@dominicm maybe take a look at onyx or tesser libs?

Is there a performant way to try and wrangle a clojure object to a java class

my current approach is using Jackson and cheshire and serializing through json

But I cant help but feel like theirs a better option than basically going through a java version of JSON.parse(JSON.stringify(..))

maybe it's better for cli to provide a error message rather than throw a NullPointerException when there's no -main?

@emccue everything in jvm clojure is an instance of some java class - what class do you need to create?

POJOs from other parts of the codebase

I would just write coercion functions. If you really want to get fancy, write a macro that loads the class, uses reflection, and writes your coercion functions for you.

well I want that fancy version but without me doing any work

The coercion functions wont work for the particular use case

sadly clojure only makes it easy to create objects implementing a defined interface

you coul use defprotocol to define the methods you want to implement then reify that protocol, but that's a little odd

welp, jsonifying it is

smells like transducers to me

does array-map keep the insertion order of keys?

yes

does anyone know of a good collection of Clojure benchmark code?

I want to experiment a bit with the compiler and I need a way to measure my results

Andy Fingerhut has a repo with some fine grained benchmarks. Maybe too fine.

http://jafingerhut.github.io/clojure-benchmarks-results/Clojure-benchmarks.html

perfect, thanks alex! 😄

is anyone aware of any writing about passing maps instead of single arguments in Clojure? I want to know when it is and isn’t a good idea.

I added a print statement to https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/Compiler.java#L3706 (InvokeExpr#emit) just to get an understanding of what’s going on

now, at the repl, I see this:

user=> (defn foo [a b] (+ a b 3))
#'user/foo
user=> (foo 1 2)
fn -> EMIT name:user$eval8278 internal name:user$eval8278 thisName: null
6

so code is getting emitted. my question is: where the heck does this bytecode end up?

all bytecode has to be in some sort of class, right?

looking at how the repl is defined in main.clj, I see this at the beginning of defn repl:

(let [cl (.getContextClassLoader (Thread/currentThread))]
    (.setContextClassLoader (Thread/currentThread) (clojure.lang.DynamicClassLoader. cl)))

the bytecode is loaded straight into memory by the classloader

https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/Compiler.java#L4972-L4982

this is where it happens

https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/DynamicClassLoader.java#L44-L49

Who loves fragile unit tests? raises hand

Is this a good way to implement prefix tries (for autocomplete for example) in clojure - https://gist.github.com/joelkuiper/a422cea41a26c9a3e2b35ac3abbb06f8

@bronsa ahh, thanks! Is this bytecode the same as a class file, more or less?

exactly the same

except it just gets loaded into memory instead of being written to a file and loaded from it

So when the class loader loads the class, it has to get "evaled" somehow? How do you know which method you're supposed to call on the resulting class?

the class is an IFn, the compiler invokes its invoke method

Aaaaahhhh

Now it's clear, thanks a ton for putting up with my endless compiler questions :)

I guess the class loading machinery has it's work cut out for it when dealing with Clojure... And I don't even want to think about the security implications of class loading 😂

So basically, Java has runtime eval just like clojure, did not know 🙂

@schmee The JVM does, yeah. The JVM is much more dynamic than the Java source language might lead you to believe.

classloading is how stuff like RMI is possible

Java gets a lot of hate, but the more I dig into Java, the JVM, and Clojure the more impressed I get

interjecting a little

So much exciting stuff on the horizon: value types, vector api, pattern matching, graal... Good times for JVM langs!

how do I give a type hint to clojure that a method on genclass returns a List of another class

:methods [[get [] java.util.List]]

you don't

its for gen-class

so java can see it

generics are just a synctactic feature of java

thats a bummer because I want to play nice with the java side of things

There has to be SOME information i could give so my gen'ed classes dont just get seen as returning ungenerified lists

no

it will return List<Object>

Which isnt really that helpful

¯\(ツ)/¯

Not making it easy to weasel clojure into my codebase

Im also getting package ... does not exist error when compiling through maven

http://alexott.net/en/clojure/ClojureMaven.html

using this library and the example config

so uhh, if anyone hasnt been able to switch to leiningen

help appreciated

Is there anything tying ClassLoader hierarchies and Clojure? I'm not interested in isolating clojure versions particularly, nor in total isolation, I want an actual hierarchy.

I am pretty sure it is all or nothing

the jvm really doesn't want two classes with the same name visible from the same place in the cassloader hierarchy, so to do things like pass clojure data around between runtimes you would still need to serialize and de-serialize which is as good a definition for isolation as there is

I put that name thing badly

what I mean is, runtime 1 won't be able to treat a PHM from runtime 2 as a PHM, but, depending on what you are doing it would likely recognize the PHM as a java.util.Map impl

@hiredman if they shared a clojure, that would be fine though right?

I'm not certain if that's even possible, but just as a possibility.

Then what is different between whatever you are proposing and just normally running Clojure?

@emccue Not sure if it'll help your quest but all the Contrib libraries are built and tested with Maven at the moment so maybe look at the pom.xml files there?

If they share a Clojure runtime, they are the same clojure and there is no difference hierarchy or isolated or whatever

@hiredman I want to isolate some dependencies of a dependency.

You cannot

Clojure names are globally visible within the Clojure runtime

Namespaces

And clojure code, unless explicitly written in a way to do otherwise, will all lookup classes the same way

@seancorfield Thats helpful and ill look through some more of them now

Ive gotten it to the point where calling mvn clojure:compile works fine, but It doesnt seem like its hooking in to the regular compile cycle

The very little I know about Maven, I've learned from dealing with pom.xml files in the various Contrib libs that I maintain ¯\(ツ)/¯

@emccue Something else to look at would be the pom.xml generated by clj for a deps.edn project, but I'm not sure whether it generates a complete compile/test XML file for that...

Is there a way to unquote-splice a temp-var in a macro?

`(let [x# [:a :b :c]] ~@x#)  

I feel like people who know less about maven live happier lives

@hiredman I see. That makes sense. A bit frustrating for me. My use case probably could work with a fully nested clojure, although the performance penalty might be a frustration. It's a shame I can't use shim dandy, as it's so interested in isolation.

@hiredman it might just be easier to use Mr. Anderson tbh

how do I un-namespace a var in a macro

right now im getting com.NDA.factories/ProductGenerators

I want to get out just ProductGenerators