July 2005 - Libraries

System.FilePath module
by Isaac Jones 08 Jul '05

08 Jul '05

Greetings. Back in October, there was a discussion on the System.FilePath module that's currently the hidden module Distribution.Compat.FilePath in Cabal: http://www.haskell.org/pipermail/libraries/2004-October/002591.html It was decided that we should not include it as System.FilePath because it was insufficiently consistent, had some confusing corner cases, and some folks felt that FilePath should be an abstract type. A number of people would like to use the FilePath module as it is, and someone has even packaged it up separately, so I think we should move forward in building a module that performs this task. For those who objected at the time, can someone put together a list of requirements for the FilePath module you believe we need? (Perhaps on the Wiki?) FWIW, the module we have is very similar to the SML and Python modules which perform this task, and most likely very similar to other languages' modules. In short, I think this approach works in practice, but not in theory, as Lemmih said on IRC ;) peace, isaac

3 2

docs for System.Posix.IO
by Isaac Jones 07 Jul '05

07 Jul '05

Greetings, I've been working with the System.Posix.IO module and ran into a few things that could use some documentation. I have added some, and it would be nice if folks could verify that I'm not making stuff up. I would actually be happy to paste some of the docs right from the 'read(2)' and 'write(2)' man pages if that's the correct source? I would commit this, but for some reason, 'make html' doesn't work for me no matter how many times I run cvs up, configure, and make. I always get this error: Control/Monad.hs: unknown flags in {-# OPTIONS #-} pragma: _GHC <<ghc: 9110492 bytes, 2 GCs, 72904/72904 avg/max bytes residency (1 samples), 12M in use, 0.00 INIT (0.00 elapsed), 0.01 MUT (0.09 elapsed), 0.00 GC (0.01 elapsed) :ghc>> make[1]: *** [Control/Monad.raw-hs] Error 1 Any advice? peace, isaac Index: System/Posix/IO.hsc =================================================================== RCS file: /home/cvs/root/fptools/libraries/unix/System/Posix/IO.hsc,v retrieving revision 1.15 diff -u -r1.15 IO.hsc --- System/Posix/IO.hsc 7 Feb 2005 12:03:44 -0000 1.15 +++ System/Posix/IO.hsc 6 Jul 2005 21:53:53 -0000 @@ -9,7 +9,10 @@ -- Stability : provisional -- Portability : non-portable (requires POSIX) -- --- POSIX IO support +-- POSIX IO support. These types and functions correspond to the unix +-- functions open(2), close(2), etc. For more portable functions +-- which are more like fopen(3) and friends from stdio.h, see +-- 'System.IO'. -- ----------------------------------------------------------------------------- @@ -82,10 +85,12 @@ -- ----------------------------------------------------------------------------- -- Pipes --- |The 'createPipe' function creates a pair of connected file descriptors. The first --- component is the fd to read from, the second is the write end. --- Although pipes may be bidirectional, this behaviour is not portable and --- programmers should use two separate pipes for this purpose. +-- |The 'createPipe' function creates a pair of connected file +-- descriptors. The first component is the fd to read from, the second +-- is the write end. Although pipes may be bidirectional, this +-- behaviour is not portable and programmers should use two separate +-- pipes for this purpose. May throw an exception if this is an +-- invalid descriptor. createPipe :: IO (Fd, Fd) createPipe = @@ -98,9 +103,11 @@ -- ----------------------------------------------------------------------------- -- Duplicating file descriptors +-- | May throw an exception if this is an invalid descriptor. dup :: Fd -> IO Fd dup (Fd fd) = do r <- throwErrnoIfMinus1 "dup" (c_dup fd); return (Fd r) +-- | May throw an exception if this is an invalid descriptor. dupTo :: Fd -> Fd -> IO Fd dupTo (Fd fd1) (Fd fd2) = do r <- throwErrnoIfMinus1 "dupTo" (c_dup2 fd1 fd2) @@ -116,15 +123,19 @@ data OpenMode = ReadOnly | WriteOnly | ReadWrite +-- |Correspond to some of the int flags from C's fcntl.h. data OpenFileFlags = OpenFileFlags { - append :: Bool, - exclusive :: Bool, - noctty :: Bool, - nonBlock :: Bool, - trunc :: Bool + append :: Bool, -- ^ O_APPEND + exclusive :: Bool, -- ^ O_EXCL + noctty :: Bool, -- ^ O_NOCTTY + nonBlock :: Bool, -- ^ O_NONBLOCK + trunc :: Bool -- ^ O_TRUNC } + +-- |Default values for the 'OpenFileFlags' type. False for each of +-- append, exclusive, noctty, nonBlock, and trunc. defaultFileFlags :: OpenFileFlags defaultFileFlags = OpenFileFlags { @@ -135,9 +146,12 @@ trunc = False } + +-- |Open and optionally create this file. See 'System.Posix.Files' +-- for information on how to use the 'FileMode' type. openFd :: FilePath -> OpenMode - -> Maybe FileMode -- Just x => O_CREAT, Nothing => must exist + -> Maybe FileMode -- ^Just x => creates the file with the given modes, Nothing => the file must exist. -> OpenFileFlags -> IO Fd openFd name how maybe_mode (OpenFileFlags append exclusive noctty @@ -164,10 +178,17 @@ WriteOnly -> (#const O_WRONLY) ReadWrite -> (#const O_RDWR) +-- |Create and open this file in WriteOnly mode. A special case of +-- 'openFd'. See 'System.Posix.Files' for information on how to use +-- the 'FileMode' type. + createFile :: FilePath -> FileMode -> IO Fd createFile name mode = openFd name WriteOnly (Just mode) defaultFileFlags{ trunc=True } +-- |Close this file descriptor. May throw an exception if this is an +-- invalid descriptor. + closeFd :: Fd -> IO () closeFd (Fd fd) = throwErrnoIfMinus1_ "closeFd" (c_close fd) @@ -216,10 +237,10 @@ -- ----------------------------------------------------------------------------- -- Fd options -data FdOption = AppendOnWrite - | CloseOnExec - | NonBlockingRead - | SynchronousWrites +data FdOption = AppendOnWrite -- ^O_APPEND + | CloseOnExec -- ^FD_CLOEXEC + | NonBlockingRead -- ^O_NONBLOCK + | SynchronousWrites -- ^O_SYNC fdOption2Int :: FdOption -> CInt fdOption2Int CloseOnExec = (#const FD_CLOEXEC) @@ -227,6 +248,7 @@ fdOption2Int NonBlockingRead = (#const O_NONBLOCK) fdOption2Int SynchronousWrites = (#const O_SYNC) +-- | May throw an exception if this is an invalid descriptor. queryFdOption :: Fd -> FdOption -> IO Bool queryFdOption (Fd fd) opt = do r <- throwErrnoIfMinus1 "queryFdOption" (c_fcntl_read fd flag) @@ -236,6 +258,7 @@ CloseOnExec -> (#const F_GETFD) other -> (#const F_GETFL) +-- | May throw an exception if this is an invalid descriptor. setFdOption :: Fd -> FdOption -> Bool -> IO () setFdOption (Fd fd) opt val = do r <- throwErrnoIfMinus1 "setFdOption" (c_fcntl_read fd getflag) @@ -256,6 +279,7 @@ mode2Int RelativeSeek = (#const SEEK_CUR) mode2Int SeekFromEnd = (#const SEEK_END) +-- | May throw an exception if this is an invalid descriptor. fdSeek :: Fd -> SeekMode -> FileOffset -> IO FileOffset fdSeek (Fd fd) mode off = throwErrnoIfMinus1 "fdSeek" (c_lseek fd off (mode2Int mode)) @@ -269,6 +293,7 @@ type FileLock = (LockRequest, SeekMode, FileOffset, FileOffset) +-- | May throw an exception if this is an invalid descriptor. getLock :: Fd -> FileLock -> IO (Maybe (ProcessID, FileLock)) getLock (Fd fd) lock = allocaLock lock $ \p_flock -> do @@ -314,11 +339,13 @@ int2mode (#const SEEK_END) = SeekFromEnd int2mode _ = error $ "int2mode: bad argument" +-- | May throw an exception if this is an invalid descriptor. setLock :: Fd -> FileLock -> IO () setLock (Fd fd) lock = do allocaLock lock $ \p_flock -> throwErrnoIfMinus1_ "setLock" (c_fcntl_lock fd (#const F_SETLK) p_flock) +-- | May throw an exception if this is an invalid descriptor. waitToSetLock :: Fd -> FileLock -> IO () waitToSetLock (Fd fd) lock = do allocaLock lock $ \p_flock -> @@ -328,7 +355,10 @@ -- ----------------------------------------------------------------------------- -- fd{Read,Write} -fdRead :: Fd -> ByteCount -> IO (String, ByteCount) +-- | May throw an exception if this is an invalid descriptor. +fdRead :: Fd + -> ByteCount -- ^How many bytes to read + -> IO (String, ByteCount) -- ^The bytes read, how many bytes were read. fdRead _fd 0 = return ("", 0) fdRead (Fd fd) nbytes = do allocaBytes (fromIntegral nbytes) $ \ bytes -> do @@ -339,6 +369,7 @@ s <- peekCStringLen (bytes, fromIntegral n) return (s, n) +-- | May throw an exception if this is an invalid descriptor. fdWrite :: Fd -> String -> IO ByteCount fdWrite (Fd fd) str = withCStringLen str $ \ (strPtr,len) -> do rc <- throwErrnoIfMinus1Retry "fdWrite" (c_write fd strPtr (fromIntegral len))

3 2

RE: docs for System.Posix.IO
by Simon Marlow 07 Jul '05

07 Jul '05

On 06 July 2005 23:00, Isaac Jones wrote: > I've been working with the System.Posix.IO module and ran into a few > things that could use some documentation. I have added some, and it > would be nice if folks could verify that I'm not making stuff up. I > would actually be happy to paste some of the docs right from the > 'read(2)' and 'write(2)' man pages if that's the correct source? Please commit! > I would commit this, but for some reason, 'make html' doesn't work for > me no matter how many times I run cvs up, configure, and make. I > always get this error: > > > Control/Monad.hs: unknown flags in {-# OPTIONS #-} pragma: _GHC > <<ghc: 9110492 bytes, 2 GCs, 72904/72904 avg/max bytes residency (1 > samples), 12M in use, 0.00 INIT (0.00 elapsed), 0.01 MUT (0.09 > elapsed), 0.00 GC (0.01 elapsed) :ghc>> make[1]: *** > [Control/Monad.raw-hs] Error 1 It looks like the compiler in ghc/compiler/ghc-inplace is out of date, because it doesn't understand the OPTIONS_GHC pragma. Perhaps you need to rebuild in ghc/compiler. Maybe you have sticky tags in part of your source tree, or something? Cheers, Simon

1 0

RE: Time Library 0.1
by Simon Marlow 07 Jul '05

07 Jul '05

I just had a little look around, mainly at System.Time.Calendar. I think the structure is rather complicated - I wanted to find out how to get a CalendarTime for "this time tomorrow", and ended up with this: *System.Time.Calendar> let c' = c{ztTime=zttime{dtDay=dtday{gregDay=day+1}}} where { zttime = ztTime c; dtday = dtDay zttime; day = gregDay dtday } and I found that CalendarTime isn't normalised, you have to convert to UTCTime and then back again to get normalisation (does that do proper normalisation?). Perhaps we should have normaliseCalendarTime. It would be useful to provide this: utcToLocalCalendarTime :: UTCTime -> IO CalendarTime (encodeLocalUTC is taken). I guess the main problem is the difficulty of working with CalendarTime due to the complicated type structure in this module. We should aim to make it easy to do common manipulations of CalendarTime (such as those suggested by Brian Smith). Cheers, Simon On 05 July 2005 09:23, Ashley Yakeley wrote: > Sorry for the hiatus, my job has left less free time to work on this. > > Please take a look at a first attempt at writing a replacement for the > standard time library. > http://semantic.org/TimeLib/ > > Take a look at the Haddock documentation: > http://semantic.org/TimeLib/doc/html/ > > Download the source: > http://semantic.org/TimeLib/TimeLib-0.1.tar.gz > > Or keep up-to-date: > darcs get http://semantic.org/TimeLib/TimeLib/ > > It needs GHC 6.4. Just run "make" to build the library, tests and > documentation. There's also a cabal file (as a completely separate > build process), but I've been having trouble with that on Mac OS X. > > I'm particularly interested in comments from people who try to write > little applications for it. What did you have trouble with? What made > no sense? > > Some points: > > 1. There is no leap second table provided, though there is a type > (LeapSecondTable) for such things. Any software compiled with a fixed > table would soon become out of date. > > 2. There is no table of time-zones provided, since these also change. > However, if there's a good way of getting this from the TZ database on > the machine, I'll add that. > > It is actually possible to get the local time-zone for any given time, > indeed one of the test programs finds your local summertime > transitions. The Timezone type includes name, minutes of offset, and > a "summertime" flag. > > 3. I allow multiple calendars with the DayEncoding class. GregorianDay > is the most obvious instance, but there are some other useful ones > such as ISOWeek. And anyone interested in local cultural calendars > can plug their own in here. > > 4. The TimeLocale type comes from the already existing System.Locale. > > 5. I've tried to balance simplicity and functionality. Of course, the > majority of people will only be interested in ordinary Gregorian date > and time, so I have a simple CalendarTime type synonym with associated > functions: > http://semantic.org/TimeLib/doc/html/System.Time.Calendar.html#8 > > 6. It's not possible to expunge POSIX time. I've hidden it, but it's > used behind the scenes. Why? Because it's the only way to do sensible > arithmetic on UTC times without knowing the leap second table and > without worrying about one-second offsets. > > 7. I include Data.Fixed which provides a fixed-point arithmetic type > (wrapper around Integer). It probably should be in a separate package. > It allows dealing with seconds as a single thing, rather than as an > integer/picoseconds pair. > > 8. I don't have any text-parsing functionality. This is a fair amount > of work, so it would be good to know sensible requirements.

3 2

RE: Time Library 0.1
by Simon Marlow 07 Jul '05

07 Jul '05

On 07 July 2005 09:59, Ashley Yakeley wrote: > In article > <3429668D0E777A499EE74A7952C382D103F41FAB(a)EUR-MSG-01.europe.corp.microso > ft.com>, > "Simon Marlow" <simonmar(a)microsoft.com> wrote: > >> The question is, what's the result of encodeUTC for an invalid >> CalendarTime. Does it roll over the out-of-range fields? What >> happens if you ask for Feb 29 in a non-leap year, for example? > > Right now it's undefined. What should it do? I'd say either it should be an error, or the date should be normalised to a valid date by carrying over any out-of-range values. We should have normalisation anyway, so that might be the easiest behaviour to provide. BTW, I think the library is great. As I said before, I'm a little concerned about how easy it is to do calendar calculations, but given the examples you posted earlier my worries might be unfounded. I think it would be great to collect these examples and put them in the documentation for System.Time.Calendar. Cheers, Simon

1 0

RE: Time Library 0.1
by Simon Marlow 07 Jul '05

07 Jul '05

On 06 July 2005 12:01, Tony Finch wrote: > On Wed, 6 Jul 2005, Simon Marlow wrote: > >> I just had a little look around, mainly at System.Time.Calendar. I >> think the structure is rather complicated - I wanted to find out how >> to get a CalendarTime for "this time tomorrow", and ended up with >> this: > > "this time tomorrow" is not well-defined. What happens then the clocks > change? It's well-defined unless the current time doesn't exist tomorrow, which can happen if a DST switch or leap second happens. In these cases I'd expect either (a) a sensible rollover, or (b) an error, or possibly both behaviours to be available. The question is, what's the result of encodeUTC for an invalid CalendarTime. Does it roll over the out-of-range fields? What happens if you ask for Feb 29 in a non-leap year, for example? Cheers, Simon

3 2

A bug in IntMap
by Simon Peyton-Jones 06 Jul '05

06 Jul '05

Dear library folk, Christian posted a bug to Sourceforge concerning the IntMap library, function split. Can anyone shed any light? Better still, fix it? Simon and I have too many other bugs to nail for 6.4.1 thanks Simon http://sourceforge.net/tracker/index.php?func=detail&aid=1195579&group_i d=8032&atid=108032 "keys $ split 5 $ theMap" sometimes returns values greater 5. I am still trying to pinpoint the bug, but since it shows itself only deep inside a complicated structure, I have problems doing so. One pointer might be, that my values are partially evaluated functions, but this is only a guess. (Changing the data structure to Data.Map restores normal operation in my algorithm)

4 4

another wash compile problem
by Frederik Eaton 06 Jul '05

06 Jul '05

When I compile WASH it takes forever for ghc to finish compiling HTMLMonad98, and on my powerpc system with 151MB free RAM, gcc crashes during the compile. Should some of the larger WASH modules maybe be split into multiple files? I get the sense that ghc compilation time/memory is super-linear in the size of the module being compiled. Frederik

1 1

Data.Monoid
by Ross Paterson 05 Jul '05

05 Jul '05

We can only have one instance of a class for a particular type, so how do we choose? The module Data.Monoid contains the instance instance Monoid (a -> a) where mempty = id mappend = (.) I want to argue that a more natural instance is the pointwise lifting: instance Monoid b => Monoid (a -> b) where mempty _ = mempty mappend f g x = f x `mappend` g x This instance is compositional, in that it builds instances for complex types (like functions of many arguments) out of instances for simpler ones, and it's consistent with the instances for tuples. It's also Haskell 98. If the second instance were used instead, the first could be replaced with one for data Endo a = Endo { runEndo :: a -> a } That's a bit awkward, but not as much as simulating the pointwise lifting by hand.

1 0

Re: wash compile problem
by Frederik Eaton 04 Jul '05

04 Jul '05

Actually I think there was an earlier error, it was refusing to write to package.conf for some reason. Maybe it should give up when this happens, so the relevant error message doesn't get buried? Also, how does one install WASH without root access? I tried to specify the user package.conf, but if this doesn't exist then it looks like the WASH installer won't create it, and also its pathname is not obvious so it seems unfriendly to require users to specify the literal path... Frederik On Sat, Jul 02, 2005 at 05:02:59PM +0200, Peter Thiemann wrote: > >>>>> "fe" == Frederik Eaton <frederik(a)a5.repetae.net> writes: > > fe> When I try to install the new version of WASH (2.4.6), I get errors > fe> saying that certain dependencies don't exist: > > fe> ... > fe> Reading package info from stdin... done. > fe> ghc-pkg: dependency Utility doesn't exist > > Strange. The Utility package is also provided by WASH. > Was there an error message in the Utility subdirectory? > What is the output of `ghc-pkg list' afterwards? > Did you run `make install' with sufficient karma to change the package > registry? Eg, as root? > > -Peter >

1 1