[Adam2025]

Nice work. Roaring bitmaps fit a sweet spot for fast set operations with good compression, so seeing them brought into PostgreSQL is genuinely interesting. The fact that it plugs into native types instead of reinventing the wheel is a big plus.

I would love to see real-world benchmarks. In theory this should outperform intarray and some GIN setups for certain cardinalities, but actual production numbers often look different.

If anyone has tried this with large tables or event-tracking workloads, your results would be great to learn from.