UUIDs for Types

[queen shoal]

I am using allot of virtual functions in a gui system and i want to be able to check "types",
For example:

// @TypeOf(node) == *Node
if (node.type_id == @idOfType(ClockNode)) {
  clock: *ClockNode = @ptrCast(node);
  return clock.getTime();
}```
Does zig store anything like this? using strings wouldnt be ideal and the other option is enums which is more code and less scaleable

[viscid trellis]

you can get a type id esc system with this:

pub fn typeId(comptime T: type) *const anyopaque {
    const storage = struct {
        var v: u8 = 0;

        comptime {
            _ = T;
        }
    };

    return &storage.v;
}

(EDITED)

[queen shoal]

wow thats pretty smart, ill use this if theres no other way. I would think the compiler stores IDs already but its not accessible

[viscid trellis]

i think theres an issue for this, someone got the same result by doing something funky with anyerror but i dont remember what lol

[queen shoal]

actually a ct hashing function would also work,

but with the above at least theres no collisions

[bleak monolith]

You’re probably interested in this proposal: https://github.com/ziglang/zig/issues/19858

[bleak monolith]

Yeah that experiment is here: https://discord.com/channels/605571803288698900/1272273545451274362

[viscid trellis]

ah here it is lol https://github.com/ziglang/zig/issues/19858#issuecomment-2094339450

[queen shoal]

the coolest version would be if it checked for duplicates, if we had const X = f32 and const Y = f32 they would have the same IDs, and same with structs that have identical layouts

[viscid trellis]

oh he made it into a library lol

[viscid trellis]

this does that for X == Y afaik :P #1272476391803191358 message

[queen shoal]

wouldnt for this tho

[viscid trellis]

it do

[queen shoal]

oh.

i guess thats an obvious optimisation if X == Y is allready readilly checkable

[viscid trellis]

it works at comptime/runtime, so its not really that much different from a builtin

only downside is no @intFromPtr at comptime of course

[queen shoal]

this post doesnt require it

[viscid trellis]

polluting anyerror (more than it already is) just feels wrong to me tho lol

[queen shoal]

all of this feels wrong.

[queen shoal]

this pollutes aswell

[viscid trellis]

how so?

[queen shoal]

its creating a struct, eitherway the compiler keeps track of it

or does it get disposed because its not returned?

[viscid trellis]

the only thing that gets added to the binary is the 0

[queen shoal]

i was wondering about the struct being created at comptime

[viscid trellis]

ok nvm theres a bug in this on release safe for some reason :P

[viscid trellis]

types dont exist in the binary

[queen shoal]

what bug?

reusing the same pointer?

[viscid trellis]

yea

[glacial sleet]

has to be var v: u8 = 0; to prevent merging

[viscid trellis]

ah thanks

[queen shoal]

the compiler stores them

[viscid trellis]

you mean like it could hurt compile times?

[queen shoal]

probably not lol, just none of this would be needed if the proposal gets approved