Parser Generator

I'm using lex and yacc for an assignment right now (actually, alex and happy).

The lexer generator builds the code to tokenize your source code based on your regular expressions and other rules; the parser generator builds a parser (generally LALR(1)) for your language's context-free grammar. Run yacc and lex on your files, and they crank out some C/C++/Haskell/Java source code that you can just call from your actual compiler.

I write the grammar for my "toy language" into yacc and it builds a parser for that language and outputs a syntax tree, which I can then use to evaluate or to generate machine code.

Since lexer and parser generation is boring, you tend to get boredom-induced errors when you write them by hand, which is a perfect task for a soulless, unthinking machine. [grin]

Check out the IBM Linux games tutorial for an example of a yacc and lex source file that's used for an in-game scripting language.

The question is not really "why do you need a parser generator", because the answer is "to create a parser for a language for which there is none". The actual question is "which languages don't already have a compiler?" and the answer here is: a lot. Every time someone creates a prototype language (happens on a weekly basis in academia) a parser generator is a simple way of generating a parser without having to spend days on writing it yourself.

I participated in a programming competition yesterday. I wrote a small parser that could handle logical expression of the form (((a v b) ^ (c x d)) <-> e), take some value binding in and tells you wether the expression is true or not.

I did it by hand since I didn't think using a compiler compiler would be useful for the small scope. However, the other team did it slightly faster than me using javacc.

Bottom line, ad-hoc languages are useful for some applications, compiler compilers are a good way to build an interpreter quickly.