Added A par_chunk_by method

[jakeKonrad]

Hello, I found this method useful in my code and thought it could be used here. However it relies on #![feature(slice_group_by)] unstable feature of the rust std library and you would probably want to wait for that method to be stabilized anyways so as to stay consistent with whatever naming they choose. This pull request is here for when that happens. Also if any one has any comments on the code I'd really appreciate hearing them.

[cuviper]

Thanks, it's an interesting feature! Do you think you'll try a mutable version as well?

And yes, we'll want to wait for the API to stabilize in the standard library so we can match it.

[jakeKonrad]

That should be the mutable version.

[cuviper]

Just for reference, that slice_group_by feature is tracked in rust-lang/rust#80552.

[stevenliebregt]

I'm interested in this, but I found a bug in it, the following call:

let data = vec![1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 4, 4];
let groups: Vec<_> = data.par_group_by(|&a, &b| a == b).collect::<Vec<_>>();

println!("{:?}", groups);

will result in 5 groups instead of the expected 4, these groups are:

[[1, 1], [2, 2, 2, 2, 2], [3], [3, 3], [4, 4]]

[jakeKonrad]

I think that fixed it

[stevenliebregt]

I think that fixed it

It did, nice job!

[cuviper]

FYI, that stabilization just landed with a rename to chunk_by / chunk_by_mut:
rust-lang/rust#117678

[cuviper]

I wonder if this can be refactored to use the internal SplitProducer infrastructure, because that's basically what this iterator is doing overall, splitting on a pair-predicate. That API won't work directly as-is, but I've also been working on modifications to add split_inclusive, which might be instructive on how to extend it.

[cuviper]

I wonder if this can be refactored to use the internal SplitProducer infrastructure,

I didn't find a good way to shoehorn this into SplitProducer, per se, but I've now made the implementation quite similar to that, and a test asserts that we get exactly len() - 1 comparisons.

[cuviper]

Thanks @jakeKonrad! Hope you don't mind that I wrapped this up myself.