- Haskell - Haskell.org

kind inference question
by Bernard James POPE 12 Feb '01

12 Feb '01

Hi all, I'm in the middle of writing kind inference for haskell, as part of a type checker/inferer. After reading the section in the haskell report about kind inference (section 4.6), I began to wonder why kind inference must be done in dependency order. Some friends and I came up with the following example which raises an issue: data C x = Foo (B x) (x Int) data B y = Bar According to the report (and ghc and hugs) this is ill-kinded. The reason is that kinds must be inferred in dependency order, and when parts of a kind are not fully determined they default to * (star). So: inferring the kind of B first gives: kind (B) = * -> * this results in a kind error because in the definition of C, B must have kind (* -> *) -> *, which differs from the inference made above However, an alternative kind inference algorithm might allow the above declarations, by performing kind inference of C and B together (effectively unifying the kind of x and the kind of y). This would result in: kind (C) = (* -> *) -> * kind (B) = (* -> *) -> * I'm not suggesting that allowing this inference would be a good idea, but I am wondering why haskell requires the dependency ordering. Perhaps there are better examples that elucidate the motivation for dependency ordering. Basically I'm just curious. Regards Bernie.

2 1

Contexts for data constructors.
by anatoli 11 Feb '01

11 Feb '01

A lot of thanks to everybody who helped me with fundeps. Now for another question (I asked it already but I think the message was lost). Wouldn't it be nice if we could specify contexts for data constructors? I mean something like that: data Quick a = Eq a => Unsorted [a] | Ord a => Sorted (Tree a) findquick :: a -> Quick a -> Maybe a findquick (Unsorted list) = ... -- find slowly using linear search and (==) findquick (Sorted tree) = ... -- find quickly using binary search and (<) Note that findquick does not require any context with this hypothetical extension. Indeed, in the "Unsorted" branch of findquick (Eq a) is required, but we know that it's satisfied because of the constraint on the data constructor. Similarly for the Sorted branch and (Ord a). I tried achieving this with current Haskell, but couldn't. Everything I tried gives me "findquick :: Ord a => a -> Quick a -> Maybe a". Regards -- anatoli __________________________________________________ Do You Yahoo!? Get personalized email addresses from Yahoo! Mail - only $35 a year! http://personal.mail.yahoo.com/

1 0

Fran Problems
by Steinitz, Dominic J 11 Feb '01

11 Feb '01

I downloaded Fran and had a go at running the demos\test.hs C:\hugs98>hugs +o __ __ __ __ ____ ___ _________________________________________ || || || || || || ||__ Hugs 98: Based on the Haskell 98 standard ||___|| ||__|| ||__|| __|| Copyright (c) 1994-1999 ||---|| ___|| World Wide Web: http://haskell.org/hugs || || Report bugs to: hugs-bugs(a)haskell.org || || Version: May 1999 _________________________________________ Hugs mode: Restart with command line option +98 for Haskell 98 mode Reading file "c:\hugs98\lib\Prelude.hs": Hugs session for: c:\hugs98\lib\Prelude.hs Type :? for help Prelude> ERROR "c:\hugs98\lib\Fran\src/ShowFunctions.hs" (line 5): Overlapping instances for class "Show" *** This instance : Show (a -> b) *** Overlaps with : Show (a -> b) *** Common instance : Show (a -> b) I edited ShowFunctions to get rid of the overlapping instance: -- From QuickCheck distribution. module ShowFunctions where {- instance Show (a->b) where show f = "<function>" -} But then ran into another problem: Reading file "c:\hugs98\lib\Fran\src/RenderImage.hs": ERROR "c:\hugs98\lib\Fran\src/RenderImage.hs": Entity "readFile" imported from m odule "Win32" already defined in module "Prelude" SoundB> I was a bit reluctant to modify Prelude or Win32. Can anyone help? Is it just that I'm on a back release on Hugs? Dominic Steinitz ------------------------------------------------------------------------------------------------- 21st century air travel http://www.britishairways.com

1 0

Types from values using existential types?
by Dylan Thurston 11 Feb '01

11 Feb '01

I recently had an idea to allow one to create types from values. This would have many uses in Haskell; e.g., it would be very nice to have a type of nxn matrices for n that are not necessarily statically determined. (It is easy to create a type of matrices and check that the sizes are compatible at run time; the point is to do more checks at compile time.) Of course, one has to be careful to preserve decidability of type checking. The idea is to provide a function that returns an element of an undetermined (existential) type: class Singleton a b where singletonValue :: b -> a singleton :: forall a . a -> (exists b . (Singleton a b) => b) singleton returns a token of some new type; this token can then be passed around, stored in data structures, etc., to guarantee type consistency. I have no idea what existential types would do to the Haskell type system; is there a way to add them to preserve decidability, etc, etc? They would clearly make some simple program transformations illegal: for the idea to be useful, the type of (a,a) where a = singleton 5 should be of type exists a . (Singleton Int a) => (a, a) while the type of (singleton 5, singleton 5) has to be ((exists a . (Singleton Int a) => a), (exists b . (Singleton Int b) => b)) for decidability. Are there any pointers to previous work I should look at? Thanks, Dylan Thurston

3 4

`Covertible' class. Reply.
by S.D.Mechveliani 10 Feb '01

10 Feb '01

Dylan Thurston <dpt(a)math.harvard.edu> writes > In thinking about various issues with the numeric classes, I came up > with the following question: Is there a problem with having a class > 'Convertible' as follows? > > class Convertible a b where > convert :: a -> b > > [..] The basic algebra library BAL http://www.botik.ru/pub/local/Mechveliani/basAlgPropos/bal-pre-0.01/ suggests class Cast a b where cast :: a -> b -> a If s is an element of a certain domain, then one can use the construction cast s x to convert various data x to corresponding canonical values in the domain defined by s. For example, if s <- Z[x] is a polynomial over Integer, then the expressions cast s 2, cast s (2,3) give the constant 2, considered as a polynomial, and a one-term polynomial equal to 2*x^3. Also BAL overloades +, * ... by hiding-reexporting the standard Prelude and implements the algebraic categories Group, Ring ... ----------------- Serge Mechveliani mechvel(a)botik.ru

4 3

Detail: laziness in show
by Patrik Jansson 09 Feb '01

09 Feb '01

I just uncountered an interesting example of (hidden) laziness - here is a very short session with hugs (February 2000): Prelude> undefined :: String " Program error: {undefined} Note the starting double quote on the line before the error! In a more complicated context this puzzled me for a while until I realized that there is an implicit application of show that produces the quote. The show instance for strings prints a '"' character before it starts evaluating the argument. (With the setting -u for hugs, nothing is printed.) So, what about other types? It seems that laziness is not equally distributed between the show instances: Prelude> undefined :: Char Program error: {undefined} Prelude> undefined :: [Bool] Program error: {undefined} ... even though one would expect a '\'' character and a '[' character to be printed. A closer inspection of the prelude shows that the default definition of showList is defined by cases (a showChar '[' could be factored out to mimic the String behaviour) and similarly the definition of showsPrec for Char could be changed to allow the starting quote to be output immediately. (showList for Char is an example of how this could be written) This is certainly a small detail, but as the variants have different semantics, and the Prelude is part of the Haskell Report, this is part of Haskell. To show that the choice of lazy/strict is not quite obvious we can take two more types as examples. After a quick look at the definition of show for pairs, instance (Show a, Show b) => Show (a,b) where showsPrec p (x,y) = showChar '(' . shows x . showChar ',' . shows y . showChar ')' I thought I could get hugs to print the two characters (" before the error, but this is the result: Prelude> undefined :: (String,String) Program error: {undefined} A small ~ before the pair (x,y) would change it to what I would have guessed. Up to this point I thought that maybe all show definitions should be made as lazy as possible. Finally the last example to show that the show-for-pair might not be wrong anyway: instance Show () where showsPrec p () = showString "()" This is also a strict pattern match and the alternative using ~() would print () both for bottom and for () (that is, for all values of type ()). So perhaps the odd bird is show for Strings. /Patrik

2 1

Re: FranTK
by Jan Kort 09 Feb '01

09 Feb '01

Hi Petra, Since you are a system administrator, I asume that you will also want to install FranTk so other people can use it. I'm not sure this is possible. I did manage to get it working for myself localy, but a "make install" doesn't do anything. If you are in a hurry, you can use the fixes below and let the users copy your directory and make their programs in the FranTk/demos directory. The whole directory is about 27M. Like always, a better solution is to wait a little: According to the news on the FranTk homepage the library will become part of hslibs. This means that it will be distributed with ghc and it should also have a proper installation. I would wait for that. I will CC the mail to haskell(a)haskell.org, maybe someone there can give you an estimate on when FranTk will be released as part of hslibs. Regards, Jan To get FranTk for ghc 4.06 to work with ghc 4.08 apply the following fixes: Run configure like normal, i.e.: ./configure --prefix=/scratch/kort/pkgs/frantk Change the file FranTk/src/FranSrc/Compatibility.ghc.hs to ======================================================================== -- this module is for GHC; GSL module Compatibility ( double2Float , yield , debugMsgLn , setDebug , mkWeakIORef , toInt , fromInt ) where import GlaExts import NumExts(doubleToFloat) import PrelRead(readDec) import IOExts import Concurrent import Exception double2Float = doubleToFloat setDebug :: Bool -> IO () setDebug _ = return () debugMsgLn :: String -> IO () debugMsgLn s = assert (trace s `seq` True) $ return () ======================================================================== In the file FranTk/src/FranTkSrc/Makefile change the line: HC_OPTS+= -i../FranSrc:../TclHaskellSrc -fallow-overlapping-instances -fallow-undecidable-instances -syslib misc to: HC_OPTS+= -i../FranSrc:../TclHaskellSrc -fallow-overlapping-instances -fallow-undecidable-instances -syslib data In the file FranTk/demos/Makefile change the line: HC_OPTS+= -i../src/FranTkSrc:../src/FranSrc:../src/TclHaskellSrc -fallow-overlapping-instances -fallow-undecidable-instances -syslib misc to: HC_OPTS+= -i../src/FranTkSrc:../src/FranSrc:../src/TclHaskellSrc -fallow-overlapping-instances -fallow-undecidable-instances -syslib data If the compiler complains about not finding the file "tcl.h", in the file FranTk/src/TclHaskellSrc/Makefile change the rule: tclhaskell.o: tclhaskell.c tclhaskell.h $(HC) -c tclhaskell.c to: tclhaskell.o: tclhaskell.c tclhaskell.h $(HC) -I/opt/arch/lib/tk8.3/include -c tclhaskell.c where /opt/arch... is the place "tcl.h" and "tk.h" are. In the directory FranTk/src/TclHaskellSrc do: rm *.hi *.o In the file FranTk/src/TclHaskellSrc/Makefile add TclPrim.hs to the HS_SRCS list. In the directory FranTk/src/FranSrc do: rm *.hi In the directory FranTk/demos do: rm *.hi And then continue the normal installation with: gmake boot gmake all

3 3

Chat Live with a CPA
by info＠cpadirectoryinc.com 09 Feb '01

09 Feb '01

FREE Chat with a CPA on your tax questions, wondering what 8a on your 1040 means. Our CPAs will answer your questions FREE. Need a CPA in your area? Search our 450,000 CPA database by city, state and industry. <a href="http://www.cpadirectory.com/cpadirectory/askacpa/ask_live_chat.cfm"> http://www.cpadirectory.com/cpadirectory/askacpa/ask_live_chat.cfm </a> Articles and FAQ's on Buying a Business, Estate Taxes, College Aid and Financial Investments are also on <a href="http://www.cpadirectory.com"> http://www.cpadirectory.com </a> If you have received this email in error or wish to be removed from our list, email info(a)cpadirectoryinc.com with unsubscribe as the subject.

1 0

Re: "Job Adverts" on haskell.org
by C.Reinke 08 Feb '01

08 Feb '01

> Haskell.org has a page "Job Adverts". > http://www.haskell.org/jobs.html I like the new scope. But to minimise confusion of people looking for paid jobs (which now exist), why not reflect the explanations in the first paragraph in the page structure? Two main sections should be sufficient, with the current sections falling naturally into one of the two: A Adverts for positions involving Haskell - Positions in Academia - Positions in Industry B Descriptions of jobs that need doing - new libraries/tools needed - medium-level GUI library needed - new caretakers for existing libraries/tools needed -.. - calls for contributions - donate your programs for testing and benchmarking -.. - calls for cooperations - project X.. (change order if necessary;-) > If you are working on some Haskell project and could > do with some help, then advertise here (and on the Haskell mailing > list). Even more importantly: If you built some Haskell software and > unfortunately no longer feel able to maintain it, then please advertise > here (and on the Haskell mailing list) for new people to take it over > and give it a home. Thus we may alleviate the problem of the large > number of Haskell tools and libraries that are half complete and no > longer maintained. Excellent ideas, especially the focus on keeping existing things in good working order by passing them on to new caretakers! Some aspects of the Haskell wish list might overlap with aspects of the suggested section B? I would suggest to keep the two linked to each other, with the wish list focusing on would-be-nice-to-have, and the task list focusing on needs-to-be-done and is-going-to-be-done. So that ideas might start on the wish list, but move to the task list when they become more concrete (either there exists a concrete customer, or a project is forming with the first names of contributors already written down or work having started, or a current maintainer is looking for someone to take over). Cheers, Claus

2 1

ANNOUNCE: hmake-2.02
by Malcolm Wallace 08 Feb '01

08 Feb '01

We are pleased to announce the release of version 2.02 of 'hmake', the intelligent compilation manager for Haskell compilers. http://www.cs.york.ac.uk/fp/hmake/ ftp://ftp.cs.york.ac.uk/pub/haskell/hmake/ What's new in 2.02? ------------------- This is mainly a bugfix release. * Improved error-reporting. When a module cannot be found, it now shows where the demand for the module arose, and which directories were searched. * Added knowledge of extended file suffixes like .p.o, .T.o, and .T.hi. * Fixed a bug in the generation of Makefile dependencies in the presence of a -I option. * Ghc's options "-syslib n" and "-package n" no longer need to be enclosed in double quotes to protect them from hmake. Regards, Malcolm

2 1