Name resolution

The part of Spoofax I haven't really looked at yet but sounds cool is their name binding language. A (partial) implementation of semantics using declarative rules! But Stroscot is imperative, so the interesting part is the name resolution algorithm. We have "scopes", which have a parent-child relationship where lookup goes to the parent if it's not in the child.



In this example, we have scopes 1,2, and 3, and then use sites with outgoing edges and declarations with incoming edges. There are also imports and exports. Then there's a complicated tie-breaking graph-walking algorithm which determines which use maps to which definition. I don't know, I expected more. There's another paper, they point out that name resolution and type checking are often co-dependent. It seems to amount to even more complex graph walking.

Haskell doesn't have TDNR so it has a separate "renaming" pass before type-checking.