Benchmarking is a great way to prove a point – numbers win arguments easily.
Yeah, you can do it. In fact, I think we might not even need the algo.UIDList anymore, and we can use the memory management techniques to do filtering inplace etc. Let’s do it in a series of PRs, each not more than 300 lines of changes or so… even if they break the build, it’s alright. This is a good chance to clean up our code base, and refactor code as well – so I want to reviewing every PR.
Good luck! And happy to discuss anything related to this.