Package ranksel provides a bit vector that can answer rank and select queries. More specifically, it implements the data structure described by G. Navarro and E. Providel's A Structure for Plain Bitmaps: Combined Sampling in Fast, Small, Simple Rank/Select on Bitmaps with some minor modifications.

go get github.com/robskie/ranksel

Godoc documentation can be found here.

All these benchmarks are done on a machine with a Core i5 running at 2.3GHz. Note that Select0 is slower than Select1 in most cases. This is because the implementation of Select1 is more optimized than Select0. The only case where Select0 is faster than Select1 is when the bit density is lower than 3% as shown from BenchmarkSelectDX where X is the bit density. Another thing to point out is that Select1 starts to slow down when the bit density gets lower than 3%. So you might want to use another data structure if you have a sparse bitmap and you want a fast Select1 operation.

You can run these benchmarks by typing go test github.com/robskie/ranksel -bench=.* from terminal.

BenchmarkRank1      10000000    160 ns/op
BenchmarkRank0      10000000    170 ns/op
BenchmarkSelect1     5000000    253 ns/op
BenchmarkSelect0     3000000    427 ns/op
BenchmarkSelect1D3   5000000    261 ns/op
BenchmarkSelect0D3  10000000    207 ns/op
BenchmarkSelect1D2   5000000    336 ns/op
BenchmarkSelect0D2  10000000    195 ns/op
BenchmarkSelect1D1   3000000    561 ns/op
BenchmarkSelect0D1  10000000    214 ns/op