Clojurians Log v2

@bronsa: so I've been digging on that fn .withMeta thing you pointed out yesterday, and I'm pretty much convinced that the AFn.withMeta is dead code.

ObjExpr.compile already creates .withMeta() and .meta() implementations unconditionally both for deftypes and fns. Demo of this behavior above. If the default abstract class implementation were running, then that vary-meta invocation would have returned a different class type.

don't think that's true

IIRC it does that only for reify

ompiled from "test.clj"
public final class test$fn__18 extends clojure.lang.AFunction {
  public test$fn__18();
    Code:
       0: aload_0
       1: invokespecial #9                  // Method clojure/lang/AFunction."<init>":()V
       4: return

  public static java.lang.Object invokeStatic();
    Code:
       0: aconst_null
       1: areturn

  public java.lang.Object invoke();
    Code:
       0: invokestatic  #14                 // Method invokeStatic:()Ljava/lang/Object;
       3: areturn

  public static {};
    Code:
       0: return
}

well, is that jaunt or clojure? simple_smile

I have no idea what changes you made

but in clojure that's definitely not the case

That's Jaunt. I haven't deliberately changed anything there yet...

sigh

dives into Compiler.java

may the Old Ones smile on you

oh my god

this is disgusting

probably

withMeta is only emitted if the function already has metadata at compile-time

e.g. ^:foo (fn []) will have it, (fn []) won't

Yes.

reason why you're seeing it everywhere in Jaunt is because you're attaching :uses etc

Right. Because that's how it makes the MetaExpr(FnExpr, MapExpr) work.

I agree that should be unconditional.

I guess just make supportsMeta return true in FnExpr rather than checking for hasMeta then

sgtm.

The other thing I want to do here is ditch that MetaExpr, metadata should be in a Fn's <cinit>

Eh.... that's actually not critical. Whatever.

meh.

@arrdem: or maybe even make supportsMeta test hasMeta || hasPrimSigs

that's the only case where that makes a difference

or not, if you want to avoid forcing the applyTo path I guess

Eh. I'm content to hard code supportsMeta true, right and avoid the applyTo path.

Two line change and a bunch of deletions. Best kinda patch.

downside is every fn is now ~300bytes heavier

Not even...

Derp. I didn't read carefully. You were dead on.

Eh. The price of sanity. I'm willing to pay it and after all this is my sandbox 😛