Question

Consider unordered_map:

template<
    class Key,
    class T,
    class Hash = std::hash<Key>,
    class KeyEqual = std::equal_to<Key>,
    class Allocator = std::allocator< std::pair<const Key, T> >
> class unordered_map;

I know (a==b) is faster than !(a<b) && !(b>a), but since unordered_map does not use std::less<Key> to compare/store the keys in the map, I wonder how can an implementation profit by tree data structures in the most efficient way to read/store different keys in the same bucket. It seems that a convertion from Key to a sort of KeyWrapper with operator<() defined cannot be avoided by any implementation with trees.

Answer 1

You cannot use a tree inside unordered_map, even just within a bucket. unordered_map‘s interface simply does not allow it. The key type is required only to be equality comparable, nothing more. That’s why it’s called an “unordered” map; because there is no specific ordering of elements. To use some kind of binary tree would require a strict weak ordering, which is not required.

If you want to use a variation of an unordered_map, you may. But it would not be an unordered_map as defined by the standard.