@mflatt I read a paper recently that might have some relevance to speeding up set-of-scopes expansion, so I wanted to bring it to your attention just in case.
All the stuff about the particular use case aside, the main result of the paper seems to be that bignums are a surprisingly efficient data structure to represent sparse finite sets.
In particular, individual scopes would just be prime numbers, subset checks would be divides? etc.
It seemed relatively easy to apply in your simpler models of the expander (nano, pico, etc) but I’m not really clear what complications would affect things in the big picture.
Anyhow, not sure if that’s of any use, but wanted to bring it to your attention just in case.
@mrm Thanks – I think I talked to Matt about this early on in the development of scope sets. I can’t remember the discussion in any detail, though, and I may misremember what a particular conversation was about. I’d certainly be interested in the results if anyone tries out this representation.