Krzysztof Kościuszkiewicz wrote:
Another story from an (almost) happy Haskell user that finds himself overwhelmed by laziness/space leaks.
I'm trying to parse a large file (>600MB) with a single S-expression like structure. With the help of ByteStrings I'm down to 4min processing time in constant space. However, when I try to wrap the parse results in a data structure, the heap blows up - even though I never actually inspect the structure being built! This bugs me, so I come here looking for answers.
Note that Parsec has to parse the whole file before it can decide whether to return a result (Left _) or an error (Right _). ghc would have to be quite smart to eliminate the creation of the expression tree entirely. The polyparse library (http://www.cs.york.ac.uk/fp/polyparse/) offers some lazy parsers, maybe one of those fits your needs. Text.ParserCombinators.Poly.StateLazy is the obvious candidate. HTH, Bertram