[MikkelPaulson]

Added lifetimes where possible referring to the parent HashMap/ HashSet rather than the entry itself. This is not possible for VacantEntry or by extension the Entry itself, but OccupiedEntry already opaquely returns references from the parent.

Affected methods:

• fn map::OccupiedEntry::key() -> &'a K;
• fn map::OccupiedEntry::get() -> &'a V;
• fn map::OccupiedEntry::get_mut() -> &'a mut V;
• fn set::OccupiedEntry::get() -> &'a V;

Also rewrote tests to inherently validate that the lifetimes are adequate (tests fail to compile against current main), and added some rudimentary coverage for the HashSet entry API.

[cuviper]

No, this is unsound -- e.g. it would allow multiple calls to get_mut() for aliased references, or for any of these references to become dangling by calling remove afterward. This is why into_entry and into_mut both consume the entry.

[MikkelPaulson]

Oh, good point. Would you be open to adding hash_set::OccupiedEntry::into_ref(self) -> &'a T?

[cuviper]

Hmm, right, there's no way to get the full lifetime from the set entry otherwise. Adding into_ref sounds ok to me.