Question

I’m looking for a std::map-esque data structure optimised for fast lookup.

One approach would be to implement map’s interface utilising a sorted std::vector as the underlying storage – this will offer fast binary_search thanks to random-access iterators and cache locality.

However, this sounds like reinvention of the wheel. Surely something like this already exists?

Is there an open-source ordered associative data structure which uses a std::vector for storage?

Edit:

In response to the comments suggesting just use std::map – please read here: http://lafstern.org/matt/col1.pdf

Answer 1

The Boost.Containers library has an ordered map container whose storage is backed by a contiguous array called boost::flat_map. Note however that the asymptotic, theoretic complexity is the same as for the standard map (logarithmic), and the choice which is better depends on many details of the use case: insertions vs. lookups, iterations, iterator invalidation requirements.

Since the interfaces are very similar, it should be possible to literally replace one by the other via a typedef and profile the relative performances, which is something you absolutely must do.