Translating Modern Compiler Implementation in ML

[lee-thomp]

How would I go about translating the following from Figure 1.4 of "Modern Compiler Implementation in ML" into Coalton? I'm stuck on the mutual recursion of stm and exp:

type id = string
datatype binop = Plus | Minus | Times | Div
datatype stm = CompoundStm of stm * stm
             | AssignStm of id * exp
             | PrintStm of exp list
     and exp = IdExp of id
             | NumExp of int
             | OpExp of exp * binop * exp
             | EseqExp of stm * exp

[stylewarning]

The only thing that's not directly supported are type aliases. That's a feature being designed right now. So you'd either need to just use a dedicated type for that, or don't use it at all.

In this example, I'll just make a new type called Ident. A translation would look something like this.

(coalton-toplevel
  (define-type Ident
    (Ident String))

  (define-type BinOp
    Plus
    Minus
    Times
    Div)

  (define-type Stm
    (CompoundStm Stm Stm)
    (AssignStm Ident Exp)
    (PrintStm (List Exp)))

  (define-type Exp
    (IdExp Ident)
    (NumExp Integer)
    (OpExp Exp BinOp Exp)
    (ESeqExp Stm Exp)))

I would love to see "Modern Compiler Implementation in Coalton"! 😊

[lee-thomp]

Ah okay I found the source of my issue. I'd actually translated it exactly as you described before posting my message. I'd managed to catch the issue where Id is already defined in Coalton, but not where Exp is already defined (coalton-library/math/elementary). Using Sly with Emacs and SBCL showed me a whole backtrace of type check errors but nothing specific about Exp itself. I only caught this after loading this all into a bare SBCL repl where I was told about the math package already having its own Exp.

Replacing Exp with Expr for example fixed all this and typechecks fine.

[stylewarning]

Ah great!

I recommend defining your own package and using package-local nicknames. :)