Josh Stone

I found [halfbrown](https://crates.io/crates/halfbrown) does this as an enum wrapper over `hashbrown::HashMap`. Someone could do the same kind of thing with `IndexMap`.

In retrospect, there is a way around this: > But since I cannot `impl Borrow for MyBigStruct`, I cannot use `IndexSet::get` to pass in `HashEssence`. The function signature on `IndexSet`...

Sure, it doesn't solve everything that `RawEntry` does, but it does seem like enough for what @SuperFluffy had. It's also _possible_ to use something like [nohash](https://crates.io/crates/nohash) if you control the...

WIP: https://github.com/indexmap-rs/indexmap/compare/master...cuviper:indexmap:raw_entry

> Perhaps `RawEntryApiV1`. Sure, I like that too. > Types like `RawEntryBuilder` I think needs default value for type parameter `S`. That's *possible*, though annoying to accomplish with `#[cfg(feature =...

The ideal would be to have this added *and stabilized* in `std` first, so the API precedent is locked in. AFAIK, that hasn't even been proposed yet, which should go...

This was fixed in #219 to return `(&K, &mut V)`, and `MutableKeys::get_index_mut2` for the "dangerous" behavior. I'll be releasing 2.0 soon!

Note that `hashbrown` still only provides `new` and `with_capacity` when "ahash" is enabled... those **can't** be implemented with that uninhabited dummy type (unless they would just `panic!`). I'm not really...

Note that `indexmap 2` has been released, with its change to no longer detect `std` implicitly. However, this does not change the situation around `new` and `with_capacity`, and I left...

The `Slice` API shipped in 2.0.0 with order-aware comparisons -- hope that helps! ``` use indexmap::IndexMap; fn main() { let m1: IndexMap = vec![(1, 1), (2, 2)].into_iter().collect(); let m2: IndexMap...