Allow array indexing

[schoen]

Project

core

Describe the feature

Maybe this is tricky in terms of the type system, but I think it would be great to able to do

let a: [u8; 4] = [0, 17, 255, 23];
let b: u8 = a[3];

and similar. Right now you have to actually destructure the whole array in order to get individual values. For example, p2ms.simf does

    let [pk1, pk2, pk3]: [Pubkey; 3] = pks;
    let [sig1, sig2, sig3]: [Option<Signature>; 3] = maybe_sigs;

and you can't conveniently refer to the value at position i at all. I don't think there is an architectural reason that this is meant to be prohibited.

(Among other things, this means that it's hard-to-impossible to loop over an array with for_while, although you can differently loop over an array with array_fold.)

[apoelstra]

It's easy to implement this for fixed constant indices. To implement this for arbitrary indices will likely involve code that resembles a binary search into the array based on the bits of the index.

We definitely don't want users to be looping over arrays by index like this. array_fold is far more efficient. But sometimes it's necessary, e.g. when iterating over multiple arrays in parallel (sometimes with offsets etc). We can try to add first-class language support for various iteration parameters or maybe we should just let users use indices and then attempt to fix it during an optimization pass.

[stringhandler]

This is possibly a candidate for a jet, but I'd suggest get_arr(array) -> Option<type>. We would need jets for every data type though.

[apoelstra]

I think that as a jet it wouldn't work because there are an unbounded number of array types. It could be a builtin in HL though.