November 2001 - Haskell

List module in library report exports []((:), [])
by Ian Lynagh 26 Nov '01

26 Nov '01

Hiyas A few more things I'm afraid :-( The List module in the library report exports []((:), []) which, as I mentioned in a previous thread WRT the GHC prelude, the report doesn't allow in the exports list. The block at the top listing the exports and types has been split into 2 at an apparently arbitrary point for no obvious reason? deleteFirstsBy has no corresponding deleteFirsts and is not mentioned in the text. Actually, having looked at the definitions it seems to be what we would like to call \\By, so this should be noted in the "By" section. insert and the zip/unzip stuff is also not explained in the text - don't know if this is deliberate or not (insert is also mentioned with sort). sortBy is mentioned before the "By" section unlike the other By variants for no obvious reason. "exmaple" genericIndex seems to be a generic !!, but The prefix "generic" indicates an overloaded function that is a generalised version of a Prelude function. and index is something completely different and defined in Ix. Clearly the function can't be called generic!!, so I suggest the non-uniformity be noted. I am also curious as to why other functions don't have generic equivalents, e.g. elemIndex, findIndices - is it because they were also not considered important enough (as per the By functions)? The indentation of unfoldr, maximumBy and minimumBy seems to have gone wrong with some things not being indented at all (only looked at the HTML version). I think unionBy eq xs ys = xs ++ foldl (flip (deleteBy eq)) (nubBy eq ys) xs may be clearer written as unionBy eq xs ys = xs ++ deleteFirstsBy eq (nubBy eq ys) xs Thanks Ian

1 0

Library report examples
by Ian Lynagh 26 Nov '01

26 Nov '01

Hi guys, The library report defines -- Diagonal of a square matrix diag :: (Ix a) => Array (a,a) b -> Array a b diag x = ixmap (l,u) (\i->(i,i)) x where ((l,l'),(u,u')) | l == l' && u == u' = bounds x but ghc, hugs and nhc98 all loop (trying to get and test the value of l I believe). I am also curious why, for example, row :: (Ix a, Ix b) => a -> Array (a,b) c -> Array b c row i x = ixmap (l',u') (\j->(i,j)) x where ((l,l'),(u,u')) = bounds x isn't written as row :: (Ix a, Ix b) => a -> Array (a,b) c -> Array b c row i x = ixmap (l,u) (\j->(i,j)) x where ((_,l),(_,u)) = bounds x ~~~ ~~~ ~~~ Thanks Ian

1 0

RE: heap and walking through a very long list
by Simon Peyton-Jones 26 Nov '01

26 Nov '01

You can do it specifically for GHC, not not in standard Haskell (except grossly inefficiently). To see how to do it in GHC, look at the implementation in the source code (ghc/lib/std), which is online (check the "CVS repository" link on GHC's home page). Meanwhile you could use the IArray stuff. Simon | -----Original Message----- | From: Joe English [mailto:jenglish@flightlab.com] | Sent: 25 November 2001 00:12 | To: haskell(a)haskell.org | Subject: Re: heap and walking through a very long list | | | | Simon Peyton-Jones wrote: | | > There should really be a strict accumArray, just as there | should be a | > strict foldl. | | Yes, please! | | Is there a way to write a strict version of accumArray in | Haskell 98, or does this need to be done by the implementation? | | | --Joe English | | jenglish(a)flightlab.com | | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell |

1 0

Functional Metapost maintained?
by Ferenc Wagner 25 Nov '01

25 Nov '01

Hi, I got to know Haskell through Functional Metapost. Those who use LaTeX and feel like programming some graphics in Haskell, might have a look at this: http://tex.loria.fr/prod-graph/functional-metapost-latex.pdf Trying to use the package I had some problems with the current Hugs and GHC, too. It seems like the version I have is quite dated: it uses fromDouble, fromInt, fromRealFrac etc. which GHC barks at. I made some modifications to have it compile, so that I can use it know, but I still have a couple of problems that should be washed out. But before I go for it, I'd like to know if somebody have got a newer version, or somebody else maintains it, or have some relevant information about that package, or anything... On the Web I found nothing, really. Cheers: Feri.

3 5

RE: heap and walking through a very long list
by Simon Peyton-Jones 25 Nov '01

25 Nov '01

The trouble here is the usual one: accumArray is lazy, and simply accumulates a giant closure in each array slot. Then, when it is evaluated, the stack gets big. There should really be a strict accumArray, just as there should be a strict foldl. There isn't in Haskell 98, but if you are using a fixed type (like Int) you can use GHC's IArrays (see documentation on the 'lang' package). There's an accumArray in there too, and it'll be strict for sure. Simon | -----Original Message----- | From: Lukasz Pankowski [mailto:lupan@zamek.gda.pl] | Sent: 18 November 2001 23:18 | To: haskell(a)haskell.org | Subject: heap and walking through a very long list | | | | Hi. I wonder if there are any methods of walking through a | very long list without a huge heap. It is good that a | laziness makes creation of large stuctures delayed, but it | seems that destruction of never again used beginning of a | list is also delayed (probably because of other reason). | | Consider the simple program: | | main = do putStrLn $ show $ hist_inc 100000 3 | | hist_inc n b = hist (0,b) $ take n $ repeat b | | hist bnds is = accumArray (+) 0 bnds [(i, 1) | i <- | is, inRange bnds i] | | and it's output (compiled with GHC): | | $ ./hist_inc | Stack space overflow: current size 1048576 bytes. | Use `+RTS -Ksize' to increase it. | | | I am using GHC 5.0, does it have any optimization to overcome | it (`-O -fvia-C -O2-for-C' does not do this). It is obvious | that the beginning of the list is no longer needed (does | garbage collector now this?). | | If there is a way is it written somewhere in documentation? | | I feel it insane to have to have let say 256Mb of memory | because I have 1 million measurements on a list. Currently I | pipe the results to a C program, nice UN!X practice which I | would like avoid. | | | -- | | =*= Lukasz Pankowski =*= | | | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell |

2 1

virtual functions
by Rahul Bhargava 24 Nov '01

24 Nov '01

Hi, I seek an elegant solution to introduce in Haskell code (with GHC as it is currently) with capabilities akin to virtual functions in C++. Say, I have a module with functions for serial operations. I want to introduce another module with distributed operations but maintain function names. Polymorphism should resolve which underlying function is to be called by checking whether the caller wants to perform operations on a serial or a distributed structure. Example, sum sums elements in a local list. sum in the distributed module works on distributed lists. A third sum should sit on top of things and be visible to the user which chooses which sum is being talked about based on context. Cheers Rahul

1 0

RE: List.partition should be more lazy
by Simon Peyton-Jones 23 Nov '01

23 Nov '01

It was wrong in the Haskell report too (now fixed; see my home page). Also fixed in GHC 5.02. Simon | -----Original Message----- | From: Bastiaan Heeren [mailto:bastiaan@cs.uu.nl] | Sent: 23 November 2001 13:55 | To: haskell(a)haskell.org | Subject: List.partition should be more lazy | | | Using the function partition from the List module, a control | stack overflow occurs when evaluating the following expression: | | | List> head $ fst $ partition even [1..] | | (35929 reductions, 63867 cells) | ERROR: Control stack overflow | | | The definition of partition in both hugs and GHC libraries is: | | | partition :: (a -> Bool) -> [a] -> ([a],[a]) | partition p xs = foldr select ([],[]) xs | where select x (ts,fs) | p x = (x:ts,fs) | | otherwise = (ts,x:fs) | | | The helper-function select is strict in its second argument | because of the pattern match on the tuple (ts,fs). Making the | pattern match lazy by adding a ~ solves the problem! | | | partition :: (a -> Bool) -> [a] -> ([a],[a]) | partition p xs = foldr select ([],[]) xs | where select x ~(ts,fs) | p x = (x:ts,fs) | | otherwise = (ts,x:fs) | | | Cheers, | Bastiaan | | | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell |

1 0

List.partition should be more lazy
by Bastiaan Heeren 23 Nov '01

23 Nov '01

Using the function partition from the List module, a control stack overflow occurs when evaluating the following expression: List> head $ fst $ partition even [1..] (35929 reductions, 63867 cells) ERROR: Control stack overflow The definition of partition in both hugs and GHC libraries is: partition :: (a -> Bool) -> [a] -> ([a],[a]) partition p xs = foldr select ([],[]) xs where select x (ts,fs) | p x = (x:ts,fs) | otherwise = (ts,x:fs) The helper-function select is strict in its second argument because of the pattern match on the tuple (ts,fs). Making the pattern match lazy by adding a ~ solves the problem! partition :: (a -> Bool) -> [a] -> ([a],[a]) partition p xs = foldr select ([],[]) xs where select x ~(ts,fs) | p x = (x:ts,fs) | otherwise = (ts,x:fs) Cheers, Bastiaan

1 0

many urls
by Alex P.Gordienko 23 Nov '01

23 Nov '01

1 0

PhD positions in Informatics, Bergen University, Norway
by Marc Bezem 23 Nov '01

23 Nov '01

2 PhD positions Institute for Informatics University of Bergen, Norway 1. The project. --------------- MoSIS is a IKT project funded by the Norwegian Science Council (NFR) for the period 2002-2005. Its descriptive full title is: Modularity in large Software and Information Systems Modularization techniques are essential for mastering the complexity of large systems. There are many approaches: agents, components, libraries, ... The overall aim of the project is to develop a conceptual and formal framework for the composition and interaction of software modules at various levels of abstraction. The two relevant subprojects aim at applying formal techniques based on type theory, algebra and/or category theory to programming `in the large'. For more information on the project, see http://www.ii.uib.no/MoSIS 2. The profile of the candidates. --------------------------------- You have a Master degree (or equivalent) in Computer Science, Informatics, Mathematics or Logic. You enjoy working in an internationally oriented, English speaking research environment. 3. The jobs. ------------ The positions are opening up in 2002 and last for three years. There are in principle no teaching duties. There is a perspective for prolongation of at most one year as a research associate. The salary is about EUR 32.000 per year. 4. Information and how to apply. -------------------------------- More information can be obtained from Prof.Dr. M.A. Bezem (bezem(a)ii.uib.no) You are invited to send an application letter by e-mail, together with a curriculum vitae and the names and addresses of two references, before 21 December 2001. Applicants from outside the European Economic Area (EEA) will be considered. In a later stage of the procedure we may ask you to write a formal application to the faculty. 5. Bergen, Norway. ------------------ Bergen is the second-largest city in Norway, beautifully situated between 7 mountains (up to 600m) and fjords. Bergen has about 230.000 inhabitants and all facilities, including a nearby airport. The climate (including tax pressure) is milder than most people expect, but it can rain quite hard, indeed. For more information on - Bergen, see http://www.bergen-guide.com/ - Bergen University, see http://www.uib.no/ - Institute for Informatics, see http://www.ii.uib.no/ - Marc Bezem, see http://www.ii.uib.no/~bezem

1 0