On Tue, Aug 01, 2023 at 01:26:19PM +0200, Han-Wen Nienhuys wrote:

quoted

One thing that I was able to do to produce slightly smaller Roaring+Run

Just for my edification: what is "Roaring + Run" ?

Roaring+Run refers to a modified version of the Roaring bitset
compression scheme which has an additional container type to represent
runs of a single (set) bit.

The run container stores the lower 16-bits of bit position of the start
of the run, and then it uses another 16-bit value to store the length of
the run.

For more, this paper from David Lemire is a good start:

  https://arxiv.org/pdf/1402.6407.pdf

Thanks,
Taylor