andy-morris/miniml

A tiny functional language in C++ [university stuff]

MiniML—A tiny functional language

Installation

Needs a C++11 compiler, lemon, and ragel to build. (The release contains pre-made versions of the output files.) Then make should Just Work (tested with GCC 4.9 on Debian and Xcode's version of clang on OS X).

Quick language guide

• Has types int, bool, string, tuples (written (type₁, type₂, ...), and (higher-order) functions (written type₁ -> type₂).

• Operators: <-> || && < <= > >= == != + - * / with what is hopefully the obvious precedence.

  • <-> is "iff", i.e. equality on booleans, with the lowest precedence. Writing this, I just realised == doesn't work on strings, only ints. Sorry about that.

• Other expressions:

  • if branch then_case else_case
  • fn (x: type) => expr
  • expr: type (type ascription, though not actually useful…)
  • expr₁; expr₂ (sequencing)
  • (expr₁, expr₂, ...) (tuples, also ())
  • tuple.0, tuple.1, etc (0-based projection)
    • I don't mean to say it needs to be a tuple literal, just something of some tuple type with enough elements.

• REPL inputs terminated with ;;.

• Declarations:

  val x: type = def       // type optional
  val rec x: type = def   // recursive; type not optional
  fun x (y: type₁) (...): type₂ = def
      // abbreviation for:
      // val rec x: type₁ -> … -> type₂ =
      //     fn (y: type₁) => ... => def
  

• Builtins (defined in src/init_env.cxx for magic functions or prelude.mml for nonmagic ones):

  // magic:
  print: string -> ()
  app: string -> string -> string
  string_int: int -> string
  newline: () -> ()
  use: string -> ()
  // in prelude:
  println: string -> ()
  print_int: int -> ()
  even, odd: int -> bool
  

  Since use needs a reference to the environment, it's defined in Repl's constructor instead (src/repl.cxx).

• “Modules” (well, files) have syntax name => decl₁ decl₂ .... Note that ;; isn't used in files, only interactively.

• As you probably notced by now, comments are of the // variety.

Notes

• GCC doesn't check exhaustiveness of switch, which is what the #ifdef __GNUC__ business is all about.
• #pragma GCC diagnostic is understood by Clang too.
• If something seems redundant (say, non-builtin type names), it probably is. There was going to be a lot more to this but I ran out of time.