- Glasgow-haskell-users

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

RE: Release Candidate for 6.2.1 available
by Simon Marlow 06 Mar '04

06 Mar '04

> Also: I'm fiddling around with a new design for GHC's web > pages. Please > browse on over and let me know what you think, especially if > it doesn't work on your browser: > > http://www.haskell.org/ghc/staging/ For those who had layout problems with the new pages, please try again. It now works fine for me on IE, Mozilla and Konqueror. (except that there's a slight off-by-one-pixel problem in the header sometimes; I'll fix that at some point). Cheers, Simon

1 0

RE: Misleading error message
by Simon Peyton-Jones 06 Mar '04

06 Mar '04

Good story. The error message is indeed misleading. We've clarified the documentation in the HEAD, and vastly improved the error message. We'll merge the fix into 6.2 if it's not too hard to do so. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of Josef Svenningsson | Sent: 05 March 2004 11:36 | To: glasgow-haskell-users(a)haskell.org | Subject: Misleading error message | | Hi all, | | This mail is a little story about a little guy called Josef and a little | adventure he had this week together with the compiler ghc. Although | exciting, this adventure took him several hours and Josef would have been | happier without it. So he appeals to the implementors to improve on the | situation so that others may not have to embark on the same journey. | | OK, here's the story. | | Josef has a rather large program which he is writing. It contains many | haskell modules spread out over several directories. Early this week Josef | decided to add yet another module. He decided to call this module | "InterfaceParse". So he created a file "InterfaceParse.lhs", filled it | with nice haskell code and imported it in another haskell file. Now, after | a while Josef realised that the name "InterfaceParse" was not in par with | his naming convention and decided to rename the module to | "InterfaceParser". Now, here is where the real adventure started. Josef | changed the name of the file to "InterfaceParser.lhs" but forgot to change | the module declaration in the file. In the file one could still read the | text: | "module InterfaceParse ..." | | Unknowing of his mistake, Josef carried on and changed the import to | importing the module "InterfaceParser" instead of "InterfaceParse". And to | make sure everything had gone right he recompiled his project. And behold! | The compiler complained. What did it say? The compiler said it could not | find the interface for the module "InterfaceParse"! Josef was very puzzled | by this. Was his import statement wrong? No. Then why was the compiler | looking for the interface "InterfaceParse" when it should look for | "InterfaceParser"? He started to experiment with importing other modules, | modules which didn't exist. All the time, whenever he was trying to import | a non-existing module the compiler would correctly complain that the | interface was missing. But as soon as he tried to import "InterfaceParser" | the compiler would complain that it couldn't find "InterfaceParse". | | There where times when Josef thought someone was playing him a practical | joke. Maybe Simon Peyton Jones had come up with a devious plan on how to | make money from his compiler? By adding some obscure bug which would | trigger very rarely he could then ask people for money to fix the bug. | | After much desperation and consulation with others Josef finally | discovered his error. | | One might say that Josef should have understood his error right away, that | he should have checked his files better. But Josef believes that the error | message he got from the compiler was very misleading and led him away from | finding the true source of the error. Dear implementors, please improve | the error reporting in this case. | | | /Josef (first time he has ever written about himself in third person) | | By the way, Josef is using ghc-6.2 on a solaris machine. | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

1 0

Release Candidate for 6.2.1 available
by Simon Marlow 05 Mar '04

05 Mar '04

The 6.2.1 release is imminent, and we'd like to get a bit of final testing before the actual release. So for this release and in the future, at the request of some of our users (you know who you are ;-) we're doing release candidates. The idea is that the nightly builds on the STABLE branch will produce snapshots in the same way that the HEAD builds do. They will be named 6.2.<date>, eg. 6.2.20040304 for last night's build. When the tree is frozen, we declare one of the snapshots to be a release candidate. So: ghc-6.2.20040304 and later are release candidates for 6.2.1 Get them from here: http://www.haskell.org/ghc/dist/stable/dist/ Also: I'm fiddling around with a new design for GHC's web pages. Please browse on over and let me know what you think, especially if it doesn't work on your browser: http://www.haskell.org/ghc/staging/ Cheers, Simon

3 2

Parsec State Monad Question
by John Knottenbelt 05 Mar '04

05 Mar '04

Hi I was wondering if it was possible to write a function that allows changing the state type of the Parser monad from Parsec. The full parser type is: GenParser p s a, where p is the token type, s is the state type and a is the return type of the monad. However, I mostly use CharParser s a which is a synonym for GenParser Char s a. What I would like is a function that allows me to combine CharParsers with differing state types. Something like: runParser' :: CharParser t a -> t -> CharParser s (t,a) So that I can use parsers written to work with different state types together on the same input stream. Does anybody know if this is possible, or how one might do it with Parsec? Thanks John

2 1

Misleading error message
by Josef Svenningsson 05 Mar '04

05 Mar '04

Hi all, This mail is a little story about a little guy called Josef and a little adventure he had this week together with the compiler ghc. Although exciting, this adventure took him several hours and Josef would have been happier without it. So he appeals to the implementors to improve on the situation so that others may not have to embark on the same journey. OK, here's the story. Josef has a rather large program which he is writing. It contains many haskell modules spread out over several directories. Early this week Josef decided to add yet another module. He decided to call this module "InterfaceParse". So he created a file "InterfaceParse.lhs", filled it with nice haskell code and imported it in another haskell file. Now, after a while Josef realised that the name "InterfaceParse" was not in par with his naming convention and decided to rename the module to "InterfaceParser". Now, here is where the real adventure started. Josef changed the name of the file to "InterfaceParser.lhs" but forgot to change the module declaration in the file. In the file one could still read the text: "module InterfaceParse ..." Unknowing of his mistake, Josef carried on and changed the import to importing the module "InterfaceParser" instead of "InterfaceParse". And to make sure everything had gone right he recompiled his project. And behold! The compiler complained. What did it say? The compiler said it could not find the interface for the module "InterfaceParse"! Josef was very puzzled by this. Was his import statement wrong? No. Then why was the compiler looking for the interface "InterfaceParse" when it should look for "InterfaceParser"? He started to experiment with importing other modules, modules which didn't exist. All the time, whenever he was trying to import a non-existing module the compiler would correctly complain that the interface was missing. But as soon as he tried to import "InterfaceParser" the compiler would complain that it couldn't find "InterfaceParse". There where times when Josef thought someone was playing him a practical joke. Maybe Simon Peyton Jones had come up with a devious plan on how to make money from his compiler? By adding some obscure bug which would trigger very rarely he could then ask people for money to fix the bug. After much desperation and consulation with others Josef finally discovered his error. One might say that Josef should have understood his error right away, that he should have checked his files better. But Josef believes that the error message he got from the compiler was very misleading and led him away from finding the true source of the error. Dear implementors, please improve the error reporting in this case. /Josef (first time he has ever written about himself in third person) By the way, Josef is using ghc-6.2 on a solaris machine.

1 0

config.h namespace clashes
by John Meacham 04 Mar '04

04 Mar '04

The inclusion of the GHC made config.h in HsFFI.h makes it extremely difficult to use autoconf with ones own projects without dealing with horrible name clashes everywhere. Is there some way this can be fixed? perhaps pre-generating the HsFFI.h. although I don't think it is stated explicitly by the FFI report, I would assume that HsFFI.h should not clash with anything that doesn't start with Hs, hs_ or HS_. also, it would be good if config.h could be renamed to something like ghc-config.h, because otherwise the conflict between the names can hide bugs where you get the incorrect config.h or make it tricky to get at the GHC one if it is actually the one you want. John -- --------------------------------------------------------------------------- John Meacham - California Institute of Technology, Alum. - john(a)foo.net ---------------------------------------------------------------------------

2 1

RE: Generics... no Tuples > 2 either...
by MR K P SCHUPKE 03 Mar '04

03 Mar '04

with reference to Ptr, I only need an instance of Data for the Ptr () case (ie opaque pointers) ... so for generics it only needs to know that a Ptr is a Ptr, and to treat it like a value. Regards, Keean Schupke.

2 1

RE: Network.Socket on win32
by Simon Marlow 03 Mar '04

03 Mar '04

> Yup that worked. Sort of an obscure error message > to communicate "package not found".... > > Is there a convention so I know which package I > need for which module? My guess is sometimes (but > randomly) you need to add a package > on the command line corresponding to the first > part of the qualified module name. e.g. I just > figured out that > > -package text worked to get me Text.ParserCombinators.Parsec > > OTOH, if this is the rule, it seems strange that > GHC can't go and find the packages itself. As far as it can, GHC does go and find the packages itself. There's one situation it can't: when doing batch linking from the command line, because then GHC doesn't know which modules were imported by the program. If you use GHCi or --make, then GHC figures out the package dependencies for itself. To answer your question, the package for each module is listed in the documentation: each module has the package name in the title at the top, and the contents page also lists the package name alongside each module. Incedentally, -package parsec is the right way to get Text.ParserCombinators.Parsec. Cheers, Simon

1 0

RE: Network.Socket on win32
by Simon Marlow 02 Mar '04

02 Mar '04

> This program does not compile using GHC 6.2 on winXP. > > import Network.Socket > main = do putStr "" > > The error is: > > conc.o(.text+0x22):fake: undefined reference to > `__stginit_NetworkziSocket_' > > What am I doing wrong? You need to add '-package network' to the command line when linking, or compile the program with --make. Cheers, Simon

2 1