November 2008 - Haskell-Cafe

Proof that Haskell is RT
by Andrew Birkett 13 Nov '08

13 Nov '08

Hi, Is a formal proof that the Haskell language is referentially transparent? Many people state "haskell is RT" without backing up that claim. I know that, in practice, I can't write any counter-examples but that's a bit handy-wavy. Is there a formal proof that, for all possible haskell programs, we can replace coreferent expressions without changing the meaning of a program? (I was writing a blog post about the origins of the phrase 'referentially transparent' and it made me think about this) Andrew -- - http://www.nobugs.org -

17 28

HOpenGL shading problem
by Trin 13 Nov '08

13 Nov '08

Hi, last few days I spent making remake of glxgears. Nothing fancy, I just wanted easily configurable gears. The current performance is only about 70% of glxgears. But I have a problem with shading. Whatever I have been trying, I still don't understand how the shading can be activated. I have light, I have color, but not even Flat shading. Please can you give me a hand? Thanks, Martin PS: I didn't find any attachment size policy for this list, so I hope that the attachment is small enough.

2 1

re: [Haskell-cafe] ANNOUNCE: rewriting-0.1
by Greg Meredith 13 Nov '08

13 Nov '08

Thomas, Did you explore nominal rewrite at all? Do you know if it might be possible to use the scrap-your-nameplate package to implement some useful subset of the nominal rewrite machinery? Best wishes, --greg -- L.G. Meredith Managing Partner Biosimilarity LLC 806 55th St NE Seattle, WA 98105 +1 206.650.3740 http://biosimilarity.blogspot.com

2 1

Turning all the Nothings into Just defaultValue using Data.Generics
by David Fox 13 Nov '08

13 Nov '08

I want to use Data.Generics to write a function to turn all the Nothings in a data structure into Just defaultValue, as shown below. I get the following error because the compiler doesn't know enough about Maybe a for mkT to create the generic function that everywhere requires, I guess. Test.hs:26:16: Ambiguous type variable `a' in the constraints: `Typeable a' arising from a use of `mkT' at Senior/Test2.hs:26:16-30 `Default a' arising from a use of `justDefault' at Senior/Test2.hs:26:20-30 Probable fix: add a type signature that fixes these type variable(s) Here is the example. It all works except for "test". Any suggestions how to do this? {-# LANGUAGE DeriveDataTypeable, FlexibleContexts, FlexibleInstances, MultiParamTypeClasses, RankNTypes, TemplateHaskell, TypeSynonymInstances #-} {-# OPTIONS_GHC -fallow-undecidable-instances #-} module Test where import Data.Generics class Default a where defaultValue :: a instance Default Int where defaultValue = 0 instance Default String where defaultValue = "" instance Default (Maybe a) where defaultValue = Nothing data A = A {b :: Int, c :: Maybe String} deriving (Show, Data, Typeable) instance Default A where defaultValue = A {b = defaultValue, c = defaultValue} test = everywhere (mkT justDefault) (defaultValue :: A) where justDefault Nothing = Just defaultValue justDefault (Just x) = Just x

3 2

Linker errors to OpenGL with GHC 6.10.1
by Greg Fitzgerald 13 Nov '08

13 Nov '08

Do you know how I can fix these linker errors? C:\projects\fun>cat HelloWorld.hs import Graphics.Rendering.OpenGL import Graphics.UI.GLUT main = do (progname, _) <- getArgsAndInitialize createWindow "Hello World" displayCallback $= clear [ColorBuffer] mainLoop C:\projects\fun>ls lib GlU32.Lib glut32.lib OpenGL32.Lib glut.def glut32.dll C:\projects\fun>ghc -Llib -lglut32 -lglu32 -lopengl32 HelloWorld.hs --make Linking HelloWorld.exe ... C:\Program Files\Haskell\GLUT-2.1.1.2\ghc-6.10.1/libHSGLUT-2.1.1.2.a(Begin.o):fake:(.text+0x1cb): undefined reference to `glutGet' C:\Program Files\Haskell\GLUT-2.1.1.2\ghc-6.10.1/libHSGLUT-2.1.1.2.a(Begin.o):fake:(.text+0x8af): undefined reference to `glutMainLoop' C:\Program Files\Haskell\GLUT-2.1.1.2\ghc-6.10.1/libHSGLUT-2.1.1.2.a(Window.o):fake:(.text+0x3cd): undefined reference to `glutEntryFunc' C:\Program Files\Haskell\GLUT-2.1.1.2\ghc-6.10.1/libHSGLUT-2.1.1.2.a(Window.o):fake:(.text+0x40d): undefined reference to `glutVisibilityFunc' C:\Program Files\Haskell\GLUT-2.1.1.2\ghc-6.10.1/libHSGLUT-2.1.1.2.a(Window.o):fake:(.text+0x3595): undefined reference to `glutPassiveMotionFunc' ...hundreds more... Thanks, Greg

2 1

Re: [Haskell-cafe] Proof that Haskell is RT
by Andrew Birkett 12 Nov '08

12 Nov '08

Edsko de Vries wrote: > See "What is a purely functional language" by Sabry. Not quite a formal > proof about *Haskell*, but then we would first need a formal semantics > of Haskell to be able to do that proof ;-) Thanks for the reference, and also to everyone who replied - all very useful and interesting. For what it's worth, the blog posts I was writing are here: http://www.nobugs.org/blog/archives/2008/11/12/why-do-they-call-it-referent… http://www.nobugs.org/blog/archives/2008/11/12/why-do-they-call-it-referent… Andrew

2 1

ANNOUNCE: hpapi 0.1 release
by Michael D. Adams 12 Nov '08

12 Nov '08

I am pleased to announce the first release of hpapi, Performance API (PAPI) bindings for Haskell. The release is on Hackage: http://hackage.haskell.org/cgi-bin/hackage-scripts/package/hpapi The source repository is at: http://code.haskell.org/hpapi/ Please send bug reports to <adamsmd [AT] cs.indiana.edu> What is it ========== This is a simple set of bindings for the Performance API (PAPI). PAPI provides access to various CPU counters such as cache-miss, instruction and pipeline stall counts. The easiest way to use this library is the withCounts function: withCounts :: [EventCode] -> IO a -> IO (a, [Integer]) Given a list of events to count and an IO operation it will run the operation and return the resulting counts. So you could do something like: main = do [l1,tlb,fp] <- withCounters [papi_l1_dcm, papi_tlb_dm, papi_fp_ins] ioOperationToMeasure printLn (l1, tlb, fp) To use it one will of course need the PAPI C library and any necessary kernel modifications installed. Status ====== The High Level interface (which will be sufficient for most people's needs) should be usable, but the low level interface is still alpha quality. That is to say, it is ready for people to start banging on it, but it is largely untested and the modules/interfaces may be reorganized based on user feedback. All of PAPI is supported with the exception of the following "advanced" features: - Multiplexing - Threads - Locks - Overflow - Statistical Profiling Future Directions ================= Future directions will be determined largely by user feedback. Principal areas that I am looking for feedback on are: - documentation - module organization - the handling of PAPI errors (currently it throws an IO error) - the design of the Haskell interface build on top of the C level PAPI_get_opt and PAPI_set_opt - what advanced features of PAPI to build bindings for next Comparison with GHC's built-in PAPI =================================== GHC supports a built-in version of PAPI (see http://hackage.haskell.org/trac/ghc/wiki/PAPI) This system is different in a few ways. First, hpapi allows you to measure specific parts of a program while GHCI's PAPI only does whole program measurement Second, since hpapi is a library you can have fine grained and sophisticated control over how things are measured. On the other hand GHC's PAPI as part of the RTS gives you less control, but on the plus side doesn't require any modification to the program. Finally, GHC's PAPI splits apart the counts that come from the garbage collector from those that come from the mutator, while hpapi combines them. This is a limitation of hpapi we hope to correct some time in the future. It remains to be seen whether it would be worth while to unify these two systems. ================ Bug reports as well as suggestions, feedback or contributions to the API or the documentation most welcome. Michael D. Adams

1 0

GHC 6.10, strange behaviour when profiling?
by Aleš Bizjak 12 Nov '08

12 Nov '08

Hello, I'm experiencing some strange behaviour with GHC 6.10 and would like an explanation why. Here's the problem. With GHC 6.8.[23] memory usage of a program was about 250mb (computing pi to 10^6 decimals). Now I tried recompiling and running with GHC 6.10 and I got more than 1.4gb before I killed it, so naturally I tried profiling but here's the strangeness. When I compile for profiling with -prof -auto-all and run with +RTS -p -RTS memory use is still about the same as before (> 1.4gb), but when I try running with +RTS -hc -RTS memory only reaches about 250mb. What is the explanation for this behaviour? I would send a program but it requires HERA, which is not on hackage ... Here is the output of -s ./Test 100000 +RTS -s 47,467,176 bytes allocated in the heap 18,046,776 bytes copied during GC 9,215,104 bytes maximum residency (6 sample(s)) 711,520 bytes maximum slop 80 MB total memory in use (11 MB lost due to fragmentation) Generation 0: 84 collections, 0 parallel, 0.01s, 0.01s elapsed Generation 1: 6 collections, 0 parallel, 0.01s, 0.01s elapsed INIT time 0.00s ( 0.00s elapsed) MUT time 3.66s ( 3.71s elapsed) GC time 0.02s ( 0.02s elapsed) RP time 0.00s ( 0.00s elapsed) PROF time 0.00s ( 0.00s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 3.68s ( 3.73s elapsed) %GC time 0.5% (0.6% elapsed) Alloc rate 12,982,553 bytes per MUT second Productivity 99.5% of total user, 97.9% of total elapsed ./Test 100000 +RTS -s -hc 47,471,160 bytes allocated in the heap 18,637,752 bytes copied during GC 13,580,144 bytes maximum residency (37 sample(s)) 810,712 bytes maximum slop 33 MB total memory in use (7 MB lost due to fragmentation) Generation 0: 85 collections, 0 parallel, 0.02s, 0.02s elapsed Generation 1: 37 collections, 0 parallel, 0.00s, 0.01s elapsed INIT time 0.00s ( 0.00s elapsed) MUT time 3.67s ( 3.73s elapsed) GC time 0.02s ( 0.03s elapsed) RP time 0.00s ( 0.00s elapsed) PROF time 0.00s ( 0.00s elapsed) EXIT time 0.00s ( 0.00s elapsed) Total time 3.70s ( 3.77s elapsed) %GC time 0.6% (0.8% elapsed) Alloc rate 12,941,165 bytes per MUT second Productivity 99.2% of total user, 97.4% of total elapsed -- Best regards, Aleš

3 4

Searching for ADT patterns with elem and find
by Paul Keir 12 Nov '08

12 Nov '08

Hi All, If I have an ADT, say data T = A String Integer | B Double | C deriving(Eq) and I want to find if a list (ts) of type T contains an element of subtype "B Double", must my "containsTypeX" function use a second "isTypeX" function as follows: isTypeB :: T -> Bool isTypeB (B _) = True isTypeB _ = False containsTypeB :: [T] -> Bool containsTypeB ts = maybe False (\x -> True) (find isTypeB ts) I understand that while something like "find C ts" will work, "find (isTypeB _) ts" will not, but is there no such thing as a pattern combinator(?), or lambda that could help with this situation. I find I have many individual "isTypeB" functions now. Regards, Paul

6 9

abstract extensible types?
by Alberto G. Corona 12 Nov '08

12 Nov '08

Is there any abstract container that permits the addition of new types of data? I know how to simulate the extension of Algebraic datatypes, but this does not permit the addition of data with new types in the same container and recover them in a type-safe way. Did I reinvent the Weel? I found something, that permits this for any Typeable datatype. For example x=5 list= [put x, put "hello"] [t1,t2 ]= list x' = get t1 :: Int s = get t2 :: String print (x' +1) -- 2 print s -- "hello" x''= get t2 :: Int --"get: casting from String to type Int" The code: data Abstract= forall a. Typeable a => T !String a class FromToAbstract x where put :: x -> Abstract get :: Abstract -> x unsafeGet :: Abstract -> x -- get(put x)== x instance Typeable x => FromToAbstract x where put x= T (typeString x) x get (Abstract type1 a)= if type2 == type1 then v else error ("get: casting "++ "from type "++type1++" to type "++type2) where v= unsafeCoerce a :: x type2= typeString v unsafeGet (Abstract type1 a)= unsafeCoerce a typeString !x= tyConString $ typeRepTyCon $ typeOf x instance Typeable T where typeOf _= mkTyConApp (mkTyCon "Abstract") []

3 2