March 2004 - Glasgow-haskell-users

Re: Polymorphic lists...
by MR K P SCHUPKE 09 Mar '04

09 Mar '04

I have written a first attempt at a fold function for the heterogenious list: class RFold i r where rFold :: (forall a . a -> i -> i) -> i -> r -> i instance RFold i RNil where rFold f i RNil = i instance RFold i r => RFold i (a `RCons` r) where rFold f i (x `RCons` xs) = f x (rFold f i xs) This works providing the folded 'op' has the type: forall a . a -> i -> i which means it does not work for functions like show :: forall a . Show a => a -> i -> i as the types are different. I have not figured out a way to make it accept a constraint like Show for example. Here is an example: length = rFold (\_ -> (+1)) 0 relation The use of such a function seems limited, if constraints like Show cannot be used, as most useful applications of fold would require some kind of class membership for example: string = rFold shows "" relation This fails to compile because shows has type: shows :: forall a . Show a => a -> i -> i but fold expects op :: forall a . a -> i -> i Regards, Keean Schupke.

1 0

Re: Polymorphic lists...
by MR K P SCHUPKE 09 Mar '04

09 Mar '04

I did not know about Oleg's posting, as I originally said, I based my implementation on a paper by Conor McBride. Oleg is addressing the question of type safe casting, rather than generic storage, so his code is a bit different. Infact his class: > class TypeSeq t s where > type_index:: t -> s -> Int > fetch:: t -> s -> t > alter:: t -> s -> s > > instance (PList Cons t r) => TypeSeq t (Cons t r) where > type_index _ _ = 0 > fetch _ (Cons v _) = v > alter newv (Cons v r) = Cons newv r > > instance (PList Cons t' r', TypeSeq t r') => TypeSeq t (Cons t' r') where > type_index v s = 1 + (type_index v $ cdr s) > fetch v s = fetch v $ cdr s > alter newv (Cons v' r') = Cons v' $ alter newv r' This stores unique types in a list that can be indexed by their types. Actually last night (before I read this code) I came up with something similar: data MNil = MNil deriving (Show,Data,Typeable) data MCons l a r = MCons l a r deriving (Show,Data,Typeable) class MLookup l r a | l r -> a where mLookup :: r -> l -> a instance MLookup l (MCons l a r) a where mLookup (MCons _ x _) _ = x instance MLookup l r b => MLookup l (MCons m a r) b where mLookup (MCons _ _ xs) l = mLookup xs l This is indexed by a unique type, but stores a second independant type. The allows a kind of static finite map, which is pretty cool! Here's an example: data TmId = TmId data TmVal = TmVal data TmFloat = TmFloat data TmName = TmName testMap :: MCons TmId Int (MCons TmVal String (MCons TmFloat Float (MCons TmName String MNil))) testMap = MCons TmId 1 $ MCons TmVal "Hello" $ MCons TmFloat 1.2 $ MCons TmName "World" MNil main :: IO () main = do putStrLn $ show $ testMap `mLookup` TmId putStrLn $ show $ testMap `mLookup` TmVal putStrLn $ show $ testMap `mLookup` TmFloat putStrLn $ show $ testMap `mLookup` TmName Index types don't need to be unique, the first match from the head of the list will be returned. No match will result in a compile time error. Regards, Keean Schupke.

1 0

Polymorphic lists...
by MR K P SCHUPKE 08 Mar '04

08 Mar '04

I needed a list which could handle items of different types for the database code I am writing. I have written a module implementing such a list based on dependant types (from Conor McBride: Faking It; Simulating Depandant Types in Haskell). Although McBride does not mention lists/vectors with items of differing types, the solution to implementing them came from his 'nthFront' function for re-arranging the order of arguments to a function. Any type can be inserted into the list, which supports head/tail/init/last, as well as indexed lookup, and a cartesian-product (concatenating two lists together). I have included fromTuple/toTuple as well. This seems quite a useful construct, and if there is nothing similar in the standard libraries at the moment, do you think this is worth including? Regards, Keean Schupke.

2 1

RE: -static
by Simon Marlow 08 Mar '04

08 Mar '04

> In ghc/compiler/main/DriverFlags.hs machdepCCOpts includes a "-static" > flag for some arches. Is this really necessary? I can't see > any comments > as to why, nor any real answers from a quick google. I can't remember why either. > It causes this when compiling darcs on these arches: > > /usr/lib/ghc-6.2/libHSunix.a(User.o)(.text+0x2a2c): In > function `s6T9_ret': > : warning: Using 'getgrgid_r' in statically linked > applications requires at runtime the shared libraries from > the glibc version used for linking > > which I assume would lead to too strict dependencies on glibc being > necessary. What platform? Does everything work if you remove the -static? Cheers, Simon

2 2

RE: #! for GHC?
by Simon Marlow 08 Mar '04

08 Mar '04

> On Sat, Mar 06 2004, Dean Herington wrote: > > Can GHC be invoked somehow via the #! mechanism? Put > another way, is > > there a GHC analogue to runhugs? > > Well, ghc now has an -e flag to evaluate stuff directly from > the prompt > (e.g. ghc -e Main.main Main.hs). But you can't execute shell-scripts > from a shell-script with #! so you can't just put it in your Haskell > module. Neither can you use the binary directly since you can > only give > one argument to programs executed with #! and you need both -e and -B. Aha! I knew there was something else. However, I just tried it on a recent Linux (RedHat 9) and it seems to accept nested #! scripts. There's a rather annoying limit on the length of the #! line, too. Cheers, Simon

2 1

RE: #! for GHC?
by Simon Marlow 08 Mar '04

08 Mar '04

> Can GHC be invoked somehow via the #! mechanism? Put another way, is > there a GHC analogue to runhugs? Almost, but not quite. Since 6.2, GHC has the -e switch for invoking expressions from the command line, which gets a step closer. GHC doesn't ignore the '#!' line yet, but I can add that. Then, as far as I'm aware, it should work, albeit a little slowly because it goes through the ordinary GHCi startup which loads some packages etc. Cheers, Simon

1 0

#! for GHC?
by Dean Herington 08 Mar '04

08 Mar '04

Can GHC be invoked somehow via the #! mechanism? Put another way, is there a GHC analogue to runhugs?

2 1

RE: [Haskell] performance tuning Data.FiniteMap
by MR K P SCHUPKE 07 Mar '04

07 Mar '04

I was thinking about improving array performance, and was wondering if a transactional model would work well. You would keep a base copy of the array, and any writes would be written to a delta style transaction list. A reference to the array would be the list plus the base array. Different references to the same array would just reference different points in the delta list. The garbage colector would eat the delta list from the tail, merging writes into the base array once references to that part of the list are discarded. Writes would be very fast - just the time to add a delta to the transaction list. Reads would slow down as the transaction list grows, but the list would only be as long as the oldest reference, so providing references to very old copies of the array are not required, the transaction list would remain short. It would be even more efficient if the 'liveness' of references could be checked when writing the array - at the cost of a slight slowdown in write performance. I would be interested in any comments... I suspect somebody has done this before, but I havent looked for any papers yet. Regards Keean Schupke.

4 4

-static
by Ian Lynagh 07 Mar '04

07 Mar '04

Hi, In ghc/compiler/main/DriverFlags.hs machdepCCOpts includes a "-static" flag for some arches. Is this really necessary? I can't see any comments as to why, nor any real answers from a quick google. It causes this when compiling darcs on these arches: /usr/lib/ghc-6.2/libHSunix.a(User.o)(.text+0x2a2c): In function `s6T9_ret': : warning: Using 'getgrgid_r' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking which I assume would lead to too strict dependencies on glibc being necessary. Thanks Ian

1 0

Overlapping instances and multi-parameter classes
by Mike Gunter 06 Mar '04

06 Mar '04

Given the following: > > import Data.Maybe ( fromMaybe ) > > data None > none = error "none evaluated" :: None these class and instance declarations: > class ToMaybe2 b a where toMaybe2 :: a -> Maybe b > instance ToMaybe2 a None where toMaybe2 = const Nothing > instance ToMaybe2 a a where toMaybe2 = Just compile fine (given -fallow-undecidable-instances and -fallow-overlapping-instances). And the following code: > mbPutStrLn :: ToMaybe2 String a => a -> IO () > mbPutStrLn = putStrLn . fromMaybe "None." . toMaybe2 > main = mbPutStrLn none >> mbPutStrLn "Hello, world!" works as I'd expect. However, if the parameters to the class are reversed: > {- uncomment to see error message > class ToMaybe1 a b where toMaybe1 :: a -> Maybe b > instance ToMaybe1 None a where toMaybe1 = const Nothing > instance ToMaybe1 a a where toMaybe1 = Just > -} I get the following error message (from GHC{,i} 6.2): Overlapping instance declarations: OverlapOrder.lhs:27: ToMaybe1 None a OverlapOrder.lhs:28: ToMaybe1 a a Why is this? thanks, mike

1 0