Compiling ascent at runtime about ascent HOT 2 CLOSED

Hi there. Arash graduated his PhD last year, and has less time to add new features to Ascent. This is a great idea, and I'd like consider adding it. Our lab is looking into a fork of Ascent where we support incremental compilation and similar concerns--but I believe the current Ascent has no plans to grow towards incremental compilation right now.

To give an idea of why this is hard: when compiling Datalog, you perform index selection based on the total computation (as you likely know, this is what distinguishes Datalog from SQL). When you add new rules, you may necessitate new indices, which may require a migration from a pre-existing index. However, there are trade-offs here: in some circumstances you may want to just scan-then-filter. At its limit, adding incremental rules to Datalog starts to look a lot like SQL, which is why it is an interesting problem. I am happy to brainstorm more if you'd like, but for now I am going to close the issue since I think it's not within scope of the current Ascent. Feel free to email me: [email protected]

from ascent.

Thanks for the update, congrats to Arash!

from ascent.