M1 java or Rosetta?

You don't need the static (does nothing now) or the long coercions (you do need the type hints)

M1 java

woah… 202ms when I type “clj” at the repl. so something weird is happening with my emacs repl

clj-test.core=> (time (fib 38))
"Elapsed time: 331.063833 msecs"
63245986

is your emacs compiled for Apple silicon?

hmm

oh wait 🙂

I get it… I HAPPENED to be in a namespace where I had overwritten the arithmetic operators with the generic multi-method based ones

duh

Also, this should actually be faster in the new 1.11.0-alpha4 due to the new use of Math.exact w hotspot intrinsics for the + - dec

This Math/exact really looks promising:

(defn fib ^long
  [^long n]
  (if (>= 1 n)
    1
    (+ (fib (dec n)) (fib (- n 2)))))

(quick-bench (fib 38))

;; Evaluation count : 6 in 6 samples of 1 calls.
;;              Execution time mean : 304,612157 ms
;; ...

(defn fib ^long
  [^long n]
  (if (>= 1 n)
    1
    (Math/addExact
      (fib (Math/subtractExact n 1))
      (fib (Math/subtractExact n 2)))))

(quick-bench (fib 38))

;; Evaluation count : 6 in 6 samples of 1 calls.
;;              Execution time mean : 236,714805 ms
;; ...

(defn fib ^long
  [^long n]
  (if (>= 1 n)
    1
    (unchecked-add
      (fib (unchecked-dec n))
      (fib (unchecked-subtract n 2)))))

(quick-bench (fib 38))

;; Evaluation count : 6 in 6 samples of 1 calls.
;;              Execution time mean : 201,021006 ms
;; ...

Clojurescript: https://clojurescript.org/ 😄 #clojurescript

I think you need to qualify what you mean by “use javascript with Clojure”.

ClojureScript compiles to JavaScript, Clojure compiles to Java bytecode. It is essentially the same language.

I have javascript. So I might have two options. Convert javascript to java, or rewrite the javascript in clojure. I’ve never written Java, only javascript. Either direction I go, it will probably not take me much time because my project is still very small. it will probably just make more sense to write it in clojure since it is so small right now.

You can run JavaScript code from ClojureScript. Why is ClojureScript not an option? Do you need to use a Java library too?

im not sure if i want to use clojure or clojurescript… im writing for a backend so i figured i should use clojure.

Another option might be using a JS interpreter from Clojure such as Nashorn, but perhaps rewriting in Clojure is the best option like you wrote.

Also possible: using Clojurescript on the server by generating to a node.js target and running it via node. I haven't seen a lot of this but if your infrastructure/server skills are already node-flavoured this could be a somewhat gentler on-ramp.

Not a lot of blog post guidance on this direction relative to the advice you can find on JVM server work, but it's do-able.

Because I’ve been using it lately, I’ll comment on the JS engine… Nashorn is deprecated now. If you’re on JVM 15 or later, then you need to use the GraalVM https://www.graalvm.org/reference-manual/polyglot-programming/#start-language-java.

There is a blog post from Nextjournal on how to use JavaScript inside Clojure using Graal’s Polyglot (as pointed out by quoll). If you’re just starting to learn Clojure I wouldn’t suggest going that route just yet though

No. But you can have multiple calls to sh.

That is what I figured. Thanks!

will do, didn't realize that was copied over there

Same, so I plotted them out

The red ellipses indicate the type tests that the functions look for

my strong suspicion is no, as record fields map 1->1 with static data accessible in instances of a record class, and class fields don't have a concept of namespace

If you have a hash map {:foo/bar 13} and you switch to a record, that's a breaking change even without Spec in the mix, isn't it? Since the record would only have bar as a field...

I think the question is just about that breaking change to records and spec is mentioned as a motivating factor

and it would have to be a breaking change, since it's impossible to make a munge that can't clash (though you could make one that' highly unlikely to clash at the cost of severe ugliness)

In the spec, you'd change :req to :req-un and :opt to :opt-un but keep the namespaced keys for the spec names (since the -un versions allow namespaced keys for locating the spec but only use the unqualified key for the field name). So that's less of a change than all your code that pulls fields out of a map... So I guess I don't see spec as being the motivator here really?

I take people for their word regarding motivation, but I do agree that it's easier to use spec differently than it is to make namespaces work with record keys

I skipped some steps in my reasoning above re: munging. I take it as a premise that records need that 1->1 mapping of keyword to static field on the resulting object, and a field can't have a namespace, so the only other option I can see there (without some sort of complex indirection) would be munging the namespace into the field name.

Yeah, I think your reasoning is correct.

Given the way keyword invoking works on records, if you were sufficiently motivated I believe you could make your own thing that used the same inline caches with namespaced keywords

I guess I don't recall of the compiler skips the cache stuff for namespaced keywords

yeah

(deftype X [thing]
  clojure.lang.IKeywordLookup
  (getLookupThunk [_ k]
    (case k
      ::foo (reify
              clojure.lang.ILookupThunk
              (get [_ obj] (.-thing ^X obj)))
      nil)))

((fn [] (::foo (->X 1))))

@U04V70XH6 no the motivation is the change from :req to :req-un because all consuming code now has to switch to using non-namespaced keywords, which could potentially be a lot of code, and could potentially be library consumers.

And yes, I was personally considering munging of the field names as the solution to this

Thanks for the suggestion there @U0NCTKEV8