Damn I've deleted the message somehow lol Do we h...

# javascript

[Edoardo Luppi]

Damn I've deleted the message somehow lol

Do we have an identityHashCode() mechanism in K/JS?

[Artem Kobzar]

No, because there is no API to get GC ID for an object.
We can emulate it for sure, like this:

var index = 0
val jsWeakHashMap = js("new WeakMap()")

fun identityHashCode(object: Any): Int =
  if (jsWeakHashMap.has(object))
     jsWeakHashMap.get(object)
  else 
     (index + 1).also { jsWeakHashMap.set(object, it) }

[Edoardo Luppi]

I need to check how they've done it in Scala/JS
https://github.com/scala-js/scala-js/issues/1027

[Edoardo Luppi]

Looks like they were using a counter, pretty much as you did there.
I'll look on the up-to-date code

[Edoardo Luppi]

This is how it was done before they moved it to a code generation step
https://github.com/scala-js/scala-js/commit/33792738f548047ac0057c28c0945b52dbb5550b#diff-c259eba87b2eb67bfc6b869e4[…]14eb57877f1c0dc6b5947c3d69L186
Look for

// identityHashCode ---------------------------------------------------------

This is how they do it now with codegen
https://github.com/scala-js/scala-js/blob/9f5cc98f7f8f1d93a07ca6592aea2e953488b062[…]c/main/scala/org/scalajs/linker/backend/emitter/CoreJSLib.scala

[Artem Kobzar]

Mostly the same. They also use JS WeakMap

[Edoardo Luppi]

Yeah. When WeakMap is not available they use a magic field

val hash = x.asInstanceOf[js.Dynamic].selectDynamic("$idHashCode$0")

Which, as far is can tell, is simply set to nextIDHashCode()

[Edoardo Luppi]

And then there is also a fixed 42 when the class is sealed. I don't get that part.

[Artem Kobzar]

Yeap. So, we can do the same. Actually, for the JS objects hashCode we do the same thing.
https://github.com/JetBrains/kotlin/blob/8d1a90c23cbd4e04747e1e9815047bb54a987e83/libraries/stdlib/js/runtime/coreRuntime.kt#L44

[Edoardo Luppi]

So, in K/JS the hash is computed with calculateRandomHash
In Scala/JS it's only a counter++

The difference between K/JS and Scala/JS is only the WeakMap usage if it's available?

[Edoardo Luppi]

And the most significant difference is Scala/JS has two separate hash codes.
The "id" one, and the one you can specify on your class/object.

[Edoardo Luppi]

Btw, I was looking at the identity hash code because I need an identity map in ANTLR

[Artem Kobzar]

I proposed to add it into our stdlib based on JS Map.
But the problem, that IdentityHashMap breaks some heuristics of all regular Map (especially the one about using of hashCode function)
So, you can emulate it by yourself using just the regular JS Map

[Edoardo Luppi]

So, you can emulate it by yourself using just the regular JS Map

That was my initial idea, but what about the environments where a JS Map isn't available?

[Artem Kobzar]

https://github.com/medikoo/es6-map/blob/master/polyfill.js

[Edoardo Luppi]

Thanks! I wanted to avoid polyfills, but it looks easier for now instead of emulating it on my own.

At some point it would be nice to have that IdentityHashMap in the KMP stdlib tho.

[jw]

Compose also has one: https://github.com/JetBrains/compose-multiplatform-core/blob/211c3844780332d2a4253206e8482120ded57c26/compose/runtime/runtime/src/jsMain/kotlin/androidx/compose/runtime/ActualJs.js.kt#L32

[Edoardo Luppi]

Oh cool! So it looks like the use cases are there.
Now I learn that Native has a kotlin.native.identityHashCode function in stdlib.
A common function in stdlib for all targets seems doable 👀

[Edoardo Luppi]

Korge also has its own implementation
https://github.com/search?q=repo%3Akorlibs%2Fkorge5-prototype%20anyIdentityHashCode&type=code