RFC: add discriminator to idl

[seanyoung]

This adds the following syntax:

#[discriminator(0xffefabcd)]
pub fn initialize(ctx: Context<Initialize>, data: u64) -> Result<()> {
      Ok(())
}

This PR is not complete yet - the code generation does not use the discriminator yet, and the client should also use the new value.

[vercel]

@seanyoung is attempting to deploy a commit to the coral-xyz Team on Vercel.

A member of the Team first needs to authorize it.

[armaniferrante]

Can you elaborate on the use case for this macro? When would an anchor program use it?

[seanyoung]

So the plan was that #[discriminator(x)] would you allow you to override the discriminator, so that we can test an idl with different discriminator than the standard discriminator created with hashing.

As for a use case,I did imagine a usecase where you can specify a single byte as a discriminator, but this syntax does not allow this yet. This would allow you to just use a single byte rather than 8 bytes for discriminator.

What do you think of:

#[discriminator([42])]
fn foo(..)

or for two byte discriminators:

#[discriminator([42,12])]
fn foo(..)

[seanyoung]

If you don't think the attribute is useful, I can create a test using a non-anchor rust contract and a hand-written idl, which is then used from rust and ts.

[stegaBOB]

I think this could be interesting. How would you handle potential collisions and/or ambiguity with discriminators of different lengths?

[jacobcreech]

Closing as stale + implemented in 0.31