Current implementation of the compiler walks the parse tree generated by ANTLR, which is a bit cumbersome because of the way the grammar has to be written to deal with priorities/ambiguities. Luckily, many of such issues are solved by the interpreter implementation, which walks the parse tree and generates a neat, simpler Abstract Syntax Tree.

This tasks is to refactor our current compiler to walk that AST, instead of the larger parse tree.

Benefits:

• less "visit" cases in the visitor
• more opportunities to "optimize" the tree (e.g, tree rewrites, constant folding)
• more opportunities to refactor tree walking phases (e.g. we could add a "type-evaluation" phase)
• a chance to refactor code gen utilities to support classes, and simplify the code in general

• cons: it takes a bit of time, but the AST nodes are really a handful, so it should not take that much