I've been thinking really hard today about type-checking and type inference for this funny BI-like type system I want to get working. I have some notions about how to simplify the thorny general problem to make it more tractable, so I want to get some implementation up and running to see whether it works on the examples I care about. I poked around the MLKit a little, believing it was designed to be more "hackable" than smlnj or mlton, but I am not totally convinced it is. Any opinions from the peanut gallery as to what the shortest path would be to hooking in a little bit of fancy refinement checking into an SML compiler that doesn't actually affect the operational semantics at all?
I've been thinking really hard today about type-checking and type inference for this funny BI-like type system I want to get working. I have some notions about how to simplify the thorny general problem to make it more tractable, so I want to get some implementation up and running to see whether it works on the examples I care about. I poked around the MLKit a little, believing it was designed to be more "hackable" than smlnj or mlton, but I am not totally convinced it is. Any opinions from the peanut gallery as to what the shortest path would be to hooking in a little bit of fancy refinement checking into an SML compiler that doesn't actually affect the operational semantics at all?
- Post a new comment
- 14 comments
- Post a new comment
- 14 comments
