#The Monkey Programming Language

Python 3 implementation of the interpreter for the Monkey language from Thorsten Ball's Writing an Interpreter in Go.

#Notable Implementation Deviations from the Book

• The interpreter in the book does not track the locations of tokens within source code. This interpreter tracks the line number (as delimited by the ASCII line feed \n) of tokens in via a SourceLocation data structure and uses that line number when reporting parse errors:

  $ cat examples/parse-error.monkey
  let add = fn(x, y) { x + y }
  add(1, $)
  $ ./monkey.py examples/parse-error.monkey
  [examples/parse-error.monkey, line 2] Expected expression, found ILLEGAL($)
  

• The interpreter in the book uses the type int64 for the Value field of the IntegerLiteral struct. This interpreter uses Python's builtin arbitrary precision int type to represent the value of integers.
• The interpreter in the book propagates nil up the stack upon parse failure. Rather than return the equivalent None this interpreter chooses to instead raise a ParseError exception.
• The interpreter in the book creates and reuses a single instance of the boolean-true, boolean-false, and null objects. This interpreter creates new instances of these object any time such values are produced during evaluation. The book correctly notes that using a single instance of these objects is (likely) faster and will save on resources, but in my opinion it creates much uglier code, and performance is not a particularly important issue for a toy interpreter.
• The interpreter in the book returns nil as the result of a call to evaluator.Eval when passed an ast.LetStatement. Returning nil from a function that should always return an object.Object violates type safety, so this interpreter chooses to instead return a null object.
• The interpreter in the book creates a struct, HashKey, for holding the keys of a hash data type. Rather than define a separate type, this interpreter uses Python's builtin __hash__ and __eq__ methods for hash key comparison.