Re: [Haskell-cafe] implementing a csv reader

22 Aug 2006


      On Tue, Aug 22, 2006 at 08:59:40AM -0700, Daan Leijen wrote:
...
...
...
2. I am looking for a parser, but I don't know Haskell parsers.  Is
   Parsec a good choice?
Parsec is definitely a good choice, but beware that it parses the whole
input before returning, thus it may consume a huge batch of memory. As
CSV is a line oriented format, you should make your parser lazy. Search
the mailing list archive for "lazy parser".
A good trick here is to first use "lines" to break up the input into
lines and than map a Parsec parse for each line to those lines
(returning a list of Maybe a or ParseError a results).
You can also create a lazy "many" parser using (get|set)ParserState. The
benefit is that this will also work if your elements are not in one-to-one
relation with lines and that it automatically takes care of maintaining
position in the input (for error messages).

  lazyMany :: GenParser Char () a -> SourceName -> [Char] -> [a]
  lazyMany p file contents = lm state0
    where
      Right state0 = parse getParserState file contents

      lm state = case parse p' "" "" of
                      Left err -> error (show err)
                      Right x -> x
        where
          p' = do
            setParserState state
            choice
              [ do
                  eof
                  return []
              , do
                  x <- p
                  state' <- getParserState
                  return (x : lm state')
              ]

Best regards
Tomasz

Re: [Haskell-cafe] implementing a csv reader

Tomasz Zielonka