How are SPL program discriminators determined

Question

I assume that there is a naming scheme being used for the account state discriminators as seen here in the token metadata extension

    const SPL_DISCRIMINATOR: ArrayDiscriminator =
        ArrayDiscriminator::new([112, 132, 90, 90, 11, 88, 157, 87]); 

My questions are:

1. Whether the slice is chosen randomly or if there's meaning to each value in said slice.

2. Why an array discriminator as opposed to a hashed constant string value?

Answer 1

The discriminator actually is calculated with a hash! You can see the test for it at https://github.com/solana-labs/solana-program-library/blob/57bc2457827775b3299fa9efc02ce7e50d34135f/token-metadata/interface/src/state.rs#L131-L136

The spl-discriminator library was very new at the time, and didn't have the ability to use a macro to create a discriminator, but now it's possible to do something like:

    #[derive(SplDiscriminate)]
    #[discriminator_hash_input("global:my_instruction_with_one_generic_and_lifetime")]
    pub struct MyInstruction5<'b, T> {
        data: &'b [T],
    }