On Thu, 2010-02-11 at 13:34 -0600, Antoine Latter wrote:
On Thu, Feb 11, 2010 at 1:27 PM, Maciej Piechotka <uzytkownik2@gmail.com> wrote:
On Thu, 2010-02-11 at 11:00 -0500, Gregory Collins wrote:
Maciej Piechotka <uzytkownik2@gmail.com> writes:
On Tue, 2010-02-09 at 16:41 +0000, John Lato wrote:
See http://inmachina.net/~jwlato/haskell/ParsecIteratee.hs for a
valid
Stream instance using iteratee. Also Gregory Collins recently posted an iteratee wrapper for Attoparsec to haskell-cafe. To my knowledge these are not yet in any packages, but hackage is vast.
Hmm. Am I correct that his implementation caches everything?
The one that John posted (iteratees on top of parsec) has to keep a copy of the entire input, because parsec wants to be able to do arbitrary backtracking on the stream.
Well. Not quite. AFAIU (and ByteString implementation indicate so) the uncons have a type uncons :: s -> m (Maybe (t, s))
Where s indicates the position on the stream. Since it is impossible to get back from having s alone the GC should be free to finalize all memory allocated to cache the stream before the first living s.
I'm not sure that this is correct - parsec believes that it is free to call 'uncons' multiple times on the same value and receive an equivalent answer.
That's what I meant. But it have to keep the reference to the first element. Consider example with list: text = 'L':'o':'r':'e':'m':' ':'i':'p':'s':'u':'m':[] ^ ^ ^ s1 s2 s3 uncons s1 == Identity (Just ('e', 'm':' ':'i':'p':'s':'u':'m':[])) uncons s2 == Identity (Just ('p', 's':'u':'m':[])) uncons s3 == Identity Nothing However we will never get (nor we keep reference to): 'L':'o':'r':'e':'m':' ':'i':'p':'s':'u':'m':[], 'o':'r':'e':'m':' ':'i':'p':'s':'u':'m':[] 'r':'e':'m':' ':'i':'p':'s':'u':'m':[] so those values can be freed as they are before first pointer (namely s1). Regards