David, thank you for your helpful explanations. So I need to define a new datatype encapsulating the two I want to parse and then create a new Stream, is that correct? Reading your comments and more of the haddock documentation of Parsec, this came to my mind: say I want to skip over some binary parts of the input (which is a PDF, containing ascii and deflated parts as well as images). I found the functions setPosition/getPosition, which seem to build on the notion of line and column number. Is there a way to say that I want to skip N bytes along the stream? I didn't find an obvious one myself when browsing through the docs. Thanks again, Christian 2010/8/6 David Virebayre <dav.vire+haskell@gmail.com<dav.vire%2Bhaskell@gmail.com>

On Fri, Aug 6, 2010 at 11:03 AM, C Gosch <ch.gosch@googlemail.com> wrote:

...

You're right, I probably used the wrong words .. I meant that apparently the tokens Parsec uses are of type Char, and I would actually at some point like to continue parsing, but using different tokens. Sorry if I still got it wrong, I'm new :) I can post some code later, as I don't have it here right now.

Parsec (at least version 3) uses any type of token you want.

Quick example off the top of my head, I didn't check if it compiles:

-- you have to make lists of your token type an instance of Stream :

instance Stream [ MyTokenType ] Identity MyTokenType where uncons [] = return Nothing uncons (x:xs) = return $ Just (x,xs)

-- your parser type is going to look like this :

type MyParser a = ParsecT [MyTokenType] () Identity a

-- assuming your toke type looks like this

Data MyTokenType = A Char | B Word8 deriving (Show)

-- you need a basic parser from which you can make more complicated ones

satisChar :: ( Char -> Bool ) -> MyParser Char satisChar f = tokenPrim prt pos match where prt = show pos p _l _cs = incSourceLine p 1 match (A c) = if f c then Just c else Nothing match _ = Nothing

satisBin :: ( Word8 -> Bool ) -> MyParser Word8 satisBin f = tokenPrim prt pos match where prt = show pos p _l _cs = incSourceLine p 1 match (B w) = if f w then Just w else Nothing match _ = Nothing

-- You can define basic parsers like this

-- parse any letter letter = satisChar (const True)

-- parse a specific char, it will return char c = satisChar (==c)

-- parse any binary word binary = satisBin (const True)

-- parse a specific binary word w = satisBin (==w)

-- now you can combine this to make more complicated parsers.

...

Ah wait ... I saw that I can get the input stream with getParserState, as it is part of the state. If that is the currently remaining stream, I could use any functions from ByteString to skip over some parts or do whatever with them, and update the state to the last position after the part I want to skip. Am I correct there, or am I then getting something in Parsec out of sync? Guess I'll just try :) Thanks! Christian 2010/8/6 Patrick LeBoutillier <patrick.leboutillier@gmail.com>

Hi,

I started using Data.Binary yesterday for a project and it's nice. I'm not sure it's possible, but perhaps you could switch tools along the way:

- Parse the ASCII bit with Parsec ; - Give the remaining ByteString to Data.Binary for the binary part ; - Switch back and forth as required.

Patrick

On Fri, Aug 6, 2010 at 5:03 AM, C Gosch <ch.gosch@googlemail.com> wrote:

...

2010/8/6 Dean Herington <heringtonlacey@mindspring.com>

...

At 11:42 PM +0200 8/5/10, C Gosch wrote:

...

Hi, does anyone here know their way around in Parsec? I'm trying to parse a file which contains some binary parts too. I have been using Parsec 3.0.1 to parse the first ASCII part, but all

the

...

...

...

parsers are returning Char type tokens, so it would not work with the binary parts .. is there any way to do this? I am using Text.Parsec.ByteString.Lazy.

Note that I'm new to Haskell and Parsec, and doing this within a small project that I basically do to get a grip on Haskell (and Parsec, because it appears to be really nifty and I want to learn more about it).

Thanks for any hints, Christian

Parsing binary parts of a file is not inherently a problem. A parser can return any type; different parsers in your application will likely return different types.

If you include some code we'll be able to help you more specifically.

Dean

You're right, I probably used the wrong words .. I meant that apparently the tokens Parsec uses are of type Char, and I would actually at some point like to continue parsing, but using different tokens. Sorry if I still got it wrong, I'm new :) I can post some code later, as I don't have it here right now.

Cheers, Christian

_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners

-- ===================== Patrick LeBoutillier Rosemère, Québec, Canada