sparsepp

If you are interested in hash tables, check out my improved version of Google's already excellent sparsehash at https://github.com/greg7mdp/sparsepp. Excellent performance, very low memory usage, grows as needed of course.

It's an improved version of Google's sparsehash, with better performance and actual key deletion (instead of erasure), though with a slight memory hit. It links to a performance comparison of their sparse++ with g++ version 4.8.4's std::unordered_map, Boost version 1.55's unordered_map, and Google's Sparsehash and cpp-btree libraries, at https://github.com/greg7mdp/sparsepp/blob/master/bench.md . (While a link to the benchmark is in the README, I managed to overlook the link because it doesn't stand out.)

The post is very light on details about the measurements, and only compares to libc++'s std::unordered_{map,set} (no version number provided). There is no mention of space usage. It's well known that std::unordered_map is slow (see panic's comment). Some more worthy opponents would be Google's dense_hash_map [1] or sparsepp [2] [1] https://github.com/sparsehash/sparsehash [2] https://github.com/greg7mdp/sparsepp

Actually, if you don't need the entries to be sorted, the sparse_hash_set from https://github.com/greg7mdp/sparsepp is much faster than std::set or a sorted vector (except for iteration) and barely uses more memory than the sorted vector.