Add <[[T; N]]>::flatten{_mut}

[Cyborus04]

Adds flatten to convert &[[T; N]] to &[T] (and flatten_mut for &mut [[T; N]] to &mut [T])

[rust-highfive]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @yaahc (or someone else) soon.

Please see the contribution instructions for more information.

[scottmcm]

Since it's now (#95016) agreed to be sound, consider adding Vec::flatten to go with these as well.

(Or maybe it should be into_flattened or something to not shadow the slice method.)

[Cyborus04]

r? @scottmcm

[Cyborus04]

Wow, is there really no usage of unchecked_* math in alloc? That can't be right...

[scottmcm]

Wow, is there really no usage of unchecked_* math in alloc? That can't be right...

It doesn't surprise me, really. As the feature explanation says, it's a niche optimization path where it's extremely rare for it to be actually relevant. And the building blocks that are foundational enough to use it tend to be in core.

[Cyborus04]

Oh, I missed that before I pushed that commit. Should I just used normal * multiplication then?

[scottmcm]

Oh, I missed that before I pushed that commit. Should I just used normal * multiplication then?

Naw, it's fine. It's plausibly helpful for optimization, since otherwise LLVM doesn't know that the slice didn't get shorter.

[leonardo-m]

Another use case is to convert &[[T; N]; M] to &[T; N * M]. That's a case of array reshaping in general.

[Cyborus04]

🤦🏻‍♂️

[Cyborus04]

oh uh messed this one up a bit

[scottmcm]

As far as I know we don't want to use anything that's still incomplete_features in the bootstrap compiler. Unstable is fine when it's decently-baked, but if compiler says it's incomplete, we don't use it in libs.

---

I found a few more nits, and I have one extra request: Can you add #[should_panic(expect = "... tests for the overflow cases, please?

So something like

#[test]
#[should_panic(expect = "vec len overflow")]
fn vec_into_flattened_panics_for_capacity_overflow() {
    let n = 1 << (usize::BITS / 2);
    let _ = vec![[(); n]; n].into_flattened();
}

which would probably go in https://github.com/rust-lang/rust/blob/master/library/alloc/tests/vec.rs

And analogous ones for the two slice methods.

When you're done, please comment @rustbot ready to update the tags accordingly.

[Cyborus04]

wait i put it in the wrong file 😅

[Cyborus04]

What do I do about the feature flag issue? There's other tests that use unstable features without any #[feature(...)] attributes

[scottmcm]

I think you need to allow the new feature in the tests library, like all of these ones

rust/library/alloc/tests/lib.rs

Lines 1 to 40 in c2afaba

	#![feature(allocator_api)]
	#![feature(alloc_layout_extra)]
	#![feature(assert_matches)]
	#![feature(box_syntax)]
	#![feature(cow_is_borrowed)]
	#![feature(const_box)]
	#![feature(const_convert)]
	#![feature(const_cow_is_borrowed)]
	#![feature(const_heap)]
	#![feature(const_intrinsic_copy)]
	#![feature(const_mut_refs)]
	#![feature(const_nonnull_slice_from_raw_parts)]
	#![feature(const_ptr_write)]
	#![feature(const_try)]
	#![feature(core_intrinsics)]
	#![feature(drain_filter)]
	#![feature(exact_size_is_empty)]
	#![feature(new_uninit)]
	#![feature(pattern)]
	#![feature(trusted_len)]
	#![feature(try_reserve_kind)]
	#![feature(unboxed_closures)]
	#![feature(associated_type_bounds)]
	#![feature(binary_heap_into_iter_sorted)]
	#![feature(binary_heap_drain_sorted)]
	#![feature(slice_ptr_get)]
	#![feature(binary_heap_retain)]
	#![feature(binary_heap_as_slice)]
	#![feature(inplace_iteration)]
	#![feature(iter_advance_by)]
	#![feature(round_char_boundary)]
	#![feature(slice_group_by)]
	#![feature(slice_partition_dedup)]
	#![feature(string_remove_matches)]
	#![feature(const_btree_new)]
	#![feature(const_default_impls)]
	#![feature(const_trait_impl)]
	#![feature(const_str_from_utf8)]
	#![feature(nonnull_slice_from_raw_parts)]
	#![feature(panic_update_hook)]

[scottmcm]

Let's see if it'll just let me commit these suggestions...

[Cyborus04]

Ohhh, thank you! I've been confused about this all day, glad it's finally solved 😀

[Cyborus04]

uh oh i think i destroyed your commit by accident, sorry. forgot to pull first

[scottmcm]

Looks good! Thanks for your perseverance through all the hiccups.

@bors r+

[bors]

📌 Commit 06788fd has been approved by scottmcm

[scottmcm]

Congrats on your first PR landing, @Cyborus04!