it's not just a convenience - being able to do something in a local scope is an important feature - the semantics are different

it's not a primitive.

is there a way to group related libraries under a single name?

i completely agree.

@bcbradley what do you mean? spreading ns stuff across multiple files?

hrm, no nothing like that

clojure.set/blah, clojure.string/blah

you are aware of https://github.com/ztellman/potemkin?

is clojure. just part of the name?

or is it like the major thing in maven? I want to say something like lwjgl-clojure.package-name/function-name

bcbradley there are no rules about package names, except for how they map to the class path

package name doesn't imply access the way it does to javac

and it need not be related to the artifact or group defining it

(but good practice is to use package names that make sense of course...)

so you are saying i could use the name lwjgl-clojure.glfw for one "package" aka "library", and lwjgl-clojure.openal for another?

I don't know how that would make them separate libraries, but OK

@bcbradley sorry, i don't follow you. please post the code that is giving you trbl.

each of them have dozens of classes and i planned on making each class a namespace

https://javadoc.lwjgl.org/org/lwjgl/opengl/package-summary.html

you can nest packages however you like

for instance each class there would be a namespace-- according to what i had planned to do

you can have as many packages as you want in one library

(or as few)

hold on i'm a bit confused now

i thought a clojure library was a collection of namespaces

right, and each namespace has a package

the package is the prefix leading up to the namespace name

i thought each namespace was a collection of symbols (function defs for instance)

Um no sorry

when functions inside a namespace are compiled, they become classes, and that defines the package those classes will be in

Clojure really has no concept of packages.

@bcbradley : what exactly is causi g trouble. please post code so we do not need to guess.

Maven might, but Clojure just looks for classes on the class path. And files that match the namespace name

this isn't a coding issue its a deployment issue

i'm wondering how i should organize what i'm making, not fretting over a line of code

@tbaldridge but when my function foo is compiled to foo.class the jvm thinks that class is in a package, no?

not that this "package" does much - but it's there

@bcbradley well i'd like to help, but unless you can be more specific i cannot.

here is the gist of it: C has no namespaces so when you look at opengl code you see that every function begins with gl, and its annoying. It wouldn't be too annoying to deal with except that java has no namespaces either (it has classes) and with lwjgl opengl is a package that contains a bunch of classes (for instance gl11), so the fact that both of these languages don't have namespaces forces you to say gl twice if you want to use it through interop in clojure

i want to make it less painful to use these native libraries from clojure, so i'm making a clojure library or set of libraries to strip out the nonsense and just use namespaces (since clojure has them)

@bcbradley just stick it all in one dependency, with 30 namespaces. No reason not to do it that way

(but to be honest, writing gl 4x times really isn't a show-stopper in any GL app I've worked on)

@tbaldridge so each "package" in lwjgl would map to a "namespace" in clojure?

I thought about that but what do the classes become?

gl30 for instance, has the functions for the 3.0 version of opengl

and thats a class

some of the classes have functions with identical names, so i can't just flatten it into one namespace

i figured the only way to do this would be to have each package become its own "library" or dependency, and each class to become its own namespace

that way I don't risk any conflicts

but something about that feels like overkill, does anyone have any suggestions?

why make them separate dependencies? why not have multiple groups of namespaces in one library?

"groups" of namespaces?

is there some way to group up namespaces in one dependency?

you can nest the directory paths however you like if you feel like grouping them

I don't even know why they would need grouping, but sometimes that makes the relationships clearer

i mean i guess i could just make one dependency and have each namespace name be the concatenation of the abbreviated lwjgl package name and class name

do you think that would be better?

rather than multiple dependencies in order to use one library, yes

well its more like a collection of libraries but i see where you are coming from

It seems to me the simplest solution would be something like clojure-gl.gl.10/vertexf

So top level is library name, then "package", aka, GL, CL. Then the version number. And in that namespace are all the functions for that version/subpackage combo

how about lwjgl-clojure.opengl.gl10/vertexf

People can use (:require [clojure-gl.gl.10 :refer [vertexf]) to remove some of the name redundancy

yeah, that works too

based on what noisesmith was saying it would be easier for people to depend on if all the libraries came together under one dependency

the only thing that makes me anxious about that is how i will handle the natives

lwjgl comes with a bunch of native binaries that it depends on, depending on which packages you need to use

Well to be honest, at that point ^^ all you're saving is this: in lwjgl I can simply do (:import [.... GL]) then do: GL/glVertexF which not only matches all the OpenGL documentation (that's huge), but isn't that much writing either

the extra gl is annoying, besides i've already produced the data structure that describes lwjgl

its not just the gl

But one more point: Any serious GL work is going to use shaders and vertex buffers, so you end up writing GL even less...and you write it once, and then you're done.

i spent some time going through the entire library and making a record of all the annoying repetitive namespaces being prepended to the function names

Personally I don't find library that reduce a few characters in typing that useful.

the other thing i wanted to do is have a clojure library for this stuff where you could pull up the documentation in clojure

besides, using java interop for this stuff can be annoying too once you start to try to build abstractions on it, often you have to wrap the java interop in an anonymous function to use ifN

But you have to consider this: going agains a few decades of OpenGL programming guides isn't going to be that helpful. I would rather be able to download the NeHe tutorials for Java and just copy most of them

bleh blaze a trail

i'm already halfway there

i realize most people wouldn't go through all this just to sanitize some names and i guess it is a lot of work for something trivial

in any case its already basically done, so that isn't really a valid concern (would have been a few weeks ago)

so if each package has a native dependency associated with it, would i be better off making each package a separate library?

is there something like group-by crossed with index? Like group-by, except instead of {:key [item]}, {:key item} (I know keys are unique.)

Normally I do (into {} …) but it’s not super clear to me that that’s optimal 🙂

I usually do (defn index-by [k coll] (into {} (map (juxt k identity)) coll)) - uses a transducer, it's a small bit of code, it works

definitely not something that exists in clojure.core

yep, that’s what I was about to do. Thanks 🙂

note the usage of map transducer instead of calling map on the coll

I’m writing here because #clojure-sanfrancisco seems to be deserted. Anyone in SF Bay Area? Are there any meetups here?

@bcbradley I don't follow everything in above discussion but seemed like inadvertent 3-way misunderstandings to me. Noisesmith's point is that you can add as many dots to a namespace name as you want, as long as you put source files in a corresponding directory structure. If you have a lot of namespaces that will give them more order, but won't do anything else. Maybe that's already clear. Maybe that's all you need? Or is the only misunderstanding mine?

is it possible to filter maps using specter? (and get another map)

ido yes, use setval

dev=> (setval [MAP-VALS (selected? even?)] NONE {1 1 2 2 3 3 4 4})
{1 1, 3 3}

@schmee I was expecting select will do this

I’m trying Lacinia, the GraphQL library from Walmart. Where is the best place to ask questions about it?

#graphql

(I am doing the same :))

Is anyone doing clojure app clustering? What do you use for remote logging/monitoring instances? Doesn’t matter if it’s a commercial service

@singen we've been using ELK for logging... pipe stdout logs to host (mounted) syslog in docker container -> logstash-forwarder -> logstash -> elasticsearch+kibana

@ido @schmee no need for selected? there, [MAP-VALS even?] is equivalent

@singen ELK here too, all logging to stdout.

discussed between logstash and fluend

Guys, I need help. It’s really important. 😞

Is there someone ?

just ask your question

We have docker swarm system.

I want to run live/repl mode in docker swarm. When I try to run lein repl swarm service with this code; docker service create -w /olexibus --mount type=bind,source=/root/app/corporate/olexibus,target=/olexibus -p 3000:3000 --name corporate --network olxnet pandeiro/lein repl I get full repl initialization while repl input turn. It’s says "Bye For Now” like end of repl session. But I didn’t enter anything.

Why leiningen doesn’t initialize normally ?

You probably need to allocate a pseudo-terminal (with -t). Otherwise, the repl doesn't have stdin.

fwiw. This is a docker topic, not specific to Clojure. https://docs.docker.com/engine/reference/commandline/service_create/#add-bind-mounts-or-volumes

It read my project file and tried to start repl, tried to get deps.

is there some obvious reason that there is no map-zipper in clojure.zip that I’m missing?

seems like xml-zipper is a special case of a map zipper

I'm having trouble java-interopin' these chained calls: String logoutUri = UriBuilder.fromUri("http://localhost:8180/auth") .path(ServiceUrlConstants.TOKEN_SERVICE_LOGOUT_PATH) .queryParam("redirect_uri", "http://localhost:8080/product-portal") .build("Realm") .toString(); to clj: logoutUrl (-> (UriBuilder/fromUri "http://localhost:8180/auth") (.path ServiceUrlConstants/TOKEN_SERVICE_LOGOUT_PATH) (.queryParam "redirect_uri" "http://localhost:8080/product-portal") (.build "Realm") (.toString)) I get "java.lang.ClassCastException: java.lang.String cannot be cast to [Ljava.lang.Object;" (on line 2)

fbielejec that makes me think one of those calls is actually varargs

to do varargs, you need to put all the “extra” args into an array of the right type (in this case it wants Object, clearly)

(into-array Object [s1 s2 …])

I assume queryParam is vararg, from context

or maybe path is

@noisesmith it's a bit hard to debug, becasue I uberwar and deploy to a container (wildfly)

ok - it’s build that is vararg

change (.build “Realm”) to (.build (into-array Object [“realm”]))

http://docs.oracle.com/javaee/7/api/javax/ws/rs/core/UriBuilder.html

oh, and the same change is needed for queryParam

- oh, that’s a little different (.queryParam “redirect_uri” (into-array Object [“]))

hmm, still not happy, same error thrown

I could actually create a proj inEclipse and inspect, now that I think 🙂

fbielejec try adding items to the -> in your repl, one of those lines at a time

didn't try your second suggestion

it should need both, according to the api docs

clojure interop with java is great, except varargs are kind of clumsy (and inheritance …)

works, beauty 🙂

many thanks!

glad I could help

I wonder how hard it would be to make something like ((varargs foo object-instance arg arg2 type) arg arg arg) and turned it into (.foo object-instance arg arg2 (into-array type [arg arg arg]))

Have a luminus re-frame app, and looking to let someone run it who is new to clojure... is there a way to get a lein run that includes all the cljs steps?

that’s not an ideal syntax I guess - but there should be a nicer way to do it

albaker you can define the :prep-tasks for the profile run uses, or have them do lein do cljsbuild once, run

ah cool, thanks

lein help sample has examples for setting up :prep-tasks

@ddellacosta wouldn't you use seq-zip most of the time?

@bja I must be missing something because that doesn’t seem to know what to do with the map I’m passing in

@ddellacosta no I'm missing something

yeah, I assume that seq-zip wants a seq

would you expect a map-zip to be (zipper map? vals ...)

@fbielejec I was thinking about how clumsy varargs can be - this macro appears to work

=> (defmacro vararg
      [method object & args]
      (let [[regular [_ t & variable]] (split-with #(not= % '|) args)]
        `(. ~object ~method ~@regular (into-array ~t ~(vec variable)))))
#’user/vararg
=> (macroexpand-1 '(vararg get java.nio.file.Paths "foo" | String "bar" "baz"))
(. java.nio.file.Paths get "foo" (clojure.core/into-array String ["bar" "baz"]))
=> (vararg get java.nio.file.Paths "foo" | String "bar" "baz")
#object[sun.nio.fs.UnixPath 0x14ebb640 "foo/bar/baz”]

@bja I’ve been using this definition FWIW: http://stackoverflow.com/a/15020649, it does exactly what I need

seems like the obvious implementation in terms of how it behaves. Just curious why that or something like it isn’t included by default.

@noisesmith nice, I'm saving it

let me know if it fails in some unexpected way - for simple stuff it works so far though

I’m a little skeptical of 10 minute old macros on principle 😄

I like that macro @noisesmith

done stuff like it with ... instead of the pipe

oh yeah, … is probably more readable there

easy enough to swap in

I wonder how it handles hinting

CLJ-440 has been open for years. Probably should be closed IMHO unless someone can find some new vararg interop syntax that has 0 implications on existing code

Is it appropriate to ask the idiomatic way to do something in clojure here? I've only been doing funcitonal/clojure programming for about a week and am trying to improve on how some of my code looks/works.

@greg_arcara always better to just ask the question 🙂. But sure, here is a good place or also in #beginners

@tbaldridge #beginners sounds like a good place for me hah

thanks

can someone help me translate this psuedocode to Clojure?

01 function negamax(node, depth, color)
02     if depth = 0 or node is a terminal node
03         return color * the heuristic value of node

04     bestValue := −∞
05     foreach child of node
06         v := −negamax(child, depth − 1, −color)
07         bestValue := max( bestValue, v )
08     return bestValue

it is non-functional in every way and I’m having a hard time translating it

do you care that it’s implicit-stack recursion?

(apply max &c)

no big deal to translate one to the other I guess

what is &c in this context?

that lack of early returns is what’s making it hard for me

lvh something like this?

(defn -negamax [game depth color]
  (if (or (g/game-over? game) (= depth 0))
    (* color (evaluate game))
    (apply max (map #(- (-negamax % (dec depth) (- color))) (advances game)))))

zero?

but yeah

pretty much

I mean, you need a base case for game advances being empty

holy shit it works perfectly

thanks 😄

my tests just magically passed

so, apply max Float/NEGATIVE_INFINITY

probably

min and max don’t have nullary arities

gotcha

this is one of the few times where I’ve seen an imperative approach be much cleaner and more readable then the functional one

eh, it’s a translation of one to the other

I’m not sure there’s no more elegant functional description 🙂

yeah, wait, I’m just an idiot. I used to have a bunch of atoms in there but I don’t need those anymore

plus, regardless, I would write that second branch with ->>

so this is functional

yeah, that probably a good idea

okay, the next step for this seems a lot trickier… https://en.wikipedia.org/wiki/Negamax#Negamax_with_alpha_beta_pruning

heh, someone already did it, and got stuck in the same place as me: http://chongkim.org/programming/2013/05/29/an-elegant-negamax-with-alpha-beta-pruning-in-clojure.html

>>> Note The flaw in this code is that it is not updating the alpha after checking each child. It does get set when the cousins are calculated. This means I will need to write a new function that combines the map and take-while.

what a cliffhanger

@schmee something that might help is reduced form with reduce

specifically for the if val >= beta return val part

that's a good one!

notice that there would have been an exception thrown if it got to nil - it effectively short circuits

I can’t help but feel like alpha and beta have to be atoms

cause I don’t see how the updated values can reach the recursive calls otherwise

schmee the alteration of alpha directly translates to part of the reduce (based on the code snippet on the chong kim site)

instead of foreach you have reduce, instead of changing alpha you return an accumulator with the new alpha in it, instead of the early return of val you use (reduced val)

it all translates directly without mutation

oh, and val would also be a value in the accumulator

:thinking_face:

alright, I’m going to play around with that tomorrow, it’s way past bedtime already 😄

thanks for the suggestions 👍

is clojureql dead? all the doc sites are down and the last commits were years ago. asking because it seems to be the way to get data from a sql db into incanter.

@schmee essentially (reduce (fn [[alpha val] child] (cond (> val beta) (reduced val) (> val alpha) [val val] :else [alpha val]))) init node)

oh wait there’s also the part where val gets calculated - exercise for the reader, heh

haha, I’ll try!

but that should be pretty close to a working pure version of the foreach

you seem to have this one in the bag, I might just ping you if a need the answers sheet for this one… 🙏

Every time I see Dakrone I think "Dankthrone" which sounds like a rad Stoner Metal band

1. I am writing code in cljs. 2. This code requires a macro. 3. I wrote a macro in a clj file. 4. This macro does something stupid and throws an exception at compile time. 5. The error I get back from the cljs compiler does not seem very helpful. 6. Is anyone else running into this problem?

@qqq wrap the macro in a try catch and .println to System/out

@adambros: I'm not sure I understand. Are you suggesting:

(defmacro ...
  (try ....
    (catch / println )))

Sorry I'm on my phone, but basically yes

That way you control what you see, it's usually helped me, as the unhelpful errors are a npe and it just give me one line

It's (catch Exception e (.println System/out ...))

Is there a function like this in core, specter, or somewhere else?

(defn apply-if [f p x] (if (p x) (f x) x))