#clojure log - Aug 15 2014

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

0:17 zkanfer: I have a ClojureScript question. I'm looking to get functions and constants on the Math object if I only have the name of the function or constant as a string. I can check if they exist with, e.g. (.hasOwnProperty js/Math "E"), but how do I get the constant or function that the string corresponds to?

0:25 TEttinger: zkanfer: http://stackoverflow.com/a/23440306/786740

0:25 it isn't supported but can be done

0:29 zkanfer: TEttinger: Thanks. So I need to somehow include proxy.js in my clojurescript code, then call (js/proxy (str "Math/" "E"))?

0:32 zkanfer_: TEttinger: Ugh, Firefox just crashed. If you responded to my last question (how to use the StackOverflow answer), I didn't see it.

0:35 TEttinger: zkanfer_, good ol' firefox crashes

0:35 I'm amazed at how often it crashes now

0:36 zkanfer_: Yeah, not even sure why -- I wasn't doing anything bad, I swear!

2:23 andyf: I am guessing there is a more elegant way to do this: (for [[idx x] (map-indexed (fn [idx x] [idx x]) some-sequence)] <do stuff with idx and x here>)

2:24 pyrtsa: andyf: At least you could use (map-indexed vector) there.

2:25 I mean (map-indexed vector some-sequence), of course.

2:25 andyf: pyrtsa: Yes, that is quite a bit better

2:25 pyrtsa: Defining (def enumerate (map-indexed vector)) could be useful.

2:26 andyf: You mean with partial before map-indexed?

2:26 pyrtsa: Bah, partial yes.

2:27 Should probably have another cup of coffee this morning. :)

2:27 andyf: np. You didn't have my senior moment of writing (fn [idx x] [idx x]) at least :)

2:27 pyrtsa: :)

2:28 andyf: ,(inc pyrtsa)

2:28 clojurebot: #<CompilerException java.lang.RuntimeException: Unable to resolve symbol: pyrtsa in this context, compiling:(NO_SOURCE_PATH:0:0)>

2:28 andyf: shoot, I don't use this stuff often enough to remember the right syntax

2:28 pyrtsa: :|

2:29 Without a leading comma, I guess.

2:29 andyf: (inc pyrtsa)

2:29 lazybot: ⇒ 7

2:29 pyrtsa: Gee, thanks!

2:32 andyf: Wow, we are up to 1332 accounts on Clojure JIRA (noticed only because of adding a count to my program that calculates weighted votes)

4:42 H4ns: s

5:04 michaelr525: hello

5:05 hyPiRion: hi

5:09 michaelr525: what's up in clojureland?

5:13 hyPiRion: not much

5:13 michaelr525: good

5:13 hyPiRion: silence this early, as all 'muricans are sleeping

5:13 michaelr525: usually there is too much so i just can't keep up with everything ;)

5:14 hyPiRion: hah

5:23 michaelr525: yogthos: hi

5:23 yogthos: your blog, does it also render on the server?

5:23 yogthos: for google and such..

5:47 Killombre: Hello max

6:36 ponpal: How can I evaluate (use 'clojure.repl) automatically everytime I use cider-jack-in in emacs?

6:39 milos_co1agen: perhaps check out http://thinkrelevance.com/blog/2013/06/04/clojure-workflow-reloaded section 'Dev profile and user.clj'?

6:40 luxbock: ponpal: you can add `:injections [(use 'clojure.repl)]` to your profiles.clj

6:40 this only makes it be present in the user namespace though

6:40 thesaskwatch: Hi, anyone has a good way to manage unbalanced parentheses when in paredit mode? Or a differerent mode/improvement to easily find/match brackets?

6:41 luxbock: if you'd like to have those functions available in all of your namespaces then you should check out vinyasa

6:41 thesaskwatch: (I mean in emacs)

6:41 ponpal: Thanks a lot, I'll check that out.

7:00 TimMc: TEttinger2: Get yourself onto the Firefox ESR (Extended Support Release) track and it's a lot more stable.

7:10 michaelr525: is there a log of the channel somewhere?

7:22 TimMc: ~log

7:22 clojurebot: see logs

7:22 TimMc: *sigh*

7:22 ~logs

7:22 clojurebot: logs is http://clojure-log.n01se.net/

7:22 hyPiRion: hah

7:22 TimMc: $log

7:22 $logs

7:22 http://lazybot.org/logs/#clojure/2014-08-15

7:25 Bronsa: michaelr525: http://logs.lazybot.org/irc.freenode.net/#clojure

7:46 michaelr525: Bronsa, TimMc: Thanks

7:46 I get this: These are not the logs you're looking for.

7:46 ah

7:46 http://lazybot.org/logs/#clojure/2014-08-15

7:47 ^^^^^^^^ this worked

8:02 phillord: hmmm -- I am confused by this

8:02 (into {} [[:a 1][:b 2]])

8:02 {:a 1, :b 2}

8:02 (into {} '([:a 1][:b 2]))

8:02 {:a 1, :b 2}

8:03 (into {} '((:a 1)(:b 2)))

8:03 ClassCastException clojure.lang.Keyword cannot be cast to java.util.Map$Entry clojure.lang.ATransientMap.conj (ATransientMap.java:44)

8:03 why does the last except?

8:04 hyPiRion: phillord: [a b] can be cast to a key entry, whereas '(a b)/(list a b) cannot

8:04 phillord: so, why not "clojure.lang.PersistantList cannot be cast to ...."

8:05 where does "Keyword" come into this?

8:10 hyPiRion: phillord: because (conj {} {:a :b :c :d}) == (conj {} (seq {:a :b :c :d})), confusingly enough

8:10 ,(conj {} (seq {:a :b :c :d}))

8:10 clojurebot: {:a :b, :c :d}

8:12 phillord: you are right

8:12 I am confused:-)

8:12 hyPiRion: yeah, it's confusing

8:13 ,(into {} [[(clojure.lang.MapEntry. :a :b)]])

8:13 clojurebot: #<IllegalArgumentException java.lang.IllegalArgumentException: Vector arg to map conj must be a pair>

8:13 hyPiRion: ,(into {} [(list (clojure.lang.MapEntry. :a :b))])

8:13 clojurebot: {:a :b}

8:13 * hyPiRion *shrugs*

8:16 phillord: does this count as a bug I wonder?

8:24 gfredericks: no, conj with maps is intentionally weird and confusing

8:24 I don't understand the intention but it's unlikely it was an accident

8:24 pyrtsa: phillord: Sounds like a question of whether the behaviour of (conj {} x) makes sense when x isn't a MapEntry nor a vector.

8:25 phillord: I should add it to my list of gotchas I guess

8:26 pyrtsa: The questionable code is this: https://github.com/clojure/clojure/blob/43cc1854508d655e58e377f84836ba128971f90c/src/jvm/clojure/lang/APersistentMap.java#L39-L45

8:26 gfredericks: merge is implemented with the questionable part of conj

8:27 pyrtsa: As well as: https://github.com/clojure/clojure/blob/43cc1854508d655e58e377f84836ba128971f90c/src/jvm/clojure/lang/ATransientMap.java#L41-L47

8:27 gfredericks: Thanks! Was about to ask next if anyone knew of code paths touching that piece of code.

9:35 TimMc: Bronsa: I see now that if (let [a 5] #{a 5}) did not fail due to the set constructor's check, then #{(gensym) (gensym)} would silently become a set of one element instead of throwing.

9:35 That would be pretty nasty.

9:36 Bronsa: TimMc: yeah

9:37 TimMc: I withdraw my objection on practical grounds. :-P

9:53 I guess set literals and code-as-data just don't mix all that well.

10:06 bobwilliams: anyone have experience with: clj-http.client?

10:06 dakrone: bobwilliams: yes a little ;)

10:07 bobwilliams: dakrone: haha!

10:07 i'm using your post to make a call to cleverbot's http://www.cleverbot.com/webservicemin

10:08 all is good but getting an error on invalid cookie header. MalformedCookieException Unable to parse expires attribute

10:08 is there a way to set this?

10:08 i'm a n00b so sure i'm just missing something simple

10:08 dakrone: bobwilliams: how are you sending the request? what options are you passing?

10:09 bobwilliams: :body <url encoded param string> :client {"http.useragent" "clj-http"}

10:09 i have :debug and :debug-body both set to true

10:10 that's all though

10:10 dakrone: so you're getting that exception when the cookie header is parsed then

10:10 bobwilliams: using requestmaker.com and/or httpclient.app it works for me

10:10 believe so

10:11 dakrone: bobwilliams: do you have a stacktrace with a line-number in clj-http?

10:11 bobwilliams: MalformedCookieException Unable to parse expires attribute: "Fri, 15 Aug 2014 15:07:08 GMT" org.apache.http.impl.cookie.BasicExpiresHandler.parse (BasicExpiresHandler.java:60)

10:43 MaxDH: Hi, I'm trying to use 'use-fixtures' in conjunction with test selectors to stop Travis from running certain tests that require database connections. However the build is failing due to these tests within use-fixtures. So I was wondering if there is a way to use test selectors and use-fixtures together?

10:43 Thanks

10:47 stompyj: does anyone here use clojure for “big data” purposes?

10:50 TimMc: stompyj: Define "big data". :-)

10:50 stompyj: I know

10:50 I hate even saying the word

10:51 but what I mean is, using clojure for working with spark, kafka, cassandra, etc

10:51 I’m about to build an event-based system

10:51 llasram: stompyj: Yes. We use Clojure on Hadoop MapReduce and HBase

10:51 I wrote https://github.com/damballa/parkour/ and use it daily on multi-TB datasets

10:51 stompyj: and I really want to use clojure, but scala seems to have a lot of momentum in the space

10:51 llasram: thanks, I’ll check this out

10:52 I know clojure is well supported w/r/t storm and hadoop

10:58 TimMc: stompyj: But seriously, what are your needs?

10:59 stompyj: TimMc: just kafka, storm and spark, to start

10:59 I guess I’m just asking a broader question, data transformation seems like a slam dunk for clojure

11:00 but there seems to be less emphasis on it, atm (which is fine)

11:30 llasram: stompyj: What do you mean by "less emphasis on it"?

11:32 ToxicFrog: "Wrong number of args (-1) passed to: core/filter"

11:32 um

11:32 stompyj: the amount of perceived activity and energy around that space seems less enthusiastic then in the core.async / clojurescript / web programming / other topics

11:32 llasram: ToxicFrog: Now that is a sweet error

11:33 stompyj: I think there's just fewer people doing distributed data-processing in general, and no big companies using the language for distributed data-processing the way e.g. Twitter is with Scala

11:33 ToxicFrog: llasram: I know!

11:33 stompyj: llasram: 100% agreed

11:34 thats why I came here to see if people are doing it, but not talking about it

11:34 because clojure actually seems a better language fit for this stuff then scala

11:34 and as I take this new gig, I’m going to have to think very hard about whether i should use scala or clojure

11:34 if*

11:34 ToxicFrog: The actual error is that I typed (filter (fn [[k v]] (keyword? k) xs)) rather than (filter (fn [[k v]] (keyword? k)) xs)

11:35 But what an error message

11:35 llasram: stompyj: I don't know Scala's collection libraries well enough to compare, but I do find Clojure's reducers abstraction (and probably transducers) to be well-suited to writing programs over distributed data

11:36 stompyj: llasram: yeah, exactly

11:36 Bronsa: ToxicFrog: there's already a ticket in JIRA for that

11:36 stompyj: we’re on the same page

11:36 most of the shops here in NYC are using Scala to do this kind of stuff atm

11:36 at least the ones I’ve talked to

11:37 llasram: stompyj: Well, give Parkour a try vs e.g. Scalding or SCrunch.

11:37 stompyj: llasram: I absolutely will! thanks!

11:37 llasram: Although if you wait a few days, I should have a new release out w/ some nice new features

11:38 Cool beans

11:38 stompyj: yeah, I don’t have th gig yet

11:38 haha

11:38 so I have time

11:43 lvh: is there a println that for some value that can be described as a literal, gives me the literal?

11:43 i.e. it will quote lists etc

11:43 llasram: lvh: prn?

11:44 ,(prn [1 #{2} [3]])

11:44 clojurebot: [1 #{2} [3]]\n

11:44 lvh: maybe :)

11:44 thanks!

11:44 mdrogalis: Hey stompyj. I specialize in building batch and stream processing systems with Clojure. Can answer questions if you need. :)

11:44 stompyj: mdrogalis: hey! I’ve read a bunch of your blog posts

11:44 great stuff

11:45 mdrogalis: stompyj: D'aw. Thanks. :)

11:45 stompyj: mdrogalis: there’s nothing specific, it’s more about hiring, and defending the usage of clojure over scala, from a business perspective

11:46 mdrogalis: stompyj: I've found that these problems end up being less about language and more about mechanism to be honest.

11:46 stompyj: mdrogalis: what do you mean?

11:47 mdrogalis: stompyj: It's a little hard for me to articulate, but when I say mechanism I mean something like map-reduce, or a queueing-type system.

11:48 stompyj: ohhh

11:48 yeah, I agree

11:48 mdrogalis: There's a lot of bindings to each of those, but once you've lock into one, you're kind of there, you know?

11:48 stompyj: yeah

11:49 but this is less about code execution and more me pitching a company on why I’m going to use clojure, and how this is more efficient, not necc. from a code perspective, but from a leaning on open-source, and hiring perspective

11:49 which is why I was looking to find more people working on these kinds of problems

11:49 mdrogalis: stompyj: Got it. Sounds like you're a consulting engineer, or battling upwards in your own company?

11:49 stompyj: as positive signaling that I won’tbe writing custom libs for everything

11:49 lvh: Is there any diference between (hash-map k0 v0 k1 v1) and {k0 v0 k1 v1}?

11:50 stompyj: mdrogalis: consulting engineer to start, with the idea of running data operations for them long term, building a team, etc

11:50 mdrogalis: stompyj: Cool :)

11:51 stompyj: mdrogalis: yeah, and we may have the time to allow me to go clojure, and build libs as I need them, but I want to be able to present truthful options to the potential client

11:52 mdrogalis: stompyj: Yeah, definitely depends on the problem. Aside from that, I can at least say that Clojure's tight number of types makes working with intermediaries extremely easy, since you never get far away from the basic data structures.

11:52 There's no boiler plate object transformation code to custom interfaces or something like that

11:52 boxed: man, the clojure reddit is so damn poisonous :(

11:53 stompyj: mdrogalis: exactly!

11:53 boxed: “I disagre with his point, DOWNVOTE”

11:53 stompyj: thats why I’d have thought lots of data aggregation / event streaming enthusiasts would be using clojure for these problems

11:54 between transducers, core.async, immutable collections, you have everything you need

11:55 mdrogalis: stompyj: Most people I meet who sell themselves as engineers for bigger data sets are Java enterprisey people. :

11:55 :/*

11:55 stompyj: mdrogalis: yeah, its deeply frustrating hahaahha

11:56 at this stage of my career I just want to solve big problems, I’m less concerned with language, so I’d use scala if I had to, however, I’m MUCH rather use clojure :D

11:56 I’d*

11:57 mdrogalis: Understandable.

11:58 whodidthis: what happened to all the cool presentation vidyas from euroclj/strangeloop

11:58 or did they happen (2014)

12:00 mdrogalis: stompyj: Is the client looking at a batch or event stream processing problem?

12:00 stompyj: mdrogalis: they don’t know, I’m pitching starting with an event stream

12:00 then as we collect data, and understand our needs better

12:01 bringing in something like spark for the batch piece

12:01 mdrogalis: stompyj: Pretty cool.

12:02 stompyj: mdrogalis: yeah! we’ll see if it happens, would be super exciting

12:02 mdrogalis: I'm going to be open sourcing a batch/streaming hybrid for Clojure in about a month. Hopefully it'll wake up some of the dormant Clojure space, kind of like you mentioned.

12:04 llasram: mdrogalis: Stand-alone, or built on top of existing tech?

12:05 mdrogalis: llasram: Its a standalone.

12:06 stompyj: mdrogalis: nice! I think this is a space that clojure could really dominate

12:07 if we can get some more wind behind the sails,

12:07 sounds like llasram and you are doing some great work on this already tho

12:07 I’m curious what nathan marz is up to as well

12:07 mdrogalis: He's been pretty quiet for a while.

12:08 stompyj: yeah

12:11 Major_Crixus: !rules

12:12 gf3: yogthos: hey cool secretary article 😏

12:14 arohner: I have an Om single-page app. Is goog.events/listen the preferred way to catch navigation events?

12:16 listen on an goog.history.Html5History

12:16 bbloom_: arohner: om/react don't help with html5 history at all, so yeah, you can use goog.events or you can use whatever other js lib you prefer for that task

12:17 arohner: I'm seeing what appears to be an advanced compilation bug (on google's side), so I'm not sure if I'm on the right track

12:18 which is obviously unlikely

12:18 gf3: arohner: you actually might be interested in https://github.com/gf3/secretary

12:18 arohner: which works with both history and html5history (or anything else really)

12:19 arohner: gf3: I'm using secretary, but this code catches events through google.history.Html5History and then dispatches to secretary

12:20 but the event google passes in to the event listener appears to be missing the critical field under advanced compilation, but not in whitespace

12:20 gf3: arohner: ohhh i think we ran into that once

12:21 arohner: this might be related → https://github.com/gf3/secretary/issues/14

12:22 arohner: gf3: I don't understand how?

12:23 gf3: arohner: what info is missing?

12:23 arohner: gf3: the 'token' field on the event

12:24 michaelr525: arohner: how Html5History is beter than History?

12:24 gf3: arohner: and how're you grabbing it?

12:24 arohner: michaelr525: I don't know, this is inherited

12:24 gf3: michaelr525: it doesn't use `document.location.hash`

12:24 arohner: gf3: aget. also missing when logged via console.log

12:24 gf3: michaelr525: it uses the HTML5 push/popstate api

12:25 michaelr525: oh

12:25 gf3: thank

12:25 gf3: btw, i'm using secretary too in a small om app :)

12:25 gf3: michaelr525: yayyyy :)

12:26 arohner: are you able to provide me with a paste via refheap.com ?

12:26 arohner: yeah, one sec

12:27 the app is based on omchaya, and looks like there's been a fix to omchaya. testing that first

12:27 yeah, ok

12:27 using (.-token e) rather than (aget e "token") fixed it

12:28 which is kind of obvious now

12:28 I'm kind of impressed closure strips fields out of events

12:28 gf3: arohner: :) excellent

12:28 arohner: gf3: thanks for your help

12:29 gf3: arohner: np—glad it worked out

12:30 ToxicFrog: So, I can destructure maps where the keys are keywords with {:keys [x y z]}, which will fish out :x :y :z from the map.

12:30 Is there an equivalent for when the keys are symbols rather than keywords?

12:30 arohner: :syms

12:30 michaelr525: gf3: i wonder, why didn't they just add the pop/push state capabilities to History instead of creating another class?

12:31 gf3: michaelr525: because it's not applicable in all situations

12:31 arohner: ,(let [{:syms [a b c]} '{a 1 b 2 c 3}] b)

12:31 clojurebot: 2

12:31 arohner: ToxicFrog: ^^

12:31 gf3: michaelr525: for instance older browsers don't support it, or there might be cases where you need to use the hash

12:31 arohner: there's also :strs

12:32 michaelr525: gf3: so they wanted to keep the interface for backward compatibility, or something

12:32 abaker: there a clojure version available in maven/clojars that has transducers, like 1.7.0-alpha1 or whatever?

12:32 michaelr525: gf3: i'd rather have one class which uses the html5 features when available and falls back to hash otherwise..

12:33 _alejandro: abaker: yes

12:33 [org.clojure/clojure "1.7.0-alpha1"]

12:33 abaker: cool, thanks

12:37 gsdfvu: warning

12:37 you may be watched

12:37 do usa&israel use the internet(facebook,youtube,twitter, chat rooms ..ect)to spy??

12:37 do usa&israel use the internet 2 collect informations,,can we call that spying??

12:37 do they record&analyse everything we do on the internet,,can they harm you using these informations??

12:37 تحذير

12:37 lazybot: gsdfvu: What are you, crazy? Of course not!

12:37 gsdfvu: Definitely not.

12:37 gsdfvu: Definitely not.

12:52 mdeboard: dnolen_: Any possibility of adding some repl options to mies? I'm struggling with how to set up a mixed-source (clj+cljs) project such that the repl is useful in both (including connecting browser to repl etc.)

12:53 or I guess if anyone can point me to such a thing

12:53 all the repl documentation is confusing since there's like a four-square grid of meanings to "repl": clj vs. cljs, in-emacs vs. command-line repl

12:54 dnolen_: mdeboard: not planning on adding any REPL support to mies in the near future

12:55 mdeboard: ok. thanks

12:57 dnolen_: So do you just use the command line repl for cljs work then? That seems the most clear-cut wrt integration with the browser and so forth.

12:57 dnolen_: mdeboard: inferior-lisp in Emacs

12:58 mdeboard: cemerick's "austin" lib works perfectly with command line, but I'm not sure where hte line is between austin & piggyback

12:58 oic

12:58 dnolen_: mdeboard: been thinking about improving REPL support but don't know when I'll get around to it

12:59 mdeboard: I understand

12:59 bbloom_: i've found weasel to be much more reliable

12:59 mdeboard: weasel?

12:59 clojurebot: weasel is https://github.com/tomjakubowski/weasel/

12:59 mdeboard: thanks clojurebot

13:00 bbloom_: oh interesting

13:00 I guess I need to understand how nrepl works

13:00 I think that's my prob

13:00 nrepl?

13:00 clojurebot: nrepl is a network REPL implementation: http://github.com/clojure/tools.nrepl

13:00 bbloom_: ~botsnack

13:00 clojurebot: Thanks! Can I have chocolate next time

13:01 mdeboard: You'll have botsnacks and you'll like it.

13:02 Maybe I should sign up to do an nrepl talk or something

13:16 TimMc: $botsnack

13:16 lazybot: TimMc: Thanks! Om nom nom!!

13:18 {blake}: I have a multimethod with a :default but I'm still getting a "No method in multimethod for dispatch value" error. How is this possible?

13:20 Must the namespace that contains the new type use the namespace that contains the :default method?

13:21 hiredman: {blake}: yes, you have to load code for it to do anything

13:21 llasram: {blake}: You do need to ensure that an implementing namespace is loaded in order for the implementation to exist

13:22 {blake}: The code is loaded. I have a system that includes a number of types, :int, :decimal, :float, etc., and I've added a new type in a new file.

13:23 And the default gets called for those other types. And the only difference I can observe is that the new type is in its own file.

13:23 llasram: {blake}: Can you produce a minimal reproducing example?

13:24 Because my understanding of the behavior you are describing is that it's not what should happen

13:24 {blake}: llasram, Not easily. OK, if it's not what should happen the problem must lie some place else. I'll continue to investigate.

13:24 ToBeReplaced: justin_smith: i reworked register! and added some other goodies... getting close to a 0.1.0 release

13:27 justin_smith: ToBeReplaced: oh, nice, I'll check it out

13:27 I still plan to work on it more, had some distractions in my personal life recently that got me off track

13:28 {blake}: llasram, hiredman OK, mystery solved. In the particular test cases the code WASN'T being loaded. Thanks!

13:29 (inc llasram)

13:29 lazybot: ⇒ 33

13:29 {blake}: (inc hiredman)

13:29 lazybot: ⇒ 52

13:56 martinklepsch: I use lein-garden which requires an old version of garden — is there any way to override a dependency's version of dependencies. (huh!)

13:56 I tried :exclusions but then it seems to be completely gone and can't be found anymore

14:05 somehow I must have messed up things really well... leiningen constantly tries to use garden 1.1.5 although I have garden 1.2.1 in :dependencies & made my own lein-garden that depends on 1.2.1

14:05 (instead of 1.1.5

14:12 SegFaultAX: martinklepsch: Whatever you did, it didn't work.

14:12 Just thought I'd clear that up for you. :)

14:13 llasram: martinklepsch: The project :dependencies are the dependencies for any project JVMs, which are different from the dependencies for the JVM for Leiningen itself

14:13 If `lein-garden` is in your :plugins, then it and its dependencies will be pulled into the Leiningen JVM (which sounds like what you want)

14:14 If you built a new version of lein-garden w/ the new dep, then make sure your :plugins specifies the correct variant version you built

14:14 grache28: "apply 0 f x = x ; apply n f x = f ( (apply (n-1) f) x)" - What's going on there? It's calling f which returns x, n times?

14:14 Sorry if syntax is off and for very basic question

14:15 llasram: grache28: Wrong channel?

14:16 TEttinger: grache28, I'm guessing that's a haskell question?

14:18 grache28: TEttinger: ok, sorry for the mix up, the language wasn't stated in the example

14:18 TEttinger: where'd you get it I wonder?

14:19 SegFaultAX: Looks like iterate

14:19 mdeboard: huh. TIL about inferior-lisp

14:19 grache28: TEttinger: university material - can ask prof. after weekend

14:19 TEttinger: as in "lisp that was not invented here" mdeboard?

14:20 mdeboard: TEttinger: hehe no the emacs command

14:20 TEttinger: that's exactly what I mean, mdeboard

14:20 doesn't it only consider emacs lisp full lisp?

14:20 clojurebot: excusez-moi

14:20 amalloy: yeah, looks like iterate in haskell

14:21 SegFaultAX: A strange way to write iterate.

14:21 mdeboard: TEttinger: idk, it's what dnolen_ said he uses for cljs repl

14:21 amalloy: well, a little weird. it's like !! and iterate glued together

14:21 SegFaultAX: Yea

14:21 TEttinger: and apply is a pretty overused name to be declared in user code like that

14:21 SegFaultAX: Usually `iterate f x = x : iterate f (f x)` will do ya

14:22 amalloy: maybe they should call it `data` instead, TEttinger

14:22 SegFaultAX: TEttinger: No kidding.

14:22 TEttinger: it doesn't appear to be creating a seq, does it?

14:22 mdeboard: TEttinger: Doing M-x inferior-lisp in a cljs buffer definitely opens a clojure repl so it (inferior-lisp) may have a hook that clojure-mode hooks into

14:22 amalloy: no. like i said, it's iterate and !! (which is nth)

14:22 SegFaultAX: TEttinger: That's why it's iterate + !!

14:22 Jinx

14:22 TEttinger: I didn't recognize !! , not being a type nerd :)

14:23 SegFaultAX: ?

14:23 It's just a function

14:23 TEttinger: I don't use haskell

14:23 I tried to learn it once

14:23 SegFaultAX: But then you took an arrow... I'm so sorry.

14:24 TEttinger: this was when people insisted on callin monads "warm fuzzy things" to avoid having to think about them

14:25 SegFaultAX: Was this like 15 years ago?

14:25 TEttinger: closer to 8

14:25 I'm sure haskell

14:26 I'm sure haskell's a fine language if you think in typesystems

14:26 http://tpolecat.github.io/assets/list.png

14:26 SegFaultAX: Haskell is no different than any other tool. It has a range of problems it's well suited for, and a range of problems it isn't.

14:26 TEttinger: that's Scala's List class superclass diagram

14:26 bbloom_: TEttinger: that graph makes me laugh every time

14:26 SegFaultAX: Depending on the fanboyism of whom you're talking to, that range will grow and shrink accordingly

14:27 bbloom_: i mean, clojure's seq/coll stuff is pretty complex too

14:27 but that's just fun

14:27 TEttinger: bbloom_, yeah I guess

14:27 SegFaultAX: bbloom_: If by laugh, you mean weep tears of blood, yea me too.

14:27 TEttinger: Scala seems to have a phobia of nil

14:27 mdrogalis: Does anyone know when the Conj talks get announced? Can't remember if it was today or next week.

14:27 SegFaultAX: TEttinger: It's a good phobia to have.

14:27 TEttinger: clojure just handles nil...

14:28 ,(seq nil)

14:28 clojurebot: nil

14:28 SegFaultAX: Because Clojure tries to make nil act like nil in some other lisps (which is the same as ())

14:28 Unless it doesn't, then NPE wheeee

14:29 TEttinger: SegFaultAX, I can't even remember the last NPE I got in clojure

14:29 bbloom_: the NPE issue is b/c nil isn't an object, like in smalltalk or ruby or whatever

14:29 it's not a problem with the concept of a universal nil value, it's a problem with the implementation of it as a null pointer

14:29 SegFaultAX: Correct

14:30 TEttinger: I do recall some fun bugs to track down in C#

14:30 bbloom_: but we can extend protocols to nil :-)

14:30 SegFaultAX: TEttinger: Again, because many times you're working with things that expect sequences, and nil is just an empty sequence in that context.

14:30 This is not an unusual property of lisps.

14:30 TEttinger: if you use GDI+ to try to get a subset of an image that is larger than the image itself, it doesn't give a nice error, it gives an Out of Memory exception

14:31 this was hard to track down

14:31 SegFaultAX: Maybe they meant "out of the memory region of this image!"

14:31 And thought OOM sounded nicer.

14:31 TEttinger: which would have been nice to know!

14:31 stompyj: but type systems!

14:31 TEttinger: stompyj, heh

14:32 stompyj: sorry, I saw Haskell in the convo, and thought it would be appropriate :D

14:32 I have equal love for clojure and haskell

14:32 martinklepsch: can I somehow list out :plugin dependencies?

14:32 arrdem: stuck in IO monad pls send help

14:32 SegFaultAX: TEttinger: The thing about advanced type systems is... it's truly shocking just how much stuff can be decided statically (eg at compile time)

14:33 TEttinger: and yeah C# has been moderately-decent for what I've been using, which is basically array-heavy code that sometimes needs to break out into functional guitar solos

14:33 SegFaultAX: Haskell certainly tries to push that to the logical extreme in some cases.

14:33 And dependent types allow you to take it further still.

14:33 But being able to identify a huge category of bugs without ever having to /run your program/ is a pretty nice property.

14:34 TEttinger: oddly enough, the fastest speedup in my C# code has been a relatively small, relatively challenging function that uses raw memory writes. and that didn't throw memory errors, ever

14:35 SegFaultAX: And that property extends over most statically typed languages to a certain degree. Even relatively crappy ones like C, C++, and Java.

14:36 TEttinger: the problem I had was entertaining to look at and possibly informative. https://dl.dropboxusercontent.com/u/11914692/Artillery_S_Large_face0_0.png is how it should look.

14:36 but it ended up generating https://dl.dropboxusercontent.com/u/11914692/Artillery_S_Large_face0_0_nogreen.png

14:37 the issue had to do with ARGB beingstored as BGRA in memory (big endian vs. little endian)

15:37 akhudek: I just want to double check that my understanding is regarding om. You generally shouldn’t use set-state outside of one of the om events: will-update, did-updated, etc. E.g. you shouldn’t use set-state in a callback?

15:47 bbloom_: akhudek: you can use set state in a callback

15:47 akhudek: http://facebook.github.io/react/docs/forms.html

15:48 akhudek: bbloom_: hmm. Ok curious. I seem to be encountering a case where state is not triggering a re-render. I’ll do into some more. I’m probably not on the latest om either, so will check that. Thanks.

15:48 s/do/dig/

16:09 martinklepsch: is there something like lein deps :tree for plugins in my project?

16:10 xeqi: martinklepsch: lein deps :plugin-tree

16:13 supersym: cool..didnt know that one

16:14 martinklepsch: xeqi: thanks!

16:16 I have [garden "1.1.5"] nowhere in there but keep getting this error + a stacktrace that (I think) wouldn't come up if the right version is used

16:16 (WARNING!!! version ranges found for:)

16:16 ([garden "1.1.5"] -> [com.keminglabs/cljx "0.3.1"] -> [org.clojars.trptcolin/sjacket ""] -> [org.clojure/clojure "[1.3.0,)"]

16:16 Consider using [garden "1.1.5" :exclusions [org.clojure/clojure]].)

16:21 Any ideas what could cause that? Here is my project.clj: https://gist.github.com/1fbb4961f902daf4e851

16:23 hyPiRion: martinklepsch: could you provide the output of `lein deps :tree`?

16:24 martinklepsch: oh wait, https://github.com/noprompt/lein-garden/blob/master/src/leiningen/garden.clj#L112-L114 seems to be injected

16:25 I guess you should add an issue to the plugin, because that's not good style

16:27 martinklepsch: hyPiRion: ohhh!

16:27 that makes sense!

16:27 akhudek: hmm, strange, I can clearly see that it’s calling the set-sate, yet om/react doesn’t always trigger a render from it

16:28 hyPiRion: I think a short article on how to write proper lein plugins would be worth looking into.

16:28 martinklepsch: hyPiRion: is there a reason why that injection there could make sense?

16:31 hyPiRion: I'd happily write a patch but I'm not sure what's the right way — should I just remove the profile/with-profile stuff?

16:34 hyPiRion: martinklepsch: they are running an `eval-in-project` call, so they want to ensure the project depends on garden and that other dependency. If you add ^:displace to both dependency entries, I think it should be fixed

16:35 that is, it should be {:dependencies '[^:displace [garden "1.1.5"] ^:displace [ns-tracker "0.2.1"]]}

16:42 wei: what’s the recommended way to do interprocess communication between clojure programs?

16:51 justin_smith: wei: you could use a java IPC like RMI

16:51 or whatever is easiest to use on a platform specific basis I guess

16:54 wei: something like netflix’s ribbon? i was originally thinking of using redis queues, but that doesn’t have easy two-way communication

16:55 akhudek: wei: we use rabbitmq

16:58 wei: akhudek: and Langohr, i presume?

16:58 akhudek: wei: yep

16:58 wei: why did you choose rabbitmq over the alternatives? e.g. redis

16:59 guaranteed delivery?

16:59 akhudek: that, plus durability, high availability and a message passing abstraction just worked better for us

17:00 processes pass larger pieces of data around via postgresql

17:01 but that doesn’t help indicating that something has changed and work needs to be done

17:01 that’s where rmq comes in

17:01 bbloom_: guarenteed deliverability is a lie...

17:01 http://aphyr.com/posts/315-call-me-maybe-rabbitmq

17:01 better to plan for dropped messages than to cross your fingers and hope

17:03 technomancy: bbloom_: sure, but compared to redis?

17:03 bbloom_: technomancy: i wasn't comparing to redis

17:03 im just saying: in general, queues drop messages

17:05 mdrogalis: bbloom_: I think that's a pretty sharp statement to make.

17:06 There're a lot of queues that have transactions, and do durable writes.

17:06 wei: bbloom_: so, how do you generally plan for dropped messages, if you require guaranteed deliverability? e.g. payment notifications

17:07 bbloom_: wei: retries, idempotency, associativity, commutativity, etc

17:09 wei: bbloom_: is there an article discussing it in more depth?

17:10 e.g. how do you make payments commutative?

17:11 bbloom_: wei: the same way banks do it: with a line of credit :-)

17:11 mthvedt: wei: a surprising amount of payments in the financial system are assumed in the short run to be commutative

17:11 technomancy: bank accounts are the worst examples for consistent distributed systems documentation

17:11 bbloom_: technomancy: yeah seriously

17:12 everybody is like "so you want to do an atomic account decrement" but then nobody actually ever does that

17:12 they do double entry accounting, etc

17:12 mthvedt: technomancy: they have the best consistency implementation in existence. if you cause an inconsistency you pay a $35 overdraft fee

17:12 who needs fancy computer science theory

17:12 mdeboard: lol

17:12 social engineering uber alles

17:12 bbloom_: mthvedt: consistency by fiat

17:12 technomancy: mthvedt: oh man I would love to see that proposed at a startup

17:12 wei: this is bitcoin related so credit doesn’t quite work. if someone withdraws their “credit”, it’s gone

17:12 mdeboard: pavlov-driven development

17:13 bbloom_: wei: when you deal with external systems that don't support sensible distributed semantics, you need to be more cautious....

17:13 bitcoin itself is robust(ish?) to this sort of thing

17:13 but random REST APIs for it aren't going to be

17:14 wei: for one, it’s idempotent thanks to payment hashes

17:14 bbloom_: organizations like coinbase operate like banks, they have extra money/bitcoin to cover discrepencies, and they work it out later

17:15 technomancy: eventual consistency is a thing

17:15 wei: although with that systme, if your missed / doubled transaction is for a huge amount relative to your reserves, you’re screwed

17:16 bbloom_: technomancy: unless your hewitt, in which case you don't think any consistency is a thing

17:16 wei: you ever have to wait for a check to clear?

17:16 wei: and for a toy project like mine that’s entirely possible

17:16 bbloom_: banks do things to mitigate problems like that

17:17 when you start talking about larger amounts of money, consistency becomes even more strange

17:17 like if you have more money than the FDIC insurance limit (250k?) then suddenly all sorts of stuff is slower and weirder

17:17 wei: i guess that’s why coinbase much longer than the blockchain to confirm a transaction

17:17 technomancy: bbloom_: I'm no platonist, but I'm inclined to agree when it comes to the physical realm.

17:18 mthvedt: bbloom_: when finance companies deal with each other, there basically aren’t any rules

17:18 when i worked at an exchange every so often we’d get a call

17:18 “hey you misplaced some trades/a few million dollars"

17:18 bbloom_: technomancy: it's like how all machines are actually finite state machines... sure you've got a finite amount of memory, but you still treat it like a turing machine... same deal w/ eventual consistency. it may not actually exist in practice, but it's a useful concept to reason about

17:19 technomancy: brb staring at flickering shadows

17:21 wei: anyways, thanks for the thoughts. i’d appreciate any literature you guys have on the subject. i basically want guaranteed deliverability and de-duping, whether it’s guaranteed by the messaging system or by a particular system design

17:23 mdrogalis: wei: I can share a little from the toolchain that I work with. HornetQ supports transactions across queues with durable writes. That might buy you what you're looking for.

17:24 I think ActiveMQ can do the same.

17:24 akhudek: mdrogalis: bbloom was pointing out that it’s still possible in weird conditions to lose messages or get duplicates

17:25 technomancy: dilemma: do you use the systems that aphyr *has* blogged about even though they are all subject to various failures, or the ones he *hasn't* blogged about, because you know, they could be good. =)

17:25 mdrogalis: akhudek: Got it.

17:25 bbloom_: mdrogalis: it's worth reading the entire call-me-maybe series

17:26 akhudek: technomancy: better the problems that you know about?

17:26 technomancy: it’s unlikely that others have any better solutions since these problems are genuinely difficult

17:26 mdrogalis: bbloom_: I've read a lot of it. But at the end of the day, I actually have to ship stuff without rewriting huge pieces of infrastructure that I woulnt get 100% correct

17:28 akhudek: ack, ok, om is 100% not always triggering a rerender on state change

17:28 bbloom_: mdrogalis: i'm not suggesting you write custom infrastructure, i'm suggesting you put a little extra data in to your messages and process those messages in a way that is tolerant network truths

17:28 akhudek: very sodd

17:28 odd even

17:29 mdrogalis: bbloom_: No disagreement there, for sure.

18:07 akhudek: any om wizards around?

18:08 I have a case where I call set-state! in a callback yet even should-update isn’t triggered

18:33 danielcompton: Is clojars.org unavailable for others or just me?

18:34 Bronsa: it's working for me

18:38 johnwalker: i'm familiar with lein do {task}, {task}, {task}

18:39 is there a way to change the project context for successive tasks ?

18:39 for example - lein do new mies-om, {change-ctx}, ancient upgrade

19:03 xnull: hei

19:03 oops

19:06 technomancy: johnwalker: sure: lein do thingy, with-profile +p1 other-thingy

19:07 johnwalker: if your first task returns a project map there is also https://github.com/technomancy/lein-thrush but I'm not as confident of that

20:35 TimMc: technomancy: What's the associativity of ,? :-P

21:32 ben_vulpes: i've got an enlive question - this refheap ((html/clone-for [[ad] [advertisements]]

21:32 [:tr.ad-data]

21:32 (html/clone-for [[copy url bid] [ad]]

21:32 [:td.copy] (html/content (str copy))

21:32 [:td.url] (html/content (str url))

21:32 ah nuts.

21:33 https://www.refheap.com/89215 <- that refheap is only returning the td.copy node, and not the url or bid nodes.

21:33 i've verified that copy url and bid are all in scope at that time, but can't seem to get the clone-for to make all the nodes i'm looking for

22:08 ToBeReplaced: [ANN] https://github.com/ToBeReplaced/nio.file 0.1.0 , for those interested in using java.nio.file.* classes and methods from clojure

22:25 ben_vulpes: ToBeReplaced: neat stuff. looking forward to it.

22:26 ToBeReplaced: thanks

23:45 arrdem: ToBeReplaced: so at what point is your name just never going to be replaced :P

23:47 ToBeReplaced: arrdem: idk, it seems like a nice UUID, and keeps in-person separate from not-in-person

Logging service provided by n01se.net