# Finding the language generated by a context-free grammar

This is a question from the Dragon book (I apologize for translation mistakes, I don´t have the English version on hand):

What language is generated by this grammar?

$S \rightarrow a S b S \mid b S a S \mid \epsilon$

I don't know what I'm supposed to do here. The definition in the book about languages says this (and that's pretty much it in the chapter):

a language is the set of all words that can be produced by any parse tree.

So, if I want to make "any" parse tree out of this grammar, I can recursively keep building it, using just the first two rules. I searched a bit and got the impression that every rule has to be used once, but I'm not sure. It would be very helpful if someone were able to provide some tips on solving these sorts of problems.

• Hint: Use regular expression – Bartosz Przybylski Mar 18 '13 at 22:20
• For tips see answers below. In answer to your question: no, it is not necessary to use every rule at least once. Start with the start symbol (or axiom) and apply the rewrite rules until you are left with only terminal symbols (here lower case). – Hendrik Jan Mar 18 '13 at 22:53
• assuming the empty String is not a terminal Symbol, to my understanding it´s not possible that only terminal symbols are left, or am I missunderstand something? – dan Mar 19 '13 at 6:17
• – Raphael Mar 19 '13 at 10:12
• @dan. The empty string vanishes, so you can end up with terminals only: $S \Rightarrow aSbS \Rightarrow aaSbbS \Rightarrow aabbS \Rightarrow aabbbSa \Rightarrow aabbba$. For example. – Hendrik Jan Mar 19 '13 at 11:04

Hint: What can you say about the number of $a$s and $b$s in the produced words?