#clojure log - Feb 03 2016

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

1:36 marcfontaine: using cursive, I am getting the following error when compiling, there are no errors when sending the form directly to the repl.

1:36 CompilerException java.lang.IllegalArgumentException: Value out of range for char: -61, compiling:(extract_from_quotes.clj:80:18)

1:36 (def invoice-ids (apply str (interpose "," (map #(:invoiceId %) (flatten (map vals clearquotes)))))

1:36 )

1:36 => #'csv.extract-from-quotes/invoice-ids

1:38 TEttinger: try checking the result of just (map :invoiceId (flatten (map vals clearquotes)))

1:38 wait

1:39 ,(flatten [{:invoiceId 1 :a 2} {:invoiceId 3 :a 4}])

1:39 clojurebot: ({:invoiceId 1, :a 2} {:invoiceId 3, :a 4})

1:39 TEttinger: ah, wasn't sure if flatten would do something to maps

1:40 marcfontaine: (map :invoiceId (flatten (map vals clearquotes))) =>("1000882" "1000883" "1000924" “1000926" …. )

1:40 TEttinger: that “1000926" is odd

1:41 one quote is a smart quote, not sure if that's just IRC

1:41 ,(interpose "," ["1000882" "1000883" ])

1:41 clojurebot: ("1000882" "," "1000883")

1:42 marcfontaine: yeah that’s just me typing

1:42 TEttinger: is clearquotes from a file?

1:43 ,(apply str (interpose "," ["1000882" "1000883" -61]))

1:43 clojurebot: "1000882,1000883,-61"

1:44 TEttinger: is this online, marcfontaine ?

1:44 the source I mean

1:46 amalloy: clojure.string/join

1:46 also, clojure.core/mapcat

1:47 TEttinger: amalloy: thoughts on the out of range for char: -61 error?

1:49 amalloy: i think -61 is not a very good character

1:50 marcfontaine: ok great pointer, the clearquotes are coming from the deserialization of php data.

1:50 TEttinger: indeed. any idea what could cause that?

1:50 oooh

1:50 encoding maybe?

2:40 marcfontaine: ok problem was converting result (a bytearray of valid data size nb) to string using (subs (map char result) 0 nb) instead of (subs (String. result “UTF-8”) 0 nb) so the discrepancy between compilation and repl was due to how the byte-array is initialized in the compiler vs repl.

2:47 or better yet (String. result 0 nb "UTF-8")

5:06 anti-freeze: Hi everyone. I'm having issues with core.async and transforming values. I have a list of files, I want to read them and then convert them to js/Blob objects before passing them to the out channel. I've tried this, but to no avail https://www.refheap.com/114412

5:07 io/read-file returns a channel

8:00 jeaye: I see this when I try to minify http://dpaste.com/3X3PTKJ#wrap

8:00 Any ideas what'd cause this?

8:04 CaptainLex: jeaye: I guess the first thing that springs to mind is the real file path you're trying might not exist or it may be malformed

8:04 jeaye: It seems like the kind of error that could be caused by a null file

8:04 jeaye: What does it output if you use a file you /know/ doesn't exist?

8:06 jeaye: So, I tried specifying full paths for in and out. It gives me an error saying out must be relative. If I make in absolute and out relative, the same initial error shows.

8:06 CaptainLex: If I specify files that don't exist, the same error comes up.

8:06 CaptainLex: jeaye: That might be a permissions issue then. Do you own the file you're trying to use, and does boot run as your user?

8:07 jeaye: I would expect bo things to be true, but one can never be too careful

8:07 jeaye: Everything is owned by my user and being run by the same user.

8:12 CaptainLex: jeaye: Sorry, lost connection. Did I miss anything?

8:13 jeaye: CaptainLex: Someone in #bootclj has it figured out, I think.

8:13 I'll follow up.

8:13 CaptainLex: jeaye: Thanks!

8:26 jeaye: CaptainLex: It wasn't part of the fileset `boot show -f` so minify couldn't find the file. I thought I could run it on arbitrary files, but I can't.

8:26 CaptainLex: If I reference something in the src directories, the error doesn't show up.

8:27 CaptainLex: jeaye: Aha! Yep, I would have never been able to come up with that one. I'm glad it worked out for you!

8:27 jeaye: CaptainLex: Thank you!

8:54 Kamuela: Elixir must be a lisp in the same way that javascript is a lisp

8:55 because it feels like a basic language with basic syntax to me

8:55 is “has first-class functions” all it takes?

9:05 tdammers: IMO neither JS nor Elixir are lisps

9:06 mpenet: indeed

9:06 wmealing: elixir is nowhere near a lisp

9:06 * wmealing boggles at that one

9:06 tdammers: defining feature for being a lisp would be s-expressions, and how they are used for both code and data, essentially making code first-class for free

9:06 mpenet: it's more about the homoiconicity

9:06 wmealing: there is lisp flavored erlang, dont get me wrong

9:07 * wmealing goes back to his hole in the ground

9:10 tdammers: there is some heavy lisp influence in both JS and elixir though

9:11 mpenet: there are just first class functions the rest is pretty common

9:11 tdammers: first class functions are also pretty common

9:11 kvtb: question, in ClojureScript, (max nil nil) returns nil, but in Clojure, (max nil nil) gives NullPointerException. How can I ensure (max nil nil) in Clojure returns nil like in ClojureScript?

9:12 tdammers: lots of details that resemble scheme though - scope, closures, defining functions as lambdas bound to variables, etc.

9:12 (in JS, that is)

9:12 CaptainLex: kvtb: You could try catching that exception and returning nil, right?

9:14 kvtb: thanks CaptainLex did not not think of that :)

9:15 CaptainLex: kvtb: No problem!

9:15 wmealing1: i often wonder if we're going to be stuck in js land forever in the browser.

9:16 its a little weird.

9:16 CaptainLex: wmealing1: JS as a target or JS as a high-level language?

9:16 wmealing1: js as the target

9:16 MJB47: does asm.js count?

9:17 wmealing1: it still feels like js

9:17 CaptainLex: Yeah, I worry about that too. If it works "well enough" then there will be no push to move to something nicer and more efficient

9:17 On the other hand, using only JS until every browser agrees on a different standard would be bad for everyone

9:18 wmealing1: so, thats it then, we're stuck

9:18 i for one welcome our new js overlords.

9:19 the browser is going to be around for a long, long time

9:19 changing / adding another language is like playing chicken with the user

9:19 while i know other things can transpile to javascript

9:20 mpenet: given that transpiled code often ends up more performant than hand writen code I don't think js is a huge problem anymore

9:20 tdammers: to be honest, none of the existing compile-to-JS solutions really convince me

9:20 wmealing1: it seems that js tends to encourage some neurotic programming.

9:21 mpenet: at least we have options now

9:21 wmealing1: that is true

9:21 CaptainLex: Yeah, the asynchronicity instead of real multi-threading leads to some really clunky workarounds for really common paradigms

9:21 If we can't figure out a patch over that, every JS-targeting language will suffer from it

9:22 mpenet: this cuold evolve, we didn't have xhr until "recently"

9:22 could*

9:23 tdammers: frankly, I believe there isn't a compelling reason why JS couldn't be extended to a multi-threaded runtime

9:23 shared-mutable-state multithreading would be problematic due to the assumptions you can currently make, but with decent concurrency semantics, I believe this could be solved

9:24 but you'd have to sandbox the whole threading runtime to avoid rogue JS eating up all the available OS threads

9:24 and given how difficult sandboxing JS already is, ...

9:25 wmealing1: on the side note, i really wanted to like elixir, but clojure won me over elixir due to clojurescript

9:25 having the same language on both client and server, and one thats not insane.. is nice

9:26 CaptainLex: wmealing1: Yeah, what sold me on the pairing was shoreleave. Seamlessly sharing data between client and server! A dream come true!

9:26 tdammers: except that clojure and clojurescript are different enough for it to matter

9:27 wmealing1: tdammers: brain context switching from clojure -> clojurescript is easier than elixir -> javascript

9:27 i can't fit that in my head

9:27 tdammers: maybe I'm different that way

9:28 wmealing1: (i also need to know C and c++)

9:28 and erlang

9:28 tdammers: ah, hm

9:28 I'm fairly comfortable with C myself, which helps a lot

9:28 wmealing1: so ive dropped elixir and erlang..

9:28 tdammers: my go-to combo, tentatively, is haskell + js

9:28 wmealing1: i thought about learning haskell, it seemed.. to be a language people talk about

9:28 but dont get things done in

9:29 but again, thats my casual observation

9:29 tdammers: idk, haskell works fine for me

9:29 CaptainLex: tdammers: Have you looked into Haskell-y JS languages like Elm or PureScript?

9:29 tdammers: yeah

9:29 purescript is still on my list

9:29 elm looked promising at first, but I ended up dropping it

9:29 wmealing1: tdammers: i'd love for it to "work" for me too.. it looks sane as far as a language goes

9:30 tdammers: for two reasons; one, it is too simple, and two, it's unstable and the toolchain is a bit brittle

9:30 wmealing1: brittle like, unstable ?

9:30 tdammers: yeah

9:30 CaptainLex: Unstable like the API keeps changing?

9:30 tdammers: yes

9:30 the API and the language itself change

9:31 and the toolchain is supposed to deal with it automagically

9:31 CaptainLex: Ahhhh, yes. So good to play around with, but not to write code you need maintain in

9:31 Not yet, anyway

9:31 tdammers: which is great, but when the toolchain fails to resolve those issues, you have nowhere to start looking

9:31 there was this one thing that cost me several hours to figure out, where they had added a validator for the required package field that pointed at a github repo for the package

9:32 and the default value that the package generator inserted was invalid

9:32 wmealing1: ouch

9:32 tdammers: which is doubly frustrating because I wasn't ever going to publish this package, and it has never been nor will it ever be on github

9:33 wmealing1: tdammers: i read that many people complain about debugging in haskell

9:33 tdammers: do you use one, or do any ?

9:33 tdammers: use what, a debugger?

9:33 wmealing1: yes

9:33 tdammers: I don't

9:33 wmealing1: ive been tossing up haskell/ocaml on another project

9:34 tdammers: I don't generally have to do a lot of debugging, most of it is a matter of fixing compiler errors

9:34 the rest I can almost always resolve using automated tests

9:34 haskell shines at automated testing

9:34 wmealing1: ok

9:34 is there a midje like tool ?

9:35 sdegutis: Good morning all.

9:35 * wmealing1 waves to sdegutis

9:35 sdegutis: Hi there.

9:35 wmealing1: if clojure talk happens in here, i'll mute.

9:35 sdegutis: Haha don't worry about it, we love talking about Haskell in here.

9:35 tdammers: wmealing1: or head over to #haskell ;)

9:35 wmealing1: tdammers: heh

9:36 sdegutis: Question: What are the valid namespace characters?

9:36 wmealing1: tdammers: i was in there for a bit, i felt so many wooshes… listening

9:36 sdegutis: Is it just alphanumeric plus hyphens?

9:36 Or can you also use something like % ?

9:36 tdammers: anyway, there's quickcheck for property tests, hunit for unit and integration tests, and tasty to provide a seamless unified API over them and a bunch of other testing libraries

9:36 wmealing1: i'll take a look !

9:36 tdammers: so the approach is different from midje, but definitely powerful

9:37 CaptainLex: sdegutis: I think you can use any identifier-valid character for namespaces, but be careful your OS can find those special characters on the filesystem

9:37 sdegutis: But I don't know for sure

9:37 tdammers: and, again, a lot of the stuff you'd normally do in unit and property tests can be offloaded to the type checker

9:37 * wmealing1 nods

9:37 sdegutis: Good morning justin_smith.

9:37 But this only matters on the machine that is compiling the uberjar for deployment, right?

9:38 CaptainLex: sdegutis: That sounds about right to me, yeah. But I am not a jvm or systems or Clojure-compilation expert of any kind!

9:38 wmealing1: hmm the jvm itself would need to unpack the filepaths

9:38 so that could also get wierd

9:39 tdammers: "which characters are valid in namespaces" is one of those questions where the answer is "if you need to ask, you're doing it wrong" :D

9:39 (like "what is the limit for how many columns I can have in a database table")

9:39 CaptainLex: Hehehe

9:41 wmealing1: tdammers: my favourite is 'how many subdirectories deep does the operating system support'

9:41 what.. is someone doing to require that !?

9:42 tdammers: well, things like sysfs and such might hit the limit at some point

9:42 wmealing1: oh this person was -definitely- messing around on ext4

9:44 MJB47: at work we had a node project that couldnt run on windows because the node modules sub directories were too deep

9:44 was annoying

9:44 CaptainLex: That's a complex situation, morally

9:44 On the one hand, what the hell Microsoft, as usual

9:44 wmealing1: running node, or running windows ?

9:44 CaptainLex: On the other, how many directories deep was it??

9:45 MJB47: idk what the number was

9:45 but it worked fine on every other OS

9:45 wasnt a terribly big project

9:46 TMA: on Windows(R) the PATH_MAX is the problem

9:46 CaptainLex: PATH_MAX sounds like a problem

9:46 wmealing1: you can work around it with relative referencing

9:47 Glenjamin: unc paths also work around afaik

9:47 MJB47: idr how we fixed it

9:47 wmealing1: but its annoying

9:47 ah

9:47 MJB47: everyone used unix systems

9:47 Glenjamin: npm v3 also made some changes to avoid path lengths

9:47 MJB47: except 1 guy

9:48 wmealing1: that guy..

9:48 MJB47: ikr

9:54 sdegutis: Hi.

9:55 CaptainLex: sdegutis: Hello again!

10:59 Shayanjm: ping justin_smith

11:47 sdegutis: Hi.

11:48 How are you?

12:41 justin_smith: Shayanjm: hello

12:42 Shayanjm: ello justin_smith - managed to get the author on skype to help me debug this thing

12:43 justin_smith: Shayanjm: OK. It's a kafka replacement not a kafka library, right?

12:43 Shayanjm: justin_smith: It's a kafka library but it uses redis instead of ZK on the consumer side because of scalability/reliability concerns at scale

12:44 the producer side publishes messages directly to kafka via the brokers

12:44 justin_smith: oh I had no idea that was an option...

12:44 the redis part that is

12:44 publishing via brokers is totally standard

12:44 Shayanjm: choosing redis over zookeeper for reliability sounds really sketchy to me

12:45 Shayanjm: justin_smith: Yeah, I wanted to try it out because it sounded interesting

12:45 if you read the readme on the project he has some points about why he made that decision

12:48 justin_smith: Shayanjm: btw in terms of reliability, out of CAP (where you can pick at most two) redis picks the empty set - it provides none of the three

12:49 Shayanjm: the amount of data you can lose is bounded by the replication lag

12:50 Shayanjm: I don't doubt that redis is easier to use and easier to scale, but reliability-wise it isn't even near zookeeper's league

12:52 amalloy: i bet it's faster though

12:52 justin_smith: amalloy: I don't doubt it at all - though this is why kafka uses zookeeper for coordination not for messages

12:52 once you have redis replacing zookeeper, why use kafka even?

12:53 amalloy: who needs reliability when you're blazing fast

12:53 justin_smith: amalloy: it makes me cry tears of blood because it's true

12:58 Shayanjm: justin_smith: got it working, but it's a bit slow

12:58 going to see if increasing JVM RAM allocation will help

12:59 dysfun: anyone want to give me some feedback on this pre-release library please? :) https://github.com/irresponsible/emotional

13:00 justin_smith: Shayanjm: what is "slow"? I don't know what your domain is, but based on my vanilla kafka usage, if kafka was my speed bottleneck I would have an amazingly performant system.

13:00 CaptainLex: dysfun: That's a good URL.

13:00 dysfun: yeah, it is :)

13:00 justin_smith: haha

13:00 Shayanjm: justin_smith: it's the way the consumer is built I think

13:00 dysfun: hadn't really occurred to me haha

13:00 Shayanjm: at least on his system

13:00 it hangs for at least a minute before returning

13:01 CaptainLex: dysfun: Although assuming your nick is short for dysfunctional, you should figure out how to work that in too

13:01 Shayanjm: he provides a vagrant build that can be used for testing, so I might test on that to see if its faster

13:01 dysfun: CaptainLex: it is, but i'm content to leave that out for this one project. there will be others :)

13:01 justin_smith: Shayanjm: oh wow - so in my usage the consumer doesn't return - I leave a consumer polling for messages (in a go loop or thread) and give it a function to call for each message it gets

13:01 dysfun: part of me is curious about what is the limit for people being prepared to use your library's name in front of their managers

13:02 justin_smith: I mean eventually the consumer returns when I forcibly shut it down but that's different

13:02 dysfun: i mean if you call your library something downright offensive, you're liable to get entire swathes of users refuse to use it

13:02 Shayanjm: justin_smith: Yeah for right now (testing purposes) I wanted to manually see if I could get the messages and do stuff with them

13:02 so I wanted the consumer to sit there, read the messages, and stick them somewhere

13:03 amalloy: kafka is for throughput, not latency, isn't it?

13:03 justin_smith: dysfun: (:require [emotional.core :as rational])

13:03 dysfun: justin_smith: this is my 'irresponsible clojure organisation', there is no rational

13:03 justin_smith: amalloy: throughput plus flexible routing with a log doubling as message queue, yes

13:04 plus reliability

13:04 Shayanjm: yeah, he has a good point, if latency is more important than throughput kafka might not be your bag, unless you want the other features badly enough that you can cope

13:10 Shayanjm: you mention that this kafka re-implementation has a rationale in the readme, but I'm not finding it. Is this the right README? https://github.com/gerritjvv/kafka-fast/#kafka-clj

13:10 Shayanjm: https://github.com/gerritjvv/kafka-fast/#consumer

13:11 the readme(s) are a bit weirdly set up

13:12 justin_smith: Shayanjm: OK, so he choses fast+simple at the expense of reliable

13:13 Shayanjm: basically, with the assumption that by clustering redis you can approach the reliability of ZK at scale

13:13 justin_smith: Shayanjm: that bullshit

13:13 sorry, *that's bullshit

13:13 Shayanjm: justin_smith: Yeah I don't know enough about ZK to speak intelligently about the approach

13:13 but I thought it was interesting that you could use redis + kafka together without them fighting

13:14 so i'm poking at it to see how it works

13:14 it's basically using redis to store the offsets of each tracked topic

13:14 justin_smith: Shayanjm: by reliability I mean "knowing your message gets from one end to the other", redis can't offer that... but OK

13:14 Shayanjm: justin_smith: Yeah I think he means "if one redis instance falls over, you can still guarantee some level of data integrity"

13:14 justin_smith: so the failure isn't "redis not running" but "silent data loss"

13:15 dysfun: yay!

13:15 justin_smith: I mean if continuously available while dropping some messages fits your use case, go for it

13:15 Shayanjm: yeah that ^^^ I asked him about silent losses earlier and he said "he hasn't seen any with a production deploy of a single redis instance"

13:15 amalloy: justin_smith: if you use any of the consumers built into kafka you can't actually know that either

13:15 mgaare: it depends on the semantics your app requires for message processing.

13:16 with Kafka, in the case of zk/redis failure you wouldn't get dropped messages, you'd have messages consumed more than once

13:16 amalloy: because the consumers ACK the message as soon as they receive it, before your client code can confirm it's acted on the message

13:17 justin_smith: mgaare: oh, interesting.

13:18 mgaare: justin_smith: zk (and presumably redis in this alternate implementation) are only storing a given consumer's position in the topic, not any message data.

13:18 justin_smith: mgaare: of course, yeah - I should have connected those dots

13:18 I do know about the partitions on disk vs. the coordination on zk

13:19 mgaare: I don't think you can be reasonably blamed for a mistake here. services written in java and distributed systems both seem to have documentation that obfuscates the basic idea of what's going on - kafka even more so because it's both

13:20 justin_smith: amalloy: yeah, we end up using zookeeper directly because of this - we get a message via kafka, if the message describes a job to start zookeeper is used to mark the fact that it was started and in case of a system failure another box can pick up the orphaned job

13:20 Shayanjm: mgaare justin_smith: mgaare is right. redis just holds the offsets on each topic for the consumers

13:20 I'd assume that's exactly what zk does

13:20 amalloy: right. you're just having to rebuild reliability on top of kafka, justin_smith

13:21 justin_smith: amalloy: right, fair enough.

13:21 amalloy: which like, yuck, isn't that supposed to be the point

13:21 justin_smith: the workaround i used was to instead disable automatic ACKs, and write consumer offsets to zk myself

13:22 justin_smith: amalloy: to me distsys is like security - we have a huge number of known was to totally fuck it up, a few heuristics for what works better, and a lot of trying and hoping...

13:22 amalloy: interesting, I'll have to look into that some time

13:22 dysfun: justin_smith: pretty much

13:22 amalloy: i wish hadoop could be described so positively

13:23 dysfun: yes but it's difficult to even build hadoop, what do you expect? :p

13:24 justin_smith: oh, reminds me of this tweet from sorenmacbeth yesterday https://twitter.com/sorenmacbeth/status/694679147784765440

13:24 mgaare: amalloy: what lib are you using to talk to zk?

13:25 amalloy: i don't remember. zookeeper-clj or something

13:26 justin_smith: Shayanjm: so it sounds like this redis for offsets thing might have legs - but I don't see where it would help with data throughput

13:26 amalloy: justin_smith: not related except that it's an image on twitter: https://twitter.com/HenryHoffman/status/694184106440200192

13:26 Shayanjm: I'm still piecing it together myself justin_smith

13:26 justin_smith: amalloy: yeah, I retweeted that one

13:28 Shayanjm: I still can't figure out why the fuck it's hanging with 6 background polling threads

13:28 mgaare: justin_smith Shayanjm: I'd be inclined to look at it from a developer and ops friendliness angle rather than throughput/distributed theory. I think redis is a lot easier to work with than zookeeper for both dev and ops.

13:29 Shayanjm: mgaare: It could be, but for the most part this library has/is trying to abstract a lot of that away

13:29 justin_smith: /dev/null has a pretty awesome api

13:29 really easy to use!

13:29 :P

13:29 dysfun: maybe that's why mysql writes too much data there :p

13:31 amalloy: i think a lot of people doing distributed stuff have zookeeper already

13:31 justin_smith: amalloy: and if they don't, they'll probably need it soon

13:32 dysfun: only if they're in the habit of shipping in big use-all-the-things tools

13:45 justin_smith: dysfun: SELECT 0='banana' http://grimoire.ca/mysql/choose-something-else

13:55 dysfun: looking at irresponsible/emotional - I never know whether to expect someproject.core to contain "the core" as in the definitions that everything in the project uses vs. the top level functionality a client is looking for

13:55 dysfun: yup

13:55 justin_smith: my solution to this is to never have a .core ns

13:55 dysfun: a blunt instrument, but sure

13:57 justin_smith: what's the rationale for having one? the fact that lein created one? the fact that clojure has one?

14:01 amalloy: that many other projects have one

14:01 dysfun: justin_smith: mostly that single segment namespaces are disapproved of and you need *somewhere* to act as the main point of interaction

14:02 justin_smith: dysfun: my github group is noisesmith, if I make a new project foo, the top level ns that clients use will be noisesmith.foo

14:02 dysfun: yeah i thought of that, then though that irresponsible is a long word and i'm lazy

14:02 justin_smith: if I followed java conventions more closely it would be org.noisesmith.foo

14:02 haha

14:03 dysfun: i wish i was joking. i think :)

14:03 justin_smith: dysfun: in fact this is what lein does for you if you run "lein new irresponsible/emotional"

14:03 dysfun: well i don't apply the group to all of my things

14:04 things that i don't expect anyone to make use of

14:04 sometimes these get promoted to projects i feel are worth sharing

14:04 this project started off as "wildebeest" heh

14:05 justin_smith: there's also src/foo/foo.clj if you don't want org in the package name, but I like the java style package names more and more too

14:05 dysfun: i like minimal

14:05 justin_smith: but .core isn't minimal, it's just arbitrary

14:06 dysfun: mostly because i am a bear of little brain and i like to be able to understand my code afterwards

14:06 yes, it's entirely arbitrary. and commonly used

14:06 justin_smith: dysfun: but we already established .core doesn't provide information - project by project it might be the core definitions other namespaces use, or the top level ns that clients access

14:06 it doesn't add info, it adds ambiguity

14:07 dysfun: maybe i can rename it emotional.to-the-core for you

14:07 justin_smith: haha, sorry, I know everyone uses .core and not just you, so I don't mean to pick on you in particular

14:07 just giving my case against this convention

14:08 dysfun: i think i was hoping for something a bit more gripping than conventions

14:08 amalloy: justin_smith: do you actually own noisesmith.org? pretty sweet website there

14:08 justin_smith: amalloy: WIP :P

14:08 amalloy: I had plans, then I got a job

14:09 hfaafb: did you make it with omnext

14:09 justin_smith: hfaafb: lol

14:09 dysfun: that's so web 0.1

14:09 ystael: that's my school of web design

14:09 * justin_smith is tempted to replace it with a rant about .core namespaces.

14:10 * dysfun thinks justin_smith must have a pretty sweet life that this is the most annoying thing to him

14:10 justin_smith: dysfun: I can't lie, life as a professional clojure dev is pretty awesome.

14:11 I have other things I could rant about, but they would be so off topic on this forum...

14:11 dysfun: concur

14:11 although i'm startuping, so i don't get vast quantities of money deposited in my bank account every month

14:15 spuz: justin_smith, whereabouts are you based?

14:15 (just wondering what companies are doing clojure out there)

14:15 justin_smith: spuz: PDX, if you are a clojure dev who can do frontend looking for work send me a DM

14:16 spuz: where is PDX?

14:16 dysfun: you should use my frontend library

14:16 justin_smith: spuz: out here Walmart, Staples, Puppet are all using clojure, plus many a scrappy startup like mine

14:17 spuz: 'here' being?

14:17 justin_smith: spuz: sorry, PDX is the airport code for Portland, Oregon

14:17 spuz: ah right

14:17 dysfun: we'll be hiring a clojure developer this year in amsterdam

14:17 spuz: i'm based in london but it's interesting who else is doing clojure

14:17 dysfun: i was in london until last month heh

14:18 there are a few clojure-using companies in london, but it's not a common tech yet

14:19 spuz: nope

14:20 am currently brushing up on clojure to apply for a position

14:20 dysfun: the only way to change that is to use it and build software in it

14:21 i find it a bit miserable how many scala positions there are relative to clojure ones. i wish they were more comparable in popularity

14:25 justin_smith: i've just created a new core.cljc, thought i'd let you know :p

14:25 justin_smith: haha

14:25 * justin_smith rubs his hands, scheming, while runing "lein new core.core.core"

14:26 * dysfun creates core/core/core.cljc

14:27 dysfun: core.clj/core.cljs/core.cljc # core.clj.core.cljs.core

14:28 justin_smith: lol

14:28 dysfun: ^{:doc "Because calling a module 'core' annoys justinsmith"}

14:29 * dysfun defns core

14:31 ajb: How would you get a value that is in a 2 level nested map while using another top level value from that map as a key for it?

14:31 justin_smith: ajb: with get-in

14:31 ajb: e.g. {:word "foo" {:terms {:foo "testing :bar "more testing"}}}

14:32 hfaafb: he only has the value

14:32 ajb: while using :word as the key to access the string "testing

14:32 hfaafb: oh v0v

14:32 ajb: v0v?

14:33 hfaafb: shrug emote

14:34 justin_smith: ajb: that's not a valid hash-map, what would the actual map look like?

14:34 ajb: same thing but with the string testing a full string, typo'd and forgot to include the closing quote

14:34 so {:word "foo" {:terms {:foo "testing" :bar "more testing"}}}

14:34 justin_smith: what?

14:34 clojurebot: what is short for ,(doc ...)

14:35 amalloy: that doesn't fix the issue ajb

14:35 justin_smith: that's still not a valid hash map

14:35 amalloy: {:x 1 2} is not good

14:35 ajb: OH, sorry, wasn't reading it correctly

14:35 so {:word "foo" :terms {:foo "testing" :bar "more testing"}}

14:36 justin_smith: ,((fn [m] (get-in m [:terms (keyword (:word m))])) {:word "foo" :terms {:foo "testing" :bar "more testing"}})

14:36 clojurebot: "testing"

14:36 justin_smith: I mean I don't know if that helps or not

14:37 amalloy: also, strongly recommend against using keywords for keys if it means you'll be converting string<=>keyword to look things up. it sounds like your terms are really just strings

14:37 justin_smith: I would either use a string as the key in the nested map, or use a keyword as val under :word to avoid the silly keyword conversion though

14:37 exactly

14:37 keywords are not a magic type for hash map keys

14:38 ajb: so, just use the string?

14:38 justin_smith: ,((fn [m] (get-in m [:terms (:word m)])) {:word "foo" :terms {"foo" "testing" :bar "more testing"}}) ; yeah, like this

14:38 clojurebot: "testing"

14:45 justin_smith: dysfun: also on my annoyances list, using keywords as keys in maps when you already had data that made sense (especially horrific keywords like :1 ...)

14:46 dysfun: yeah, there's no accounting for taste

14:46 yes, keywords are cool, but we got over them years ago, not *everything* must be one

14:47 i do like them

14:48 justin_smith: ,(def annoying-things {:1 (keyword "core cargo culting")})

14:48 clojurebot: #'sandbox/annoying-things

14:48 justin_smith: ,annoying-things

14:48 clojurebot: {:1 :core cargo culting}

14:48 amalloy: ,(keyword 1)

14:48 clojurebot: nil

14:49 tolstoy: I think that works in ClojureScript (or did at one time). Which caught me "keywording" UUIDs.

14:49 dysfun: ,'((((unnecessarily :nested))))

14:49 clojurebot: ((((unnecessarily :nested))))

14:49 justin_smith: tolstoy: in my repl it returns nils

14:50 tolstoy: for both longs and uuids

14:50 (in my cljs repl that is)

14:50 tolstoy: Ah.

14:50 Maybe it's fixed, now?

14:51 justin_smith: I mean it doesn't fail, it just happily returns nil

14:51 clojurebot: Excuse me?

14:51 tolstoy: Seems like one of the Clojure's allowed it.

14:51 justin_smith: which would still be a bug

14:51 depending on how you go on to use the value at least

14:51 dysfun: ,(-> 1 str keyword)

14:51 clojurebot: :1

14:51 tolstoy: Anyway, apropos of inapproprite keywording, I was keywording UUIDs which mostly worked, until one of them starting with a number.

14:52 justin_smith: haha, must be an old cljs behavior then

14:52 tolstoy: Could be. It was a long time ago.

14:52 dysfun: that's a 10/16 chance. not good odds

14:54 tolstoy: Hm. Keyword (uuid) in regular clojure works just fine.

14:54 :17faa74a-c108-4139-aa5e-1e09d8f004b2

14:54 justin_smith: tolstoy: happily returns nil, I mean that won't be correct code

14:55 wait, what clojure version does that?

14:55 tolstoy: 1.8

14:55 justin_smith: ,(keyword (java.util.UUID/randomUUID))

14:55 clojurebot: nil

14:55 justin_smith: tolstoy: citation needed

14:55 was it actually a keyword?

14:55 tolstoy: ,(keyword "2")

14:55 clojurebot: :2

14:55 justin_smith: tolstoy: I mean was it actually a uuid?

14:55 tolstoy: (keyword "17faa74a-c108-4139-aa5e-1e09d8f004b2")

14:55 ,(keyword "17faa74a-c108-4139-aa5e-1e09d8f004b2")

14:56 clojurebot: :17faa74a-c108-4139-aa5e-1e09d8f004b2

14:56 dysfun: yes, that's a string

14:56 Bronsa: ,:2/consistency

14:56 clojurebot: :2/consistency

14:56 justin_smith: that's not a UUID, that's a string

14:56 tolstoy: Right.

14:56 Bronsa: ,:2/3

14:56 clojurebot: #<RuntimeException java.lang.RuntimeException: Invalid token: :2/3>

14:56 dysfun: strings will keywordify quite easily

14:56 tolstoy: So, keywords can begin with a number?

14:56 RedNifre: Clojure newbie here. If keywords can be arbitrary strings, why would I use them instead of strings?

14:56 Bronsa: tolstoy: no pls

14:56 dysfun: the problem isn't keywords, it's that the keyword function doesn't convert numbers to keywords

14:56 Bronsa: it's just an accident

14:57 dysfun: why should it?

14:57 tolstoy: Bronsa: Right. But it "works" in Clojure, not in ClojureScript?

14:57 dysfun: Bronsa: i didn't say it should :)

14:57 Bronsa: tolstoy: a lot of things "work" in clojure, doesn't mean you should use them

14:57 justin_smith: RedNifre: keywords should be used when the data "stands for itself" and isn't meaningful as a string.

14:57 RedNifre: if it exists only as a key in a map or a special arg to a function, and you won't need string operations

14:58 tolstoy: Bronsa: I'm not at all advocating that we should. In fact, I was supporting the idea that not everything should be a keyword by mentioning how I got screwed.

14:58 justin_smith: RedNifre: also, the keyword function accepts all kinds of crap that isn't valid in keywords, and will generate invalid keywords, it does no validation for correctness

14:59 Bronsa: tolstoy: sorry, I chimed into the conversation after it had started and I missed that :)

14:59 justin_smith: ,::1 :P

14:59 clojurebot: :sandbox/1

14:59 Bronsa: :(

14:59 justin_smith: yeah, it's sad that that works

14:59 tolstoy: Bronsa: I don't have a CLJS repl handy, but it might actually balk at (keyword "2akjdas").

15:00 justin_smith: tolstoy: my cljs repl thinks that is just hunky dory

15:00 Bronsa: it doesn't

15:01 justin_smith: ~keyword is gigo

15:01 clojurebot: You don't have to tell me twice.

15:01 tolstoy: Okay. Maybe it used to? Anyway, I swear I was bitten using keywords build from the string representation of a UUID.

15:02 justin_smith: ,(keyword (pr-str #uuid "a02a5e9a-34bf-4175-b0fa-76b5a6597a87")) omgbarf

15:02 clojurebot: :#uuid "a02a5e9a-34bf-4175-b0fa-76b5a6597a87"

15:02 RedNifre: justin_smith I'm not sure I understand. I interpreted keywords as values of a global enum, so I'm surprised that you can create them dynamically with the keyword function. When would you turn strings into keywords or create keywords programmatically?

15:03 justin_smith: RedNifre: people like to do this when consuming json. I actually think this is a terrible idea since many valid json keys are not good keywords, but it's popular to do

15:04 tolstoy: I bet it's popular because it just looks better in the code. (:id (str->json blob))

15:04 justin_smith: RedNifre: my take is that when doing this people are letting aesthetics or an odd sense of convention eclipse common sense

15:05 dysfun: i actually quite like it, because if the structure has been validated, you can easily query it (keywords as functions)

15:05 but in this case you're free to only choose keywords that make sense

15:06 RedNifre: I guess it's a good idea if you are sure that the JSON keys can be valid clojure keywords, huh?

15:06 dysfun: as always, check your input

15:06 justin_smith: RedNifre: there's room for intellegent developers to disagree there.

15:07 it's also a bunch of work done for a small benefit, creating a whole mess of keywords which takes more time than parsing the json does, just so you can use one or two of them in pretty code.

15:08 RedNifre: I'm a clojure beginner so I don't see the problem with turning JSON objects into clojure dictionaries where the keys are keywords, please enlighten me :)

15:08 dysfun: rage against the machine, justin_smith

15:08 justin_smith: RedNifre: because there are valid json keys that are not valid clojure keywords, and generating all the keywords takes more time then parsing the json does

15:09 but like I said, there's room for intelligent developers to disagree here, so whatever

15:09 tolstoy: RedNifre: For me, anyway, the only problem with keywording json docs is when doing so is more than just a flag on your converter.

15:09 dysfun: damnit, i was just making popcorn

15:10 RedNifre: Nah, I agree that it's bad if there are JSON keys that aren't valid keywords. That's why I qualified it "when you are sure"

15:10 tolstoy: RedNifre: Like undoing CamelCase to camel-case, etc, etc.

15:10 dysfun: tolstoy: camel-snake-kebab

15:12 RedNifre: Why does clojure use the dash style instead of camel case?

15:12 dysfun: because it's easier on the eyes

15:12 justin_smith: RedNifre: historical reasons, and we think it's more readable

15:13 tolstoy: IMHO, because it's easier to read, but also because it can. No "infix" notation.

15:13 foo-bar doesn't mean foo minus bar.

15:13 dysfun: the range of characters allowed in symbols helps make clojure very readable

15:13 justin_smith: RedNifre: LISP was originally meant to be a parsed internal representation that programmers would not use directly, so it doesn't really support much syntax

15:14 and we borrow heavily from that tradition

15:14 RedNifre: Personally I find it harder to read because the dash looks more like a space than a camel case capital letter does but maybe I just need a font with thicker dashes.

15:14 justin_smith: RedNifre: or it could be you are used to languages where - is not word-constituent?

15:14 tolstoy: My issue with CamelCase is its easy to FatFInger.

15:15 justin_smith: I find it kind of headache-inducing to read

15:15 it's telling that we don't write that HeadacheInducing

15:15 dysfun: justin_smith: i use camelcase for protocols and records. feel sick yet? :p

15:16 justin_smith: dysfun: that's where camelcase is expected

15:16 dysfun: since you are naming java classes

15:16 anything else would be annoying actually

15:16 dysfun: damnit, you're supposed to disapprove

15:16 RedNifre: (a-b c d-e-f g h) vs (aB c dEF gH) it's easier for me to see that the second list contains 4 things but you're right, it could be a habit.

15:17 justin_smith: the first one contains 5 things

15:17 dysfun: except they're rarely two characters long

15:17 sdegutis: justin_smith: what do you think of this idea?

15:17 dysfun: real-world example i'm writing just now. PassLocked vs pass-locked

15:18 sdegutis: (ns myapp.routes.order.orderid%) (defn invoice [request] ...) ;; becomes "/order/:orderid/invoice"

15:19 Or how about (ns myapp.routes.order) (defn orderid%invoice [request] ...)

15:19 justin_smith: sdegutis: I don't like it at all!

15:20 sdegutis: Why not?

15:20 justin_smith: magic

15:20 sdegutis: It wouldn't be dynamically found out. It would be, you'd have 'myapp.routes.order in your list of routes, which finds the namespace object and gets the info from that.

15:20 dysfun: yeah but it's specialcase for things that followed the expected rest routing

15:20 justin_smith: sdegutis: that's still magic

15:21 dysfun: and you always get edge cases in real world systems

15:21 sdegutis: It just feels too repetitive though, to always type (ANY "/order/:orderid/invoice" [request] ...) when I'm already in the myapp.routes.order namespace.

15:21 justin_smith: would creating the route automatically require?

15:21 dysfun: i mean if it were a macro i'd be okay with it

15:21 justin_smith: both yes and no lead to problems

15:21 sdegutis: justin_smith: nothing is automatically required

15:21 but yeah I think I see your point

15:22 dysfun: sdegutis: there is a macro for that in compojure as well, context

15:22 nkhodyunya: sdegutis: which library is that?

15:22 dysfun: so it's down to once

15:22 sdegutis: dysfun: it's more that I don't want the repetition between the namespace and the route path.

15:23 dysfun: but in my experience it's handy to be able to throw a few isolated handlers into one namespace for organisation

15:23 irrespective of url

15:24 there's a point at which something turns from a library to a framework. i like the former, i grudgingly use the latter

15:24 justin_smith: sdegutis: in my experience silly things like client aesthetic choices mean that routes change, and I don't want that to have any effect on my code, ideally all info about routes is totally separated from the code itself and I use bidirectional routings to figure out where my endpoints go

15:24 dysfun: i think it's the point where you invert control

15:25 sdegutis: justin_smith: good point

15:25 dysfun: justin_smith: because obviously /node/3456 is a great url!

15:26 justin_smith: dysfun: ?

15:27 sdegutis: justin_smith: thanks for helping me see that more clearly, I dunno what's wrong with my judgment today :/

15:27 dysfun: aesthetic choices in urls are potentially valid usability concerns

15:28 justin_smith: dysfun: right, I have a huge respect for that, which is why I want URL structure to be totally decoupled from my code's logic, so that we can swap endpoint names arbitrarily without breaking code

15:29 names, structure, the whole deal

15:29 dysfun: 'silly things' was what i was objecting to

15:30 justin_smith: dysfun: oh, the decoupling is implemented because it could increase usability, but in practice the changes are silly and rarely actually do increase usability, that's just the cynical side coming out

15:31 dysfun: oh i know, i used to work in marketing and you know what marketing people are like ;)

15:31 justin_smith: but it is true, I was likely overly dismissive there

15:31 haha

15:31 dysfun: my product is a tool for marketers, so :P

15:31 dysfun: fun fun fun :)

15:31 any of them want to help a struggling startup for free?

15:35 Shayanjm: justin_smith: I figured it out and this kafka-clj thing is pretty sweet

15:35 justin_smith: Shayanjm: interesting.

15:35 Shayanjm: it has some assumptions baked in about how you use kafka (i.e: message frequency) for optimization purposes

15:35 but nothing you can't fiddle with

15:36 that's why I was experiencing hanging earlier - a few settings regarding message chunking on the consumer end

15:36 justin_smith: regular kafka has a config for how long you buffere messages to combine then on a topic

15:36 that's a pretty larg default buffer

15:37 is it configured by size or time?

15:37 Shayanjm: Size with a timeout I believe

15:37 so if it doesn't hit size x by time y it still fires off

15:38 it's # of messages on 'size'. It basically accumulates messages and delivers them in chunks as 'work-units'

15:39 The default out of the box is 100k messages to be consumed per 'work unit'

15:39 justin_smith: woah

15:39 Shayanjm: yeah, I think the author was assuming you don't batch inside the messages

15:40 so each "message" is a basic record

15:41 but you can change that in the consumer config, I just set it to 1 for testing purposes. Fired off two messages and could iterate over a lazy-seq to find them

16:08 zipper: Hey a good resource for learning clojure coming from haskell?

16:08 justin_smith: zipper: Joy of Clojure is good, as is Clojure Applied

16:08 zipper: once you grok the non-syntax it should come pretty easily

16:08 zipper: justin_smith: Thanks

16:09 non syntax?

16:09 You mean like python?

16:09 justin_smith: zipper: python has a lack of delimiters and a lot of syntax, clojure has a lack of syntax and a lot of delimiters

16:10 devth__: not really a clj question, but noticed lein projects follow the convention: why are docs markdown files UPPER_CASE.md ? some gnu convention?

16:10 justin_smith: zipper: point being that (f arg another-arg) is the structure of just about everything in clojure, with a small set of exceptions that don't take long to learn.

16:11 sure, we have syntax, but we get as close to not having one as you can get

16:11 sdegutis: Hi.

16:11 Is anything shaking?

16:13 zipper: justin_smith: Okay that's interesting. So delimiters == brackets

16:13 justin_smith: () {} [] right

16:13 zipper: justin_smith: It's a lisp :D

16:14 justin_smith: zipper: right this whole delimiters instead of syntax thing is something we borrow from lisp

16:14 the origing LISP was meant to be a direct representation of program structure, and not a syntax humans would use, but as we all know usage evolved otherwise

16:15 so the tradeoff is that due to using delimiters explicitly it is very easy to use and manipulate data literals representing source code, the whole macro thing

16:18 zipper: I guess after learning clojure reading SICP will be easy :)

16:19 justin_smith: zipper: in fact people have translated parts of SICP to clojure- but SICP is not especially functional and doesn't expose the interesting parts of clojure

16:20 seubert: justin_smith: what are the interesting parts

16:20 that you're referring to, I mean

16:20 that's a broad question :P

16:20 justin_smith: seubert: using immutable values across multiple threads

16:21 being able to use persistent data structures where "modification" doesn't alter the original, nor does it require doing a full copy of the original

16:23 seubert: and pragmatically, if clojure is used well, large code bases where you don't have to deal with "spooky action at a distance" - things that would effect the result of a given function are explicitly visible where the function is defined and used and you don't have to go hunting for magic

16:23 seubert: I see

16:23 sdegutis: 3spooky5me

16:34 justin_smith: sdegutis: you have been spooked by a spoopy clojure, :DOOT

16:34 sdegutis: :D

16:36 sarcher: Is clojure well suited for writing code that reads from / writes to sockets?

16:37 justin_smith: sarcher: we've got some good libs for it

16:37 sarcher: I've dabbled in clojure from time to time, but it hasn't "stuck" yet.

16:38 justin_smith: eg. ztellman's gloss for packing data into streams of bytes, and manifold for pushing those streams around

16:38 sarcher: I haven't had the aha moment.

16:38 justin_smith: OK

16:38 sarcher: what have you wanted to do with raw sockets in clojure?

16:38 sarcher: I work for a company that has a socket-based API for real time data.

16:39 I'd like to write a consumer that receives that data, maintains state based on it, and ultimately displays it on a web page.

16:39 rhg135: even plain interop isn't too bad if not doing nio

16:39 sarcher: I have a working version written in node, but i'd like to implement with clojure to learn more about the language.

16:40 Technically written in javascript on node.

16:40 justin_smith: so two node instances communicating over sockets?

16:41 sarcher: Something like that, yes

16:56 sdegutis: ,(let [[a b :as c] nil] [a b c])

16:56 clojurebot: [nil nil nil]

16:58 noncom|2: i am creating a 1-page app with clj+cljs and i have the "#/page" style nav in the app. ok, now i want to send notifications to users which have links to access a specific "#/page/1234-some-id-1233" how do i do this?

17:00 if i just give them regular link, like "http://address/page/id" then the rest of the in-app nav goes crazy with "http://address/page/id/#/the-rest-of-the-nav-system-is-now-here" instead of "http://address/#/the-correct-nav-behavior"

17:08 justin_smith: noncom|2: link to http://address#/page/id

17:11 noncom|2: yes, probably you are right.. i have to rewrite page allowance mechanism to allow this

17:14 justin_smith: hmmm for some reason the reagent.session does not recognize the page.. it says the page is nil, though i have "/invitation/:id" in the routes now

17:15 justin_smith: noncom|2: are you sure the router is set up to route on the fragment?

17:15 noncom|2: it does ok for in-app nav, but simply going to a in-app link from nowhere lands me on page nil which in my app redirects to login

17:17 justin_smith: hmmm, well, i have it specified in the routes just the same way as the rest of the routes...

17:18 justin_smith: oh, it works

17:18 cache problems again!

17:18 damn, there seems to be simply NO WAY to ensure that page cache is cleared...

17:19 i press the ctrl+f5 for the 5th time and miracously on the 5th time it decicdes to finally clear the cache and rebuild and reload all

17:25 justin_smith: with figwheel it will push your new code as you save it

17:49 sdegutis: Hi.

18:16 justin_smith: what do you think of (for) taking an optional final argument which is executed when the seq is empty and there's nothing to enumerate?

18:16 (for [user users] (user/email user) ["n/a"])

18:16 I don't like this idea but I want to hear what you think.

18:16 Because who knows, maybe I should like it?

18:27 arrdem: silly idea.

18:28 for is a lazy/pure sequence comprehension expression.

18:28 "on the first element" "on the nth element" "on the last element" are pretty imperative concepts/states

18:28 no doubt you could do it with CL's for macro, but that macro has very very different goals.

18:35 CaptainLex: arrdem: I am on the edge of my seat!

18:38 arrdem: CaptainLex: uh... why?

18:38 CaptainLex: arrdem: I am waiting for the silly idea! Unless you unveiled before I logged on and you were just explaining it retroactively

18:38 justin_smith: arrdem: because that's next to "the last element" which he was on before, I think

18:39 CaptainLex: Ahhhhh okay

18:39 I just walked in at the wrong time

19:02 zoite: hi, i'm trying to use korma to select items in a table but with the were get populated from a loop but it doesn't seem to be making the query right and was checking if anyone could point me in the right direction? http://pastebin.com/tzMnwh9N

19:03 justin_smith: zoite: map with one arg is probably not what you want here

19:04 zoite: Ok, sorry I'm new to clojure, what should I be using instead?

19:04 justin_smith: zoite: noticfe in the dry run that where you should have a parameter vector for the parameterized query you instead have a function, returned by map when it only has one arg

19:04 zoite: you should provide a second arg to map, the collection it should map across to create your query

19:05 ,(map inc)

19:05 clojurebot: #object[clojure.core$map$fn__4781 0x19d34ba0 "clojure.core$map$fn__4781@19d34ba0"]

19:05 justin_smith: ,(map inc [1 2 2])

19:05 clojurebot: (2 3 3)

19:05 justin_smith: that's the difference

19:05 zoite: oh ok

19:05 I just need to loop over the key/value. originally I tried a for loop but then I learned that doesn't return anything

19:06 justin_smith: if we were like racket we would have a "beginner mode" where map always needs at least two args :0

19:06 amalloy: probably exactly two tbh

19:06 justin_smith: zoite: for isn't a loop, and it does return something, in fact it is designed for returning lists

19:06 amalloy: yeah, makes sense

19:06 zoite: oh

19:07 justin_smith: zoite: doseq on the other hand, is for side effects and returns nil (but still isn't really a loop)

19:07 zoite: oops yeah I meant doseq not for

19:07 justin_smith: right - changing the doseq to for might have been simpler (probably would have given you easier to read code)

19:08 but map works too if you provide the args needed

19:08 amalloy: speaking of map int, last night i introduced some clojure guys at work who are doing js at the moment to the beauty of ["1", "23", "10"].map(parseInt)

19:08 if you're not excited enough to run that in your chrome console, the result is [1, NaN, 2]

19:08 justin_smith: haha, wow

19:09 what the hell is that even doing?

19:09 amalloy: make up some more example inputs and try to figure out the pattern

19:11 hint: an exciting slight change to the input: ["1", "23", "10", "112233445566"].map(parseInt) // [1, NaN, 2, 44]

19:11 hfaafb: :D

19:12 justin_smith: I'm still not seeing it, sorry...

19:14 amalloy: well, it's a puzzle! i can't give out spoilers

19:17 hyPiRion: amalloy: oh, that is just plain evil

19:17 amalloy: hyPiRion: my posing the puzzle is evil, or you found the answer and it's evil?

19:18 hyPiRion: amalloy: the answer

19:18 amalloy: srsly

19:24 zoite: justin_smith: I tried changing it to use 'for' but it's giving back clojure.lang.LazySeq, not sure how to get past that http://pastebin.com/vjhE86j0 sorry

19:25 hiredman: zoite: you are creating a collection of things, and the dsl just doesn't know how to handle a collecction of things like that, so it doesn't matter how you create the collection

19:25 sdegutis: arrdem: good point

19:25 justin_smith: zoite: you changed a couple of things there I think - wasn't there an "and" before? also ((first x) [like (second x)]) is not likely to return anything useful

19:25 zoite: hiredman: oh, so it's not going to work?

19:25 justin_smith: ya, I added in the and and it just gave back: "SELECT `test`.* FROM `test` WHERE ((NULL, NULL))"

19:26 sdegutis: Datomic is so cool.

19:26 hiredman: if you look at your previous usage of the dsl, when you pass literal things to the dsl (and {...} {...}) you are not passing a collection of maps

19:27 zoite: well i'm trying to turn the loop into how I ran it manually, just not sure how to do it

19:27 hiredman: you need whatever the dsls version of 'apply' is

19:29 zoite: i'm looking through the documentation but I don't think it has one

19:37 rhg135: does lein trampoline not start nrepl?

19:45 justin_smith: rhg135: I don't think it does, no, with nrepl you need the two jvms, client and server

19:46 err... wait no it's saying nrepl in the boot splash

19:47 rhg135: yeah... that's what confuses me

19:47 justin_smith: but also it seems not to have opened a port

19:47 so that part of the splash is lying?

19:47 rhg135: exactly

19:47 * rhg135 looks at suspiciously

19:48 justin_smith: rhg135: lein trampoline repl :headless opens a port and then provides no repl, as expected

19:48 (I mean you would get a repl on that port, but not in that terminal)

19:49 rhg135: oh cool then, thx

19:49 this should be documented somewhere

19:50 justin_smith: rhg135: yes, it should

19:54 rhg135: meh, I prefer to launch a client even if it takes much longer

19:59 justin_smith: the part that's strange is what is it doing normally? afaik you can't use nrepl without a server.

19:59 the splash is a lie

20:00 justin_smith: yeah, I think the splash is a lie

20:00 rhg135: it would be cool if trampoline repl made it start the server and client, both in the same vm

20:01 which more closely matches the expected features of both the trampoline prefix and the repl command

20:01 * rhg135 dives head-first into lein's code

20:13 turbofail: i made a thing in clojurescript

20:13 http://funkenblatt.github.io/coriolis/

22:28 marcfontaine: is there a more functional way to write the loop ? https://gist.github.com/marcandrefontaine/103e4d2200e19eb6d5a8

22:42 justin_smith: marcfontaine: the only thing I would fix is the dangling parens

22:45 marcfontaine: justin_smith: oups, thanks.

Logging service provided by n01se.net