- Glasgow-haskell-users

RE: [Haskell] How to define tail function for Even/Odd GADT lists?
by Simon Peyton-Jones 24 Apr '08

24 Apr '08

[Redirecting to ghc-users] You're right that tailList "ought" to work. There are at least two reasons that it doesn't. First, GHC does not have a robust implementation of GADTs + functional dependencies. The interaction is very tricky. So I tried re-expressing it using type families, thus: data Even; data Odd type family Flip a :: * type instance Flip Even = Odd type instance Flip Odd = Even data List a b where Nil :: List a Even Cons :: a -> List a b -> List a (Flip b) tailList :: List a b -> List a (Flip b) tailList (Cons _ xs) = xs (I find the program quite a bit easier to read in this form.) Now I get the error (from the HEAD) Occurs check: cannot construct the infinite type: b = Flip (Flip b) In the pattern: Cons _ xs In the definition of `tailList': tailList (Cons _ xs) = xs There's a bug here -- reporting an occurs check is premature. But there is also a real problem: if you look at the type constraints arising from tailList you'll see that you get c ~ Flip b b ~ Flip c where c is the existential you get from the GADT match. Now, *we* know that b = Flip (Flip b) is always tru, but GHC doesn't. After all, you could add new type instances type instance Flip Int = Bool type instance Flip Bool = Char and then the equation wouldn't hold. The best we can hope for is to get tailList :: (b ~ Flip (Flip b)) => List a b -> List a (Flip b) although this too is rejected at the moment because of the above bug. What we really want is a *closed* type family, like this type family Flip a where Flip Even = Odd Flip Odd = Even (preferably supported by kinds) and even *then* it's not clear whether it's reasonable to expect the compiler to solve the above problem by trying all possibilities. This relates to http://hackage.haskell.org/trac/ghc/ticket/2101 That is probably more than you wanted to know. Since there's a bug here, I'll make a ticket. Simon | -----Original Message----- | From: haskell-bounces(a)haskell.org [mailto:haskell-bounces@haskell.org] On Behalf Of Ahn, Ki Yung | Sent: 23 April 2008 21:33 | To: Haskell(a)haskell.org | Subject: [Haskell] How to define tail function for Even/Odd GADT lists? | | Using GADTs and functional dependencies, we can define GADT lists | that statically distinguishes even length lists from odd length lists. | | > data Even | > data Odd | > | > class Flip b c | b -> c, c -> b where | > | > instance Flip Even Odd where | > instance Flip Odd Even where | > | > data List a b where | > Nil :: List a Even | > Cons :: Flip b c => a -> List a b -> List a c | | For example, | | Nil :: forall a. List a Even | Cons True Nil :: List Bool Odd | Cons False (Cons True Nil) :: List Bool Even | | | We were able to define the function that returns the head. | | > headList :: List a b -> a | > headList (Cons x _) = x | | However, we were not able to write a function that returns the tail. | | > tailList :: Flip b c => List a b -> List a c | > tailList (Cons _ xs) = xs | | Is there a way to define the tailList function within the current | GHC type system implementation (maybe using some other language | extensions), or do we need to improve the type system regarding | GADTs and functional dependencies? | | $ ghci -fglasgow-exts EOlist.hs | GHCi, version 6.8.2: http://www.haskell.org/ghc/ :? for help | Loading package base ... linking ... done. | [1 of 1] Compiling Main ( EOlist.lhs, interpreted ) | | EOlist.lhs:30:25: | Couldn't match expected type `c' against inferred type `b1' | `c' is a rigid type variable bound by | the type signature for `tailList' at EOlist.lhs:29:21 | `b1' is a rigid type variable bound by | the constructor `Cons' at EOlist.lhs:30:12 | Expected type: List a c | Inferred type: List a b1 | In the expression: xs | In the definition of `tailList': tailList (Cons _ xs) = xs | Failed, modules loaded: none. | Prelude> | | Note: You can save this message content as EOList.lhs and load the | script with ghci to observe the type error message above. | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org | http://www.haskell.org/mailman/listinfo/haskell

1 0

runstmt within API
by C.M.Brown 23 Apr '08

23 Apr '08

Hello, I want to use the ghc evaluator on the fly for some refactorings within HaRe. However, I notice that runstmt ::Session -> String -> IO RunResult. Is there anyway I can grab the result of the evaluated expression as a String (rather than it being outputted to the terminal)? It seems RunResult is defined: data RunResult = RunOk [Name] -- names bound by the expression | RunFailed | RunException GHC.IOBase.Exception If I can't grab its result, is there a way to query the [Name] to get a list of bindings in String format? Preferably something like "it = expression". Kind regards, Chris.

2 2

help with an intriguing problem
by Liang Guang 18 Apr '08

18 Apr '08

Hi! the ghc compiler keeps complaing " can't find module 'Char', perhaps you haven't installed the profiling libraries for package haskell98?". but actually I did install it, and i checked with ghc -v , it is right there: "wired-in package haskell98 mapped to haskell98-1.0.1.0", and Char.hi is right there in the folder haskell98-1.0.1.0.. can't figure out why it happened.. anyone please ? Thanks! Liang --------------------------------- Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now.

3 4

6.8 unable to derive with higher-kinded variable (6.6 could)
by Mike Gunter 18 Apr '08

18 Apr '08

GHC 6.8 seems unable to derive (some?) instances for data types with higher-kinded variables. GHC 6.6 manages these just fine. See below. What's the story? thanks -m :~/haskell$ cat D.hs data T w = T (w Bool) deriving (Show) data ID x = ID x deriving (Show) main = print (T (ID False)) :~/haskell$ ghci -ignore-dot-files -fallow-undecidable-instances D.hs ___ ___ _ / _ \ /\ /\/ __(_) / /_\// /_/ / / | | GHC Interactive, version 6.6.1, for Haskell 98. / /_\\/ __ / /___| | http://www.haskell.org/ghc/ \____/\/ /_/\____/|_| Type :? for help. Loading package base ... linking ... done. [1 of 1] Compiling Main ( D.hs, interpreted ) Ok, modules loaded: Main. *Main> main T (ID False) *Main> :q Leaving GHCi. :~/haskell$ ${GHC_682_BIN}/ghci -ignore-dot-files -fallow-undecidable-instances D.hs GHCi, version 6.8.2: http://www.haskell.org/ghc/ :? for help Loading package base ... linking ... done. [1 of 1] Compiling Main ( D.hs, interpreted ) D.hs:1:0: No instance for (Show (w Bool)) arising from the 'deriving' clause of a data type declaration at D.hs:1:0-48 Possible fix: add an instance declaration for (Show (w Bool)) When deriving the instance for (Show (T w)) Failed, modules loaded: none. Prelude>

2 1

gmp
by Christian Maeder 17 Apr '08

17 Apr '08

I understand that gmp is needed for the certain libraries like the Prelude with Double and Integer. But I do not understand why gmp is so deeply buried in the rts. Are the basic types Int and Pointer not enough to write a compiler like ghc? Cheers Christian

7 10

Using GHC API to generate STG
by Mark Wassell 15 Apr '08

15 Apr '08

Hello, I am looking at how GHC generates STG and I am finding with a very simple piece of Haskell my usage of the API is not generating as much STG as I see when using -dump-stg option. In particular it isn't generating a binding for the nullary constructor LNull (I hope that's the correct terminology) The code is module Ex2 where data List a = LCon a (List a) | LNull -- deriving Show main :: List Int main = LCon 1 LNull Using the API I get [sat_s1pdQ = NO_CCS GHC.Base.I#! [1]; Ex2.main = NO_CCS Ex2.LCon! [sat_s1pdQ Ex2.LNull]; main = \u srt:SRT:[(s1pdS, Ex2.main)] [] Ex2.main;] with -ddump-stg I get a_r5Y = NO_CCS GHC.Base.I#! [1]; SRT(a_r5Y): [] Ex2.main = NO_CCS Ex2.LCon! [a_r5Y Ex2.LNull]; SRT(Ex2.main): [] Ex2.LCon = \r [eta_s67 eta_s68] Ex2.LCon [eta_s67 eta_s68]; SRT(Ex2.LCon): [] Ex2.LNull = NO_CCS Ex2.LNull! []; <---- This is missing SRT(Ex2.LNull): [] (in particular the Ex2.LNull is missing from the API STG) My thrown together API code is session <- GHC.newSession $ Just path (dflags,_) <- GHC.getSessionDynFlags session >>= Packages.initPackages GHC.setSessionDynFlags session dflags {GHC.hscTarget=GHC.HscAsm} core <- GHC.compileToCore session fp case core of Just core' -> do core'' <- corePrepPgm dflags core' [] stg <- coreToStg (PackageConfig.stringToPackageId "Ex2") core'' putStrLn $ (show $ (ppr stg) defaultDumpStyle) Nothing -> return () Mark

2 1

GHC API (parsing)
by Johannes Waldmann 15 Apr '08

15 Apr '08

Dear all, how can I convince the Language.Haskell.Parser to accept "GHC Haskell" (i.e., -fglasgow-exts, e.g. for existential types) or: how can I convince the GHC API loader to parse a module from a string (not from a file)? Any hints appreciated, Johannes.

5 9

Sending binary files through a socket
by Ivan Salazar 15 Apr '08

15 Apr '08

Hello, I'm having a lot of trouble sending binary files through a socket (I'm trying to program a very minimalistic webserver) but I just can't find a way to read and send binary files, even though, plain text and html documents are sent correctly. My code looks something like this: sendResource handle name = do exist <- doesFileExist name isFile <- liftM not (doesDirectoryExist name) if exist && isFile then do resource <- Data.ByteString.readFile name hPutStr handle (header ++ fileType ++ "\n") hSetBinaryMode handle True hPutStr handle resource else do hPutStr handle page404 where fileType = maybe ("text/plain") (flip const []) (lookup (takeExtension name) extensions) Thanks for your help in advance. PS: Yes, I'm terribly newbie to Haskell.

1 0

Re: GHC API (parsing)
by Johannes Waldmann 14 Apr '08

14 Apr '08

> http://hackage.haskell.org/cgi-bin/hackage-scripts/package/haskell-src-exts… interesting, I'll look into that. * it's not exactly a drop-in replacement for Language.Haskell.* ? (HsNewTypeDecl is different?) * for the others, number of constructor arguments does not match, e.g. `HsConDecl' should have 2 arguments, but has been given 3 * how is your parser tied to ghc? I.e. what happens if the next ghc release introduces new syntax? general question * is there any parser that keeps comments (i.e. makes AST nodes from them)? what parser does Haddock use? Best regards, J.W.

4 4

Re: [Haskell-cafe] more on my ghc package issues
by Duncan Coutts 13 Apr '08

13 Apr '08

On Sat, 2008-04-12 at 23:33 -0500, Galchin, Vasili wrote: > 2) This strongly looks like the package database manager, i.e. > ghc-pkg. Source please so I can understand where this d*amn "unknown > package: unix-2.3.0.0" message is coming from and why? Yes, ghc/ghc-pkg is giving this message because you're trying to use a package that depends on a package that you unregistered. In your case, you unregistered unix-2.3.0.0 but process-1.0.0.0 depended on that. Cabal depends on process-1.0.0.0 which means you now cannot runghc Setup.hs scripts because they all import Cabal. Look at the output of ghc-pkg list. It marks in {} the packages that are broken due to missing dependencies. Remember you also have two package databases, the per-user one and the global one. The per-user one is stored under ~/.ghc You can see the exact path in the ghc-pkg list output. The packages from the user db mask those from the global db, so if you have messed up the per-user one then you will still get problems even if you reset the global one to default. If you're just trying to get back to a known state then you can delete the per-user ghc package db. ghc-pkg will re-create it next time you register a package as a user. > 3) Serious rant mode on ... I am a computer industry veteran(decades) > who is trying to convince, coggle, etc. younger colleagues about the > usefulness of FPLs for correctness, etc, and then => "unknown > package ....". The error messages from ghc-pkg are not very helpful and it does not prevent you from removing packages which other packages still depend on. The best advice at the moment is, don't un-register packages, at least not without considering what other packages still depend on them. There's no easy way of doing a reverse dependency lookup I think yet via ghc-pkg. You can see direct dependencies, eg $ ghc-pkg field containers depends depends: base-3.0.1.0 array-0.1.0.0 As it happens, the latest version of Cabal gives a much better error message in the case that you try to configure a package that depends on a broken package. It says exactly which package is broken due to which missing dependency. In this case we'd get a message like: The following installed packages are broken because other packages they depend on are missing. These broken packages must be rebuilt before they can be used. package process-1.0.0.0 is broken due to missing package unix-2.3.0.0 It would be nice if ghc/ghci/runghc could give a similarly helpful error message. Duncan

1 0