Am Sonntag 17 Januar 2010 14:30:36 schrieb Mark Spezzano:
Hi,
I am writing a Text Adventure game in Haskell (like Zork)
I have all of the basic parser stuff written as described in Hutton's Programming in Haskell and his associated papers. (I'm trying to avoid using 3rd party libraries, so that I can learn this myself)
Everything that I have works (so far...) except for the following problem:
I want to define a grammar using a series of Verbs like this:
data Verb = Go | Get | Jump | Climb | Give etc, etc deriving (Show, Read)
and then have my parser "get" one of these Verb tokens if possible; otherwise it should do something (?) else like give an error message stating "I don't know that command"
Now, Hutton gives examples of parsing strings into string whereas I want to parse Strings into my Verbs
So, if the user types "get sword" then it will tokenise "get" as type Verb's data constructor Get and perhaps "sword" into a Noun called Sword
But the Read instance can only read "Get", not "get". You'd have to capitalise the input to work with derived Read instances.
My parser is defined like this:
newtype Parser a = Parser (String -> [(a, String)])
So I CAN give it a Verb type
but this is where I run into a problem....
I've written a Parser called keyword
keyword :: Parser Verb keyword = do x <- many1 letter
case reads x of [(verb,"")] -> return verb _ -> fail "No verb" fails gracefully (assuming your Monad instance for Parser has fail _ = Parser (\_ -> []) ).
return (read x)
(read this as "take-at-least-one-alphabetic-letter-and-convert-to-a-Verb-type")
which DOES work provided that the user types in one of my Verbs. If they don't, well, the whole thing fails with an Exception and halts processing, returning to GHCi prompt.
Question: Am I going about this the right way? I want to put together lots of "data" types like Verb and Noun etc so that I can build a kind of "BNF grammar".
Question: If I am going about this the right way then what do I about the "read x" bit failing when the user stops typing in a recognised keyword. I could catch the exception, but typing an incorrect sentence is just a typo, not really appropriate for an exception, I shouldn't think. If it IS appropriate to do this in Haskell, then how do I catch this exception and continue processing.
You could try guessing what the user meant (cf. Levenshtein distance) for added comfort. Or you could ask for corrected input immediately when parsing an input fails. With the graceful failing of the parse as above, that doesn't need exceptions. If you think catching exceptions might be preferable after all, take a look at Control.Exception.
I thought that exceptions should be for exceptional circumstances, and it would seem that I might be misusing them in this context.
Thanks
Mark Spezzano