Parser generators

Are they really better than handmade parsers? The code they produce is fucking ugly, it's cluttered with cryptic variable names and you need to read entire books to understand how they really work. Yet everyone claims writing parsers is stupid because we've had parser generators for years and they will always produce much better results.

How true is this statement? Is it akin to the ``you don't need to write ASM because compilers are much better than you'' bullshit? I, for one, have never been in a situation where I need to write my own assembly, but that doesn't make GCC any less of a piece of shit.

Having written gramer specs in lex/yacc (and lots and lots of other parser generators, from antlr to attoparsec) and hand-rolled state machines, I can attest that you must be absolutely sure that you need a handrolled machine.

I have stepped into this pile of shit more than once. Yes, handmade machines usually perform better and give you more control over error reporting and passing grammar attributes around. BUT they are a device from hell when you suddenly need to change something in your grammar a month later. Then you are in a world of shit. With parser generators, easy.