在OCaml中构建AST

Question

I'm using OCaml to build a recursive descent parser for a subset of Scheme. Here's the grammar:

    S -> a|b|c|(T)
    T -> ST | Epsilon

So say I have:

   type expr = 
       Num of int | String of string | Tuple of expr * expr

Pseudocode

These functions have to return expr type to build the AST

parseS  lr =
   if head matches '(' then
     parseL lr
   else
     match tokens a, b, or c

Using First Set of S which are tokens and '(':

parseL lr = 
   if head matches '(' or the tokens then
       Tuple (parseS lr, parseL lr)
  else
       match Epsilon

My question is "How do I return for the Epsilon part since I just can't return ()?". An OCaml function requires same return type and even if I leave blank for Epsilon part, OCaml still assumes unit type.

Answer 1

As far as I can see, your AST doesn't match your grammar.

You can solve the problem by having a specifically empty node in your AST type to represent the Epsilon in your grammar.

Or, you can change your grammar to factor out the Epsilon.

Here's an equivalent grammar with no Epsilon:

S -> a|b|c|()|(T)
T -> S | S T

Answer 2

也许不是手动创建解析器函数而是使用现有方法更好：例如，基于LALR（1） ocamlyacc或基于camlp4的LL（k） 解析器 ？