Full Consistency With HashMap

[bluss]

Missing methods (stable):

• [x] with_hasher #35
• [x] hasher #35
• [x] shrink_to_fit
• [x] is_empty #22
• [x] drain #27
• [x] values, values_mut #17

New methods:

As a rule, we don't implement unstable methods; they are implemented when they are stable in Rust beta

• [x] get_key_value rust-lang/rust/issues/49347
• [x] Entry::and_modify (unstable)
• [x] Entry::or_default (unstable)
• [ ] try_reserve rust-lang/rust/issues/48043
• [x] remove_entry rust-lang/rust/issues/46344

Trait impls

• [x] PartialEq + Eq #23
• [x] Extend<(&K, &V)> #34
• [x] Ser/De #28
• [x] ESI for iterators #26
• [x] Debug for Entry
• [x] Clone for some iterators (Keys, Values, Iter, probably all of them)

Type Conflicts:

• [x] Entry<K, V> vs Entry<K, V, S> (not important) #59
• [x] retain's &K vs &mut K #39
• [ ] drain's signature (not important)
• [x] OccupiedEntry::insert's signature #68

[sugar700]

#28 implements Ser/De

[bluss]

I can't find any discrepancies between HashSet and OrderSet. A second close look is of course welcome, but I think @cuviper already did this carefully when implementing OrderSet, so it should already be fine.

[CAD97]

indexmap::map::Keys is not Clone where std::collections::hash_map::Keys is. (Just noticed this moving some code that needed stable ordering over.) Is this an innate limitation or just an oversight?

[bluss]

Just an oversight, please check all iterators you can and PRs welcome.

[cuviper]

I think we are as close as we can get at this point. New additions to std can continue being considered individually.