May 2002 - Glasgow-haskell-users

specifying common 'dis's in greencard
by Hal Daume III 30 May '02

30 May '02

Hi, I have three modules which all use the same (or some of the same) %dis directives; I'd like to pull these out into their own file (a module called "MyDIS" or something), but I can't figure out how to get green-card to look at those definitions (I've `import`ed it in the modules that need it and put it's directory in the search path with -i). Unfortunately, even so it complains that it doesn't know how to marshall my datatypes. Any suggestions? - Hal -- Hal Daume III "Computer science is no more about computers | hdaume(a)isi.edu than astronomy is about telescopes." -Dijkstra | www.isi.edu/~hdaume

3 4

RE: instance Ord FiniteMap
by Simon Marlow 29 May '02

29 May '02

I'll happily incorporate the code if someone sends me a patch... Cheers, Simon > -----Original Message----- > From: Hal Daume III [mailto:hdaume@ISI.EDU] > Sent: 29 May 2002 15:03 > To: Johannes Waldmann > Cc: GHC Users Mailing List > Subject: Re: instance Ord FiniteMap > > > I agree; the problem is that I fear that making my own > instance by using > setToList will be very inefficient (or at least much more so than an > instance which actually looks at the tree structure). > > -- > Hal Daume III > > "Computer science is no more about computers | hdaume(a)isi.edu > than astronomy is about telescopes." -Dijkstra | www.isi.edu/~hdaume > > On Wed, 29 May 2002, Johannes Waldmann wrote: > > > > for instance, Sets of Sets of things would be really nice. > > > > Sure. One could simply use lexicographic ordering > > (i. e. s1 `compare` s2 = setToList s1 `compare` setToList s2) > > or length-lexicographic ordering (for efficiency) > > ... = (cardinality s1, setToList s1) `compare` > (cardinality s2, setToList s2) > > > > As you write, there seems to be no reason not to do this. > > An Ord instance should be a linear ordering, and the above are. > > -- > > -- Johannes Waldmann ---- > http://www.informatik.uni-leipzig.de/~joe/ -- > > -- > joe(a)informatik.uni-leipzig.de -- phone/fax (+49) 341 9732 204/252 -- > > > > _______________________________________________ > Glasgow-haskell-users mailing list > Glasgow-haskell-users(a)haskell.org > http://www.haskell.org/mailman/listinfo/glasgow-haskell-users >

3 2

Re: GHC ebuild done!
by Sven Moritz Hallberg 29 May '02

29 May '02

On Wednesday 29. May 2002 17:07, you wrote: > > With the help of the excellent new guide on porting GHC > > (Thank you Simon!) > > I have just created an ebuild for GHC. I thought I'd let you > > know, in case > > you're still working on it. I don't bootstrap GHC 5 directly > > from HC files, > > but GHC 4.08.3 (as some suggested) and go on from there. I've > > submitted the > > ebuild yesterday as Bug #3092 to the Gentoo Bugzilla. Feel > > free to comment if > > you want to look at it. > > Does this mean we can claim that Gentoo Linux is supported? :) I hope it will mean that in the very near future. That is, once the ebuild is in Gentoo CVS. I'm currently waiting for that to happen, but they're probably still waiting for their compile to finish... ;^) > Is there a > useful link I can add from the GHC download pages? Hm. No. Ebuilds are the fundament of Portage, the central Gentoo Linux package system, which is directly available to anyone running Gentoo Linux. Maybe you can add a comment saying that all Gentoo users with a recent Portage tree can now simply 'emerge ghc' for much Haskell fun. But, wait with that until the ebuild enters CVS. I'll make a seperate announcement to this liste once that's happened. So long, Sven Moritz

1 0

RE: GMP from haskell
by Simon Marlow 29 May '02

29 May '02

> I should have been more specific :). I was referring to the more > complicated things, such as Lucas numbers, Binomial coefficients, etc. There isn't a full GMP binding as such. However, we do make use of more than just the simple arithmetic in GMP - for example, Prelude.gcd on Integers uses GMP's gcd, and similarly for lcm. Cheers, Simon

1 0

GMP from haskell
by Hal Daume III 29 May '02

29 May '02

The gnu web page (www.gnu.org/manual/gmp-4.0.1/html_node/gmp_70.html) claims that Haskell (GHC) has bindings to GMP. Is this true? How can I access these routines? - Hal -- Hal Daume III "Computer science is no more about computers | hdaume(a)isi.edu than astronomy is about telescopes." -Dijkstra | www.isi.edu/~hdaume

3 3

Overlapping Instance Thoughts
by Ashley Yakeley 28 May '02

28 May '02

At 2002-05-21 07:25, Simon Peyton-Jones wrote: >Overlapping instances are a terrific swamp and while they >are interesting I don't think I'm going to spend a lot of >time on them in the near future, I'm afraid. But perhaps >someone else can.. I would say the overlapping instances issue is one of the more important ones in the extended language. I'm sure most Haskell programmers have been tripped up by it at least once or twice. It's certainly the thing I most frequently bang my head against. There are in fact some different issues here. 1. There are instances that don't actually overlap, but only because of class contexts: class C a; class D a; instance C Bool; instance (D a) => C a; GHC rejects this, because it only looks at the instance heads, not the contexts. Ideally GHC would accept this as non-overlapping, but then prohibit instances of different classes that caused the two instances to overlap, such as this: instance D Bool; 2. There's also times when the programmer would like instances that really do overlap. Now GHC already has a -fallow-overlapping-instances flag to allow overlaps where one is a strict subset of the other: the more special case takes priority over the more general. But sometimes this isn't enough. For instance, while writing a Scheme interpreter, I wanted to do something like this: newtype Constant a = MkConstant a; -- members omitted class (Monad m) => MonadGettableReference m r where ... instance (Monad m) => MonadGettableReference m Constant where ... instance (MonadGettableReference m r) => MonadGettableReference (SchemeCPS r (m p)) r where ... Unfortunately the two instances overlap, here: MonadGettableReference (SchemeCPS Constant (m p)) Constant In fact it wouldn't actually matter which instance was used in this case, and I suspect that's typical of good class design. So perhaps an -fallow-incoherent-instances flag that just arbitrarily picked one might be helpful. Better would be some mechanism for allowing the programmer to specify which instances defer to which. Or else extend -fallow-overlapping-instances to allow "disambiguating" instances: class C a b; instance C a (); instance C () a; instance C () (); Here, the third instance is precisely the overlap of the other two. But apparently this too has problems: >I don't understand all the implications of this. >For example, what if we have an instance that >doesn't match C () () now, but may do 'later' >when a type variable has been instantiated. I don't doubt Simon PJ when he says this is a terrific swamp. I think any work anyone might wish to do cutting through the swamp that ended up in the language would be very valuable. -- Ashley Yakeley, Seattle WA

1 0

POpen/createPipe errors
by Hal Daume III 28 May '02

28 May '02

Hi Jens, GHC folk, I recently ran into a problem with POpen. I'm using it lots (on the order of tens of thousands) of times to run a very fast program on small amounts of data and after a little while, it dies, saying: Fail: system error Action: createPipe this can be reproduced in this program: import POpen main = mapM_ (\x -> popenEnvDir "ls" [(x:"*")] Nothing Nothing Nothing) (['a'..'z'] ++ ['A'..'Z']) i ran in to a similar problem with reading files wherein i was opening (and not closing) too many files using readFile and my solution was to write a function: readFileCloseBy :: DeeqSeq a => (String -> a) -> FilePath -> IO a readFileCloseBy f fp = do h <- openFile fp ReadMode d <- hGetContents h let res = f d res `deepSeq` hClose h res `deepSeq` (return res) this function opens and reads a file, parses it using the function provided, deepSeqs the result and closes the file. could a similar thing be done with popen? - Hal -- Hal Daume III "Computer science is no more about computers | hdaume(a)isi.edu than astronomy is about telescopes." -Dijkstra | www.isi.edu/~hdaume

1 0

Announce: HTk - a GUI toolkit for Haskell
by Christoph Lueth 28 May '02

28 May '02

It is our pleasure to announce the release of HTk, a graphical user interface toolkit and library for Haskell. Version 0.98 (World Cup edition) Htk is a typed, portable encapsulation of Tcl/Tk into Haskell. Its distinctive features are the use of Haskell types and type classes for structuring the interface, an abstract notion of event for describing user interaction, and portability across Windows, Unix and Linux. HTk is easily installed -- we provide precompiled binaries for Windows & Linux on x86 and Sparc/Solaris. All you need apart from one of these and a matching version of GHC (5.02.3) is a reasonable recent installation of the Tcl/Tk interpreter "wish". HTk should be particularly useful for teaching, but its reliability and extensive functionality also make it useful in many Haskell projects requiring easily designed graphical user interfaces. This release is for version 0.98. We hope to release version 1.00 in August. The HTk Homepage is http://www.informatik.uni-bremen.de/htk There, you will find binary downloads, source downloads, documention, and much more. HTk is (C) Universität Bremen, 2002. All rights reserved. Please do not hestitate to directed any further questions to <uniform(a)informatik.uni-bremen.de> George Russell <ger(a)informatik.uni-bremen.de> Christoph Lüth <cxl(a)informatik.uni-bremen.de>

1 0

GHC ebuild done!
by Sven Moritz Hallberg 28 May '02

28 May '02

Hi Tom, With the help of the excellent new guide on porting GHC (Thank you Simon!) I have just created an ebuild for GHC. I thought I'd let you know, in case you're still working on it. I don't bootstrap GHC 5 directly from HC files, but GHC 4.08.3 (as some suggested) and go on from there. I've submitted the ebuild yesterday as Bug #3092 to the Gentoo Bugzilla. Feel free to comment if you want to look at it. Regards, Sven Moritz Hallberg

1 0

RE: Runtime importing of modules in GHC
by Simon Marlow 27 May '02

27 May '02

> .. and now, my real reason for posting the previous message 8) > > Unfortunately, the big limitation of this is that we're going > through the FFI to get things to work. > > This kinda sucks because that means that the exported functions > have to stick to exporting one of the prim_types (Int, Char, > Float, Double, Ptr, StablePtr, and a few others). > > So, is there some wizardry out there which allows using the RTS > Linker to load up .o modules which are not created with the FFI? > I've been experimenting with it, and I just get segfaults if > I try to load the adder_closure, adder_entry, adder_fast1 or > adder_info symbols. You *should* be able to just get the address of adder_closure and coerce it to the right type. The difficulty with this is that lookupSymbol returns a Ptr, which is an unboxed type, whereas the function type is a boxed type. Coercing unboxed types to boxed types is highly unsafe, because they have different representations on the stack. However, GHCi does exactly this, and coerces from one to the other using addrToHValue# (see ByteCodeLink.lookupCE). It's dangerous, but we'd know pretty quickly if it didn't work. The Right Thing To Do would be to do the communication with the RTS linker in terms of StablePtrs, but there's a performance overhead for doing that (perhaps not too bad, I haven't measured it). Cheers, Simon

3 2