized" hosttype is a js hashmap while the extension map is a cljs hashmap?

22:34 brehaut: phil_: everything is a hashmap with an associated constructor function

22:35 phil_: if there were no types in js,then the instanceof operator would always evaluate to false

22:35 phil_: brehaut: yes, but why is the ctor optimizing access to the keys?

22:37 dnolen: phil_: if JS objects were really hashmaps, JS would still be dog slow.

22:37 phil_: i think i understand

22:38 dnolen: phil_: I think it's better to think of JS Objects like Self objects, you can create slots on the fly.

22:38 phil_: the fact that JS Objects inherits slots from prototype makes the hashmap analogy even more weak

22:39 brehaut: phil_: if you can portion off ad hoc properties in a javascript object into their own object and leave a consistent set of properties for the main type (ie, like a record does) then you maintain the 'shape' that the runtime inferencer can see, so it doesnt need to keep generating underlying types

22:39 (because generating new underlying types also throws guards onto slow paths and causes rejitting to occur)

22:40 phil_: so defrecord is creating a js prototype underneath?

22:40 dnolen: phil_: deftype/defrecord both create constructors functions

22:41 phil_: when you extend a type/record to a protocol, this is done via the constructor fn's prototype property

22:41 phil_: dnolen: ok, this clarifies things

22:43 dgrnbrg: Is there a function to check if 2 seqs are equal, or do I do a map/reduce?

22:43 dnolen: dgrnbrg: just use =

22:43 dgrnbrg: that'll work for seqs

22:43 dnolen: ,(= '(1 2 3) '(1 2 3))

22:43 dgrnbrg: ?

22:43 clojurebot: true

22:43 dgrnbrg: ,(= '(1 2 3) [1 2 3])

22:43 clojurebot: true

22:43 dgrnbrg: oh

22:44 adam__: has anyone seen, https://github.com/globulon/sicp-in-clojure ?

22:44 i was wondering if it maps to SICP well, or if its imcomplete

22:45 dnolen: adam__: I've said it once and I'll say it again. I love SICP. It's a terrible way to learn Clojure.

22:45 adam__: okay, I have some experience with clojure, and noticed clojure does a lot of the things that SICP teaches

22:45 like lazy sequences

22:46 thanks

22:55 JanxSpirit: hey I imagine you guys are familiar with 4clojure?

22:55 I'm doing this one: http://www.4clojure.com/problem/21#prob-title

22:56 I came up with a function that does it

22:56 but I'm just not sure what the problem wants in terms of input

22:56 it won't let me do a (defn ...)

22:57 tmciver: JanxSpirit: you can do a (fn [] ... or a #() instead.

22:57 JanxSpirit: OK - I probably need to go look this up but what's the difference fn/defn?

22:58 tmciver: defn is a macro that def's a var, which you can't do in the 4clojure solutions

22:58 fn returns a funtion object which can be called directly.

22:59 JanxSpirit: hmm...OK - thanks!

22:59 tmciver: ,(doc defn)

22:59 clojurebot: "([name doc-string? attr-map? [params*] body] [name doc-string? attr-map? ([params*] body) + ...]); Same as (def name (fn [params* ] exprs*)) or (def name (fn ([params* ] exprs*)+)) with any doc-string or attrs added to the var metadata"

22:59 tmciver: sure

23:01 brehaut: (comp first (partial apply drop) (comp reverse vector))

23:01 worst nth ever

23:01 (comp first (partial apply drop) rseq vector) ; slightly better

23:05 tmciver: brehaut: ha! you haven't seen my solution!

23:05 brehaut: ,((comp first (partial apply drop) (juxt second first) vector) [:a :b :c :d :e] 3)

23:05 clojurebot: :d

23:05 brehaut: thats the winning solution there

23:05 tmciver: i have not :)

23:06 tmciver: it's better that way.

23:06 brehaut: everything is better with juxt ;)

23:09 phil_: ,((comp first (partial apply drop) (juxt second first) vector) [:a :b :c :P :e] 3)

23:09 clojurebot: :P

23:09 tmciver: That's a cool solition. It's taken me this long to understand it.

23:09 s/solition/solution

23:09 brehaut: its the too much haskell coolaid solution

23:09 tmciver: ha!

23:10 brehaut: (which is to say i know enough haskell to be dangerous but not enough to be useful)

23:10 adam__: yeah, i learned enough haskell to be able to understand books on functional programming written using haskell

23:11 brehaut: that amount is worth the effort

23:11 adam__: there is a lot of great material written on FP that uses haskell from a pretty large academic community

23:12 not all of it applies though, but a lot does

23:12 phil_: yea but they lose me when they start using transformers on monad transformers :)

23:12 brehaut: haha

23:12 monad transformers are much easier to grasp in clojure

23:12 adam__: ha, learning monads is on my news years resolution list

23:12 brehaut: less type system syntax and magic to wrap your head around

23:13 adam__: i will tackle concurrency first, then go into monads

23:14 phil_: ive looked at clojure monads only superficially, but i have a gut feeling ill have t o adjust to working with monads without explicit types

23:14 (not that ive done so much with monads)

23:14 JanxSpirit: what languages do most Clojure users come from?

23:15 adam__: wasn't there a survey going around a couple weeks ago?

23:18 dnolen: http://cemerick.com/2011/07/11/results-of-the-2011-state-of-clojure-survey/

23:18 last survey - mostly Java / Python / Ruby

23:18 adam__: i think its a good thing that most people are coming from imperative langauges

23:19 no way clojure could bloom if it stayed in the lisp community

23:19 fbru02_: hey all ! how do i do a get-in to look for a key but i don't know the intermidiate keys to get there ?

23:19 dnolen: adam__: most people in the older lisp communities either don't care or are hostile

23:19 pdk: lisp world can only grow so much from within itself

23:20 also cl guys tend to hate clojure

23:20 JanxSpirit: really? old school Lispers don't like Clojure?

23:20 phil_: didnt think so much java programmers are so open minded tbh :)

23:20 dnolen: but cl guys tend to hate everyone even cl people

23:20 JanxSpirit: I come from Java/Scala

23:20 my first Lisp and it's very interesting - good exercise for the brain :)

23:21 adam__: yeah, it really has opened my mind to what it is I am doing when I write a program

23:21 phil_: JanxSpirit: wait till you come to haskell, THEN youll have an excercise for your brain :)

23:22 JanxSpirit: I read a chapter or so of my Haskell book, but it was pretty intimidating to jump into

23:22 adam__: in Haskell, implemeting state can be a phd level topic

23:22 JanxSpirit: got Clojure doing some things right out of the gate which has been fun

23:23 phil_: adam__: yea, you cant really do anything in haskell besides small 3 liner teasers without having understood monads at the very least

23:23 schleyfox: I did a lot of stuff in erlang, Clojure is nice because compiling and starting code using a few dependencies isn't a week long adventure during which I feel like the Swedish Alexander Graham Bell

23:23 brehaut: i disagree http://haskell.org/haskellwiki/Simple_unix_tools

23:24 phil_: schleyfox: rly? i found erlang very easy to set up and jump into

23:24 schleyfox: phil_: oh it is. then try to use a couple libs outside of the standard distribution

23:24 fbru02_: related : what's the best way to walk over a hash?

23:25 jeremyheiler: fbru02_: What do you need to look at?

23:25 phil_: schleyfox: i was only using mochiweb and it wasnt such a problem to set up but cant talk about more complex project setups

23:26 schleyfox: yeah, and rebar has helped a lot

23:26 phil_: yea rebar was awesome

23:26 fbru02_: jeremyheiler: i have a deep nested hash and need to find a specific key that should be at the any of those sub-hashes of any given depth

23:26 schleyfox: but compared to leiningen or anything of that nature, it's basically the worst

23:27 metajack: erlang works the same as python and ruby. there's just a path to dependent libraries.

23:27 phil_: schleyfox: yea ive been very pleasantly surprised by clojures eco system

23:27 metajack: all rebar does it set it up in a nice default way and fetch the dependencies based on git urls

23:28 in a way, clojure seems weird to me because it deals with build artifacts as dependencies. haven't really dealt with that since i spent most of my time in C

23:29 schleyfox: metajack: yeah, I mean, it's not too bad, but it just always bites me when I try and distribute applications to servers and such

23:29 metajack: if you use releases, it's kind of like an expanded uberjar.

23:29 rebar now makes that easy, but otherwise it can be a pain, sure

23:29 even easier than java, since erlang releases include the entire runtime.

23:29 schleyfox: yeah, I mean I stopped doing much erlang about the time rebar started to actually get decent

23:30 metajack: riak is a good example of releasable erlang code. you just unzip and type riak start. no muss no fuss.

23:31 schleyfox: yeah, I agree

23:32 adam__: Erlang looks pretty interesting, but Clojure STM is alot better for me than the actor model

23:33 schleyfox: I feel like if erlang just had a bigger ecosystem, it would just be awesome

23:35 metajack: erlang could use data structure literals like clojure has :)

23:35 schleyfox: erlang could use a lot of syntactic/modernization type things that clojure has

23:37 phil_: adam__: but clojure STM aint distributed

23:38 there is a lib though

23:38 dnolen: damn I love logic programming

23:38 https://gist.github.com/1667847

23:38 phil_: https://github.com/technomancy/die-roboter

23:38 icylisper: schleyfox: There is a nascent project on an Erlang-Clojure bridge using Chas Emerick's nREPL protocol. https://github.com/icylisper/isthmus

23:39 schleyfox: icylisper: nice

23:39 I think the extent of my clojure-erlang bridging though these days will be reading and writing to riak

23:40 adam__: does anyone here have experience on using clojure's concurreny along with sun grid engine ?

23:43 nighty^: pyr! ping

23:48 brehaut: re:distributed stm http://avout.io/

23:48 adam__, phil_: ^

23:49 phil_: brehaut: this looks very interesting, bookmarking

23:49 adam__: brehaut, so this would allow me to distribute concurrent jobs across cluster nodes?

23:50 brehaut: adam__: i dont know the details sorry

23:50 adam__: thanks for showing me, i'll have to read more into it!

23:50 metajack: teh think relevance podcast has an interview with the author of avout

23:50 brehaut: avout is quite new

23:50 metajack: you might start there

23:50 brehaut: metajack: good point

23:59 jeremyheiler: fbru02_: some questions: is the key in the nested structure more than once? is the value of the key another map?

Logging service provided by n01se.net