hi everyone. is there a shorter version to this?

(let [[butlast-args last-arg] [(butlast args) (last args)]])

(let [[x y] [(butlast a) (last a]])

that’s what i have 😄

yeah, joking 😅

i thought there’d be a function specifically for doing that coz i’ve seen clojure do magic 😄

you could write last-two

ah nevermind, was confused about what butlast did

funny thing was i just typed butlast assuming there’d be a function like that

and there was..

😄

what data structure is args?

you could play with split-at (dec (count coll)) but i'm not sure how you can make this more terse

((juxt pop peek) (vector args))

user=> ((juxt pop peek) [1 2 3 4 5])
[[1 2 3 4] 5]

not fair, no local bindings but nice 😉

what I'm saying is that (juxt pop peek) is more concise than split-at etc.

user=> ((fn [& args] (let [[b l] ((juxt pop peek) (vec args))] {:b b :l l})) :a :b :c :d :e)
{:b [:a :b :c :d], :l :e}

(and fairly efficient too, compared to last/`butlast`)

anyone using hugsql? Is there a way to import a file of definitions in another file? I have a bunch of snippets that I want to be used in each hugsql file I make.

If I have a ns that looks like this:

(ns my-ns.core
  (:gen-class))

(defn -main
  [& args]
  (println "------")
  (prn "ARGS:" args)
  (println "------"))

(binding [*compile-path* "target/classes"] 
  (compile 'my-ns.core))
=> my-ns.core

java -cp 'target/classes/*' my-ns.core

@kenny java -cp target/classes my_ns.core would be what I expect.

Ah, the * was causing the issue. Now getting Exception in thread "main" java.lang.NoClassDefFoundError: clojure/lang/Var. Do I also need to include Clojure on the cp?

depends how much you AOT, I've noticed that when I AOT I am only getting the immediate dependencies of my main. I'm not sure if this is intentional behaviour though.

you won’t transitively compile clojure itself, as it’s already compiled

so clojure needs to be on the classpath

@bronsa I noticed that only things I did :require on from my main were AOT'd, is that correct?

yes

Is there some sort of cache for compile? I have noticed that sometimes things in my :require are included in *compile-path* and sometimes they are not.

@bronsa is this what leiningen works really hard to workaround. I recall some workaround it performs.

@kennny no compilation cache but clojure won’t re-compile a clj file if the class file is more recent

Which class file is it using to determine that?

@dominicm yes see https://dev.clojure.org/display/design/Transitive+AOT+Compilation (and https://dev.clojure.org/jira/browse/CLJ-322)

@kenny the one you’re compiling

if you’re compiling my-ns.core that would be my_ns/core__init.class

So if I delete the contents of target/classes it would recompile everything?

just the transitive deps of my-ns.core

provided they aren’t loaded already

Still doesn't include core async in target/classes :thinking_face:

if they are already loaded they won’t be recompiled

@kenny it might be that you have already required core.async before compileing my-ns.core

Ah, yes. I definitely did that.

right

that’s the issue

That makes sense 🙂

you can replace your :requires with e.g. :require core.async :reload-all to force reloading

or you can use tools.namespace to force unloading of the transitive dependencies for my-ns.core

Gotcha. Adding clojure to the cp doesn't seem to fix the issue.

java -cp target/classes:~/.m2/repository/org/clojure/clojure/1.9.0/clojure-1.9.0.jar my_ns.core
Exception in thread "main" java.lang.NoClassDefFoundError: clojure/lang/Var
        at my_ns.core.<clinit>(Unknown Source)
Caused by: java.lang.ClassNotFoundException: clojure.lang.Var
        at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
        at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:331)
        at java.lang.ClassLoader.loadClass(ClassLoader.java:357)
        ... 1 more

I assume order shouldn't matter for the cp args

@kenny ~ won't expand when squished like that.

use $HOME instead

Or ${HOME} if you want to be super safe.

I assume I need to do the same for Spec with Clojure 1.9?

yep

any reason you’re not using clj+`deps.edn` btw?

@metametadata your byplay library looks really interesting! Somewhat related to background job processing, I'm evaluating different scheduling libraries (scheduling as in "run this task 5 days from now"). Just wondering if you have any experience or preference with libraries like quartzite or monotony?

Just trying to get a deeper understanding of how this works. Also, I wrote a clj uberjar task that does not work when dependencies on included and I want to figure out why.

:thumbsup:

Any ideas here?

java -cp target/classes:$HOME/.m2/repository/org/clojure/clojure/1.9.0/clojure-1.9.0.jar:$HOME/.m2/repository/org/clojure/spec.alpha/0.1.143/spec.alpha-0.1.143.jar uberjar_sample.core

Error was too long for slack ^

Basically Caused by: java.lang.ClassNotFoundException: org.objectweb.asm.Type. Haven't seen that class before.

that’s a missing dependency from tools.analyzer

I assume core.async is using that. Is there a reason that wasn't included when I AOT'ed my ns?

it’s a java library :)

Ah. So that needs to be manually added to the CP as well?

yes

@kenny if you want a convenient way to "dump out" the libs juxt/pack.alpha just added support for doing that in a particular branch. It's also good for uberjars, if you're looking for a library for that.

I thought clojure had org.objectweb.asm.Type

oh it does but it renames it i think

@dominicm Might be helpful to take a look. Which branch?

@dpsutton tools.analyzer.jvm uses org.objectweb.asm, not clojure.asm

@kenny it's called "skinnyjar", latest commit is 800afc066907e31fe90ce4df4f519e6224116867

clojure bundles org.objectweb.asm but it renames it

the reason why t.a.jvm doesn’t use the bundled one is that up until 2 days ago clojure shipped an ancient version of ASM and we were using modern APIs that just didn’t exist in the version bundled with clojure

ah. that's why ghadi bumped it.

well

the ASM version got bumped because the class version and minimum jvm requirement got bumped from 1.6 to 1.8 which means we can now start using modern bytecodes (indy) which the old version of ASM just didn’t support

+ there was a bug in the emission of certain static method invocations due to the old ASM not fully supporting 1.9 bytecode verification changes

https://github.com/clojure/clojure/commit/3d6c1364698a4b7020a6d97f33b9a596a1160d4b

@kenny Why are you trying to manually compile Clojure code and run it with java? That seems like a very painful way to work.

@seancorfield Oh it is 🙂 I wanted to better understand what was going on behind the scenes.

@kenny I think one way if you want to learn what is going on is to just read the source of ‘lein run’ or ‘lein test’ task. Instead of going through this painful procedure.

That exercise provided me with exactly the information I needed 🙂

@kenny Compiling Clojure is painful in all sorts of ways. We don't AOT anything at World Singles now. We still build (uber) JAR files for deployment tho' (from a mono-repo).

@seancorfield Cool. And you run the uberjar with clojure.main instead?

Yes. java -jar path/to/my.jar -m my.app.entry

(we still highlight clojure.main as the entry point)

It also has the benefit that if you want to explore the app interactively for any reason, you can run it without -m and get a REPL that has all the dependencies and source directly available.

I have a unit test with clojure.test that starts with instrument and ends with unstrument to mock out some function calls. However, it seems to sometimes call the underlying functions I’m mocking and accessing the database, which I’m trying to avoid. Any idea how to make sure there’s no race condition happening here?

combing instrument and with-redefs sounds like a bad idea

that being said, I doubt there is a race condition there, my guess is the race condition is you are spinning up threads to do something in your test, and don't ensure dynamic extent of the thread is the same as the dynamic extent of the test

Hmm, ok.

I’m still figuring out the best way to get spec.test into our pipeline.

I suspect you have a bug there

in that you well set state to the string "third"

(swap! state update-in [:task-list] conj "third") will likely do what you meant

No, I'm adding the string "third" to the list ["first" "second"]. Let me try your approach.

oh, right, I didn't read the parens right

(swap! state update-in [:task-list] conj "third") this only replaces the vector

common to see (fnil conj []) in the case the key isn't already present

also no need for an update in if the path is one segment. update :task-list

Clojure 1.9.0
user=> (def state (atom {:task-list ["first" "second"]}))
#'user/state
user=> (swap! state update-in [:task-list] conj "third")
{:task-list ["first" "second" "third"]}
user=> @state
{:task-list ["first" "second" "third"]}
user=> 

not sure what you mean by "only replaces the vector"

My bad, sorry about that 😉 This works indeed. Thanks a lot!! Clojure is sometimes a bit overwhelming to find elegant solutions

update is a newer addition, that in my opinion was silly

In what way?

I was on the fence but I use it all the time

like, we had update-in, it works fine, but someone decided to create a jira issue because they wanted a version that only worked for a single path segment, and then spend however many months to shepherd that change through and get it merged, and then it caused issues with different clojure libraries because it introduced a commonly used name in to clojure.core

what a waste of time

Ah well, but that's more of a political/project structural issue.

so now, everytime someone writes code that uses update-in and works just fine with a single segment, the first comment is someone going "but actually you should use update"

what an even bigger waste of time

sulks shamefully

To me Clojure feels somewhat "hacky", like 'emacs lisp', coming from OO languages and having some little experience with Haskell and Scala. But I guess this is a more of an evolutionary approach.

I think it would be hard to reconcile the idea it is hacky with rhickey, but it is definitely coming from a very different place

if you haven't see "are we there yet?" from almost 10 years ago, it is still very good, and I think puts on display a lot of the thought that went in to clojure

Yeah, I've watched many talks he has given and my impression is that he is less likely inclined to the way academic groups would design a language. But I like that a lot, as for example Scala seems to artificial from time to time. And not to mention Java itself, which is very robust, but lacks innovation. Once again thanks for your help @hiredman!

Im looking for a way to get the full path to leaf nodes of a map of maps, so for example if I have {:a 1 :b {:c {:d 2}}} it would return [[:a] [:b :c :d]]

e.g. the result would be vectors suitable to pass to get-in to retrieve any individual value from the original map

suggestions on approach appreciated, don't need full solution

people ask this a lot, I am not entirely sure why

user=> 
((fn map-paths [path m]
   (if (map? m)
     (for [[k v] m
           i (map-paths (conj path k) v)]
       i)
     [[path m]]))
 []
 {:a {:b 1
      :c [3]}
  :w {:x 1}})
([[:a :b] 1] [[:a :c] [3]] [[:w :x] 1])
user=> 

thanks, more than I expected to receive

and to answer your question, in this case I want to convert an object structure to a java properties list