July 2004 - Libraries

RE: copyFile function in Haddock, VSHaskell & Cabal
by Bayley, Alistair 29 Jul '04

29 Jul '04

It might be worth lifting copyContents too; I've written this and I've seen it in hws-wp. I called it pipe, and in hws-wp it's called squirt. Perhaps something like this (untested!), in System.IO: hPipe :: Handle -> Handle -> Int -> IO () hPipe hFrom hTo bufferSize = do allocaBytes bufferSize bufferedPipe where bufferedPipe buffer = do count <- hGetBuf hFrom buffer bufferSize when (count > 0) $ do hPutBuf hTo buffer count bufferedPipe buffer > -----Original Message----- > From: Krasimir Angelov [mailto:ka2_mail@yahoo.com] > Sent: 29 July 2004 08:19 > To: libraries(a)haskell.org > Subject: copyFile function in Haddock, VSHaskell & Cabal > > I have found a bug in CVS Haddock. It uses the following > definition of copyFile function: > > copyFile f = do > s <- readFile (libdir ++ pathSeparator:f) > writeFile (odir ++ pathSeparator:f) s > > (See HaddockHtml.hs) > > This works well for text files but the function fails when it > is used to copy binary files under Windows. > Similar functions are used also in VSHaskell and Cabal. I > propose to add copyFile function to System.Directory which > will take care of binary files. > The function signature is simple: > > copyFile :: FilePath -> FilePath -> IO () > > In the attached file there are two implementations. > The first is a more portable but slower and the second is GHC > specific but is faster. > > Any suggestions > Krasimir ----------------------------------------- ***************************************************************** Confidentiality Note: The information contained in this message, and any attachments, may contain confidential and/or privileged material. It is intended solely for the person(s) or entity to which it is addressed. Any review, retransmission, dissemination, or taking of any action in reliance upon this information by persons or entities other than the intended recipient(s) is prohibited. If you received this in error, please contact the sender and delete the material from any computer. *****************************************************************

1 0

copyFile function in Haddock, VSHaskell & Cabal
by Krasimir Angelov 29 Jul '04

29 Jul '04

I have found a bug in CVS Haddock. It uses the following definition of copyFile function: copyFile f = do s <- readFile (libdir ++ pathSeparator:f) writeFile (odir ++ pathSeparator:f) s (See HaddockHtml.hs) This works well for text files but the function fails when it is used to copy binary files under Windows. Similar functions are used also in VSHaskell and Cabal. I propose to add copyFile function to System.Directory which will take care of binary files. The function signature is simple: copyFile :: FilePath -> FilePath -> IO () In the attached file there are two implementations. The first is a more portable but slower and the second is GHC specific but is faster. Any suggestions Krasimir __________________________________ Do you Yahoo!? New and Improved Yahoo! Mail - 100MB free storage! http://promotions.yahoo.com/new_mail

1 0

RE: Prelude function suggestions
by Simon Marlow 28 Jul '04

28 Jul '04

On 27 July 2004 23:44, ariep(a)xs4all.nl wrote: >> {- | Split the list at the occurrences of a separator into sub-list. >> This is a generalization of 'words'. -} >> chop :: (a -> Bool) -> [a] -> [[a]] >> chop p s = >> let (l, s') = break p s >> in l : case s' of >> [] -> [] >> (_:rest) -> chop p rest > > I like 'chop'. It belongs in Data.List, I'd say. We have something very like this inside GHC, except we call it 'split'. It's very useful for splitting up file paths into components. For comparison, Perl calls it 'split'. In Perl, 'chop' does something different. Python also has a 'split' function that does something similar. Does anyone object to calling it 'split'? Cheers, Simon

8 12

RE: Combinators for ReadP
by Koen Claessen 28 Jul '04

28 Jul '04

Simon Peyton-Jones said: > Koen, not me! I know that one can get away with the local quantification by adding an extra type parameter to the type constructor: newtype ReadP r a = ... But that wouldn't do much good, would it? Regards, /Koen

2 2

RE: Prelude function suggestions
by Graham Klyne 28 Jul '04

28 Jul '04

At 10:16 28/07/04 +0100, Simon Marlow wrote: >Does anyone object to calling it 'split'? Not I. 'split' is a name I have used on occasion for my own definition. #g ------------ Graham Klyne For email: http://www.ninebynine.org/#Contact

2 1

Prelude function suggestions
by Henning Thielemann 28 Jul '04

28 Jul '04

There are some functions that I regularly use which are so general that I wonder if they should be in the Prelude. Maybe I have overlooked them or some generalizations of them. Here are the ones that I most frequently use: {- | Compositional power of a function, i.e. apply the function n times to a value. -} nest :: Int -> (a -> a) -> a -> a nest 0 f x = x nest n f x = f (nest (n-1) f x) {- | Split the list at the occurrences of a separator into sub-list. This is a generalization of 'words'. -} chop :: (a -> Bool) -> [a] -> [[a]] chop p s = let (l, s') = break p s in l : case s' of [] -> [] (_:rest) -> chop p rest {- | Returns 'Just' if the precondition is fulfilled. -} toMaybe :: Bool -> a -> Maybe a toMaybe False _ = Nothing toMaybe True x = Just x

4 3

RE: Combinators for ReadP
by Simon Peyton-Jones 28 Jul '04

28 Jul '04

| >| Maybe I should write a test suite too, while I'm at it. Is there any | >| specific location in the cvs tree where test suites should be put? | > | >Yes please! The right place is in | > testsuite/tests/ghc-regress/lib/ | >you can make a ReadP sub-directory. | | Er, doesn't that make it more difficult to create a library+test suite that | can be shipped with other compilers? | | E.g., for my XML parser work, I created a separate test suite subdirectory | in the source tree for that library. It may not be perfect, but that's the way the CVS source tree is currently arranged. Of course, a separate package can adopt whatever conventions it likes, but Martin was asking about the CVS source tree that contains the libs that come with GHC/Hugs etc. S

1 0

RE: Combinators for ReadP
by Simon Marlow 27 Jul '04

27 Jul '04

On 27 July 2004 15:16, Martin Sjögren wrote: > On Tue, 27 Jul 2004 14:31:21 +0100, Simon Peyton-Jones > <simonpj(a)microsoft.com> wrote: >> You could do this, I believe, but it'd make ReadP less covenient for >> most users, because they'd have to add these redundant (and >> inexplicable) 'r' parameters. By "inexplicable" I mean that given a >> value p::ReadP r a we can say that p is a parser that parses >> some input, returning a value of type 'a'... but what's this 'r' >> thing? >> >> Wrapping up a polymorphic fn in a newtype is very standard, even >> though it is not H98. >> >> I'd be pretty reluctant to make this change > > Oh I agree. Having to pass around the extra parameter sucks. > Unfortunately, nhc98 doesn't support forall. :) Incedentally, universal quantification in constructor fields is one of the easier extensions to implement. Much easier than multi-parameter type classes, for example. You don't have to go the whole hog and implement rank-N polymorphism ala GHC in order to get most of the benefit. Cheers, Simon

2 1

RE: Combinators for ReadP
by Simon Peyton-Jones 27 Jul '04

27 Jul '04

You could do this, I believe, but it'd make ReadP less covenient for most users, because they'd have to add these redundant (and inexplicable) 'r' parameters. By "inexplicable" I mean that given a value p::ReadP r a we can say that p is a parser that parses some input, returning a value of type 'a'... but what's this 'r' thing? Wrapping up a polymorphic fn in a newtype is very standard, even though it is not H98. I'd be pretty reluctant to make this change Simon | -----Original Message----- | From: libraries-bounces(a)haskell.org [mailto:libraries-bounces@haskell.org] On Behalf Of Martin | Sjögren | Sent: 26 July 2004 12:18 | To: libraries(a)haskell.org | Subject: Re: Combinators for ReadP | | On Mon, 26 Jul 2004 11:00:31 +0100, Malcolm Wallace | <malcolm.wallace(a)cs.york.ac.uk> wrote: | > > While we're discussing changes, is there an overwhelmingly good reason | > > for the use of local quantification beyond making the type signatures | > > simpler? Writing | > > newtype ReadP r a = R ((a -> P r) -> P r) | > > (like in ContT) would make ReadP completely Haskell98 as far as I can tell. | > | > I think you'll find that without the local universal quantifier, | > you cannot make ReadP a (useful) instance of Monad, because your | > 'r' parameter would be fixed across >>=, whereas it really needs to | > be variable. | | Uhh. But it works for ContT, so why not for ReadP? The only point time | you have to fix the r is when you want to "run" the parser, all other | parsers will be polymorphic in r: | | parseFoo :: ReadP r Foo | parseBar :: ReadP r Bar | | readP_to_S :: ReadP a a -> ReadS a | | I do exactly this in the package description parsing code of Cabal | (using a locally hacked ReadP, Cabal really must be H98...), and it | works just fine. | | | /Martin | _______________________________________________ | Libraries mailing list | Libraries(a)haskell.org | http://www.haskell.org/mailman/listinfo/libraries

2 1

RE: Haskell Extensions in various compilers (Cabal)
by Simon Marlow 26 Jul '04

26 Jul '04

On 13 July 2004 04:15, Isaac Jones wrote: > - Do any of Hugs, GHC, and NHC disallow Long.Hierarchical.Module.Names > by default? Can they all turn them off? GHC includes them by default, and they cannot be turned off. This is a bug (but we're unlikely to fix it). My only comment is that some of the extensions you've formalised are probably a bit too experimental or GHC-specific to be included in the general Extension type. I'd leave out InlinePhase for one. PatternTypeAnnotations should probably be called ScopedTypeVariables. I'm not sure what UnsafeOverlappingInstances does compared with AllowOverlappingInstances and the other instance-related flags. Cheers, Simon

4 6