Hi, So far I did not regret using Parsec for a huge language (like CASL http://www.tzi.de/cofi). Compared to other combinator parser libraries, I think, it is best documented and the fastest (if used sensibly). Error reporting could be better, but it is already better than that of happy. Usually, one does not want to stop at the first parse error, but rather try to collect several errors (that are not caused by previous ones). I assume, alex and happy generated code will be faster. (I did not try to use a seperate lexer with Parsec.) But a slow-down of the parser (say by a factor of up to 3) is usually no problem. Much more time is spent elsewhere. One disadvantage of combinator parsers is that the "grammar" is not as visible as in a happy file and needs more tuning (left factorization, etc.). The major advantage is the great flexibility in combining parsing and other computations (possibly for lookaheads or dependent parsing) in one formalism (i.e. pure hasekll in a single module). (The yacc features like precendences and associativities in grammar rules do not help much for user-definable precendences like in haskell.) Christian