Recursive Descent Parsers

From Progzoo
Revision as of 15:01, 7 March 2011 by Andr3w (Talk | contribs) (A Recursive Descent Parser in Java)

Jump to: navigation, search

A Recursive Descent Parser takes a stream of tokens and builds a syntax tree.

JSON Data Format

As an example we will take the JSON data format. This is simple, but it does have the recursive structure that shows off the technique. JSON is used to represent data. There are two primitive type: string and number.

  • "one"
  • 123

There are lists For example:

  • [123, "one" ]

There are hash arrays For example:

  • {num:123, str: "one" }

You can have a list inside a list of course:

  • [1, 2, [3, 4]]

Similarly you can have a hash array of lists, and in JavaScript you can mix types:

  • {a:1, b:[1,2,3], c:{x:2}}

Production Rules

An eta ε production allows the empty string as an option.

N -> number
N -> [ N Z ]
N -> { word : N Y }
Z -> , N Z
Z -> ε
Y -> , word : N Y
Y -> ε

Productions such as

N -> number

Are permitted because the lexer can recognise numbers, similarly strings.

A Recursive Descent Parser in Java

  • Each of the non-terminals N Z and Y have a method ParseN, ParseZ and ParseY.
  • In each case the method takes the token stream as input and does the following:
    • Consumes all tokens associated with the thing parsed
    • Returns a data structure representing the thing parsed
static Object ParseN(StreamTokenizer st) throws Exception
{
	st.nextToken();
	if (st.ttype == StreamTokenizer.TT_NUMBER)
		return null;
	if (st.ttype=='[')
	{
		ParseN(st);
		ParseZ(st);
		if (st.nextToken()!=']')
			throw new Exception("Was expecting ]");
		return null;
	}
	… section missing …
	throw new Exception("Unexpected input.");
}

Recursive Descent Parser Tutorial