Flesh out functionality

[FintanH]

We should flesh out the functionality of nonempty. A lot of it would just be the Vec implementation.
Enumerating these for clarity and also to comment on which ones we may not want:

• append
• as_mut_ptr N/A (memory is not contiguous)
• as_mut_slice N/A
• as_ptr N/A
• as_slice N/A
• capacity
• chunks
• chunks_exact
• chunks_exact_mut
• chunks_mut
• clone_from_slice - we use from_slice, we should rename This doesn't make sense either since it's just copying a slice's contents in, but we can't guarantee a length of 1 without panicing.
• concat
• connect
• contains
• copy_from_slice - we have from_slice but a Copy version would be good Looking into the semantics of this, I think it doesn't suit NonEmpty after all.
• copy_within
• clear - we probably don't want this
• dedup
• dedup_by
• drain - wouldn't make sense for NonEmpty unless it returned a Vec rather than Drain
• drain_filter - same as above
• ends_with
• eq_ignore_ascii_case
• extend_from_slice
• first
• first_mut
• from_raw_parts - this seems unecessary
• get
• get_mut
• get_unchecked
• get_unchecked_mut
• insert
• into_boxed_slice
• into_raw_parts - not sure about this one
• is_ascii
• is_empty
• is_sorted
• is_sorted_by_key
• iter
• iter_mut
• join
• last
• last_mut
• leak - seems specialised
• len
• make_ascii_lowercase
• make_ascii_uppercase
• new
• partition_at_index
• partition_at_index_by
• partition_at_index_by_key
• partition_deup
• partition_deup_by
• partition_deup_by_key
• pop
• push
• rchunks
• rchunks_exact
• rchunks_exact_mut
• rchunks_mut
• remove - should give back Vec
• remove_item - should give back Vec
• repeat
• reserve
• reserve_exact
• resize - possibly just work on the tail Vec
• resize_default
• resize_wtih
• retain - unsure if this should be kept
• reverse
• rotate_left
• rotate_right
• rsplit
• rsplit_mut
• rsplitn
• rsplitn_mut
• set_len
• shrink_to - should be greater than 1
• shrink_to_fit
• sort
• sort_by
• sort_by_cahced_key
• sort_by_key
• sort_unstable
• sort_unstable_by
• sort_unstable_by_key
• splice
• split
• split_at
• split_at_mut
• split_first
• split_first_mut
• split_last
• split_last
• split_last_mut
• split_mut
• split_off
• splitn
• splitn_mut
• starts_with
• swap
• swap_with_slice
• to_ascii_lowercase
• to_ascii_uppercase
• to_vec - the Into impl does this
• try_reserve
• try_reserve_exact
• windows
• with_capacity
• From<NonEmpty<T>> for Vec<T> - Implement and test IntoIterator #17 (comment)

[FintanH]

Hmmmm so something I'd like to discuss is the functions that come "Methods from Deref<Target = [T]>". Is this something that's possible with NonEmpty? Also, this makes me think, could we implement Deref<Target=Vec<T>>? 🤔

Otherwise, I think the way to go about it is go through the list of functions and see if they're possible to implement.

[FintanH]

As discussed async, we should use https://doc.rust-lang.org/std/num/struct.NonZeroUsize.html for things that take a length, e.g. with_capacity

[FintanH]

Something I'd also be interested in is split functions that own the T, I'm not sure if this is possible though. I tend to split and consume, but I end up having to clone :/

[fritzrehde]

In the checklist it says as_mut_slice is implemented, but I can't find that method anywhere in the repo or in the documentation.

[cloudhead]

I think it's checked for the same reason as_slice is checked: it's actually not possible to implement these since the memory isn't contiguous. I'll update the todo.

[bbstilson]

@FintanH is it cool if I open a PR for with_capacity? I have a use-case that would benefit, and I'm happy to contribute 🙂

Branch is up here: https://github.com/cloudhead/nonempty/compare/master...bbstilson:nonempty:feature/new_with_capacity?expand=1

[FintanH]

Hey @bbstilson! Yes please do, I’ll get around to merging it when I can ☺️

[FintanH]

I’m just looking at your branch and have two questions:

• Why is the capacity not allowed to be zero?
• If it’s not allowed to be zero, than why not take NonZero?

[bbstilson]

I’m just looking at your branch and have two questions:

* Why is the capacity not allowed to be zero?

* If it’s not allowed to be zero, than why not take NonZero?

Great questions. I think my brain just checked out for a bit because I originally implemented with NonZeroUsize, then deleted it. Anyways, I added it back in and opened a PR haha: #84

[bbstilson]

How desirable is it to modify the NonEmpty struct such that memory is contiguous? The current implementation prevents a lot of Vec operations (all the chunks, for example).

(edit)

I did this locally (happy to push a branch); it was pretty straightforward. The biggest consequence is constructors are no longer const compatible because you can't instantiate a vec at compile time.

[bbstilson]

@FintanH bumping ☝🏻