- Haskell-Cafe - Haskell.org

How can I get the mutable array out of an IOUArray for FFI use?
by Ryan Ingram 02 Aug '08

02 Aug '08

I wrote some fast bit-twiddling functions in C because my Haskell performance wasn't good enough. Now I'm trying to recompile with GHC6.8.3 and failing. This code worked on GHC6.6.1. I get the following error: > ghc --make main.hs Bitmap.hs:11:7: Could not find module `Data.Array.IO.Internals': it is hidden (in package array-0.1.0.0) I suppose I can declare a copy of the internal type and use unsafeCoerce#, but that seems like a terrible idea if there is a better way. What's the right way to make this work? Can I force that module to be unhidden? Should I file a GHC bug? -- ryan {-# OPTIONS_GHC -fffi -fglasgow-exts #-} {-# INCLUDE "bitmap_operations.h" #-} module Bitmap ( clearBitmap, ) where import Foreign.Ptr import Data.Array.Base import Data.Array.IO.Internals import GHC.Exts import Data.Word foreign import ccall unsafe clear_bitmap :: MutableByteArray# RealWorld -> Word32 -> Word32 -> IO () {-# INLINE unsafeGetMutableArray# #-} unsafeGetMutableArray# :: IOUArray Int Word32 -> MutableByteArray# RealWorld unsafeGetMutableArray# (IOUArray (STUArray _ _ array#)) = array# clearBitmap :: IOUArray Int Word32 -> Word32 -> Word32 -> IO () clearBitmap a1 color size = clear_bitmap (unsafeGetMutableArray# a1) color size

2 1

Using fundeps to resolve polymorphic types to concrete types
by Bryan Donlan 02 Aug '08

02 Aug '08

Hi, Is there any theoretical reason that functional dependencies can't be used to resolve a polymorphic type to a concrete type? For example: > -- compile with -fglasgow-exts > > class DeriveType a b | a -> b > > data A = A > data B = B > > instance DeriveType A B > > simpleNarrow :: DeriveType A b => b -> B > simpleNarrow = id Since 'b' is uniquely determined by the fundep in DeriveType, it seems that this ought to work; ie, since the only type equation satisfying DeriveType A b is B -> B, it should reduce to that before trying to fit its type against its body. The motivation is this case: > data ComplexType a where > SomeConstructor :: DeriveType a b => a -> b -> ComplexType a > > specialCaseFunc :: ComplexType A -> B > specialCaseFunc (SomeConstructor _ b) = b Essentially, if I have a data structure with two types used as fields, and one uniquely determines the other, I'd like to use these instances to avoid re-stating the implied one in the type equations, if possible. Is there some theoretical reason for this not to work, or is it just a limitation of GHC's current implementation? (Note, I'm testing with GHC 6.8.2, so it's possible this might be fixed in trunk already...) Thanks, Bryan Donlan

5 9

question about "faulting in" Haskell packages ...
by Galchin, Vasili 02 Aug '08

02 Aug '08

Hello, Sorry I asked this question before .. say I want to install package A that depends on package B, C, D but maybe B and D are not installed on my machine. Basically I want to "fault in" (i.e. install automatically) in B and D with "no fuss no muss". ??? In reality I would like to get my laptop which is running Ubuntu totally "faulted up" with all Haskell packages ... ;^) Regards, Vasili

1 0

Cabal files on Windows
by John Lato 01 Aug '08

01 Aug '08

Hello, Can anyone point me to a method for including path names with spaces in a cabal file? I would like to add a line similar to the following: include-dirs: C:\Program Files\program\include and of course a corresponding library as well. I've tried various methods of escaping the space and quoting, and nothing seems to work. Is this possible at all? I'm using Cabal-1.2.4.0 on Windows. I've tried running with a Windows shell and MSYS bash, with the same results. Thanks, John Lato

4 8

Analysing Haskell Program
by Roberto D'Aprile 01 Aug '08

01 Aug '08

Hello to everybody I'm using haskell for some research work and now i need to evaluate the performance of some simple haskell programs in particular information on the like, CPU cycles, bus usage, memory access and so on; so i wish to know if there is a way to compile haskell programms so they can be used with simplescalar or if there is another tool which i can use. Thank you

3 2

category theory tutorial pdfs .....
by Galchin, Vasili 01 Aug '08

01 Aug '08

Hello, Prof. Harold Simmons' tutorial IMO are like a Russian matroshka doll ... first layer is for newbie ... inner layers require more sophistication. IMO a very subtle writer ... I have every book imaginable on cat theory and topos theory so I think can compare a little. 1) http://www.cs.man.ac.uk/~hsimmons/BOOKS/books.html ... an earlier version ... "An Intro to Category Theory in Four Easy Movements" ... this version delves a little into Topos Theory ... pretty subtle .. cool 2) http://www.cs.man.ac.uk/~hsimmons/MAGIC-CATS/magic-cats.html ... newer version which also has its cool merits IMO! Regards, Vasili

2 1

a really dumb posting question ;^(
by Galchin, Vasili 01 Aug '08

01 Aug '08

Hello, What do I do to do a followup haskell cafe posting? E.g. I want to put a posting on the category theory thread! Kind regards, Vasili

3 2

an array of pointers in FFI?
by Galchin, Vasili 01 Aug '08

01 Aug '08

Hello, Is a (Ptr (Ptr a)) a polymorphic representation of an array of pointers of type "a"? I want to pass an array of pointers to a C function. Kind regards, Vasili

3 3

Point-free style in guards
by L29Ah 31 Jul '08

31 Jul '08

outStanza | (isMessage) = outMessage | (isPresence) = outPresence | (isIQ) = outIQ Why such a style doesn't work, so I must write ugly code like that: outStanza a | (isMessage a) = outMessage a | (isPresence a) = outPresence a | (isIQ a) = outIQ a so, guards can't be useful in point-free function definitions in any way

8 7

array fusion interface
by Evan Laforge 31 Jul '08

31 Jul '08

With all the noise lately about high performance array libraries with list-like interfaces, such as bytestring, storablevector, uvector, and vector, I thought I would try to make use of one in a project of mine, and I'm either bumping up against the limits of its expressiveness, or am missing out on how to express my problem. I have streams of samples with irregular sampling rates, so they look like [(Time, SampleVal)]. In practice, this means [(Double, Double)]. I can do this with storablevector by either storing two Vector Double, or by making a (Double, Double) Storable instance (more natural but requires -XFleixbleInstances). Now most functions that operate over a sample stream actually want a pair of samples so they can get an exact time with linear interpolation. Combining two streams requires them to both be resampled so their points line up, e.g. with a list implementation (ignoring the edges): resample as@((ax0, ay0) : (ax1, ay1) : _) bs@((bx0, by0) : (bx1, by1) : _) | ax1 == bx1 = (ax1, (ay1, by1)) : resample (tail as) (tail bs) | ax1 < bx1 = (ax1, (ay1, y_at (bx0, by0) (bx1, by1) ax1)) : resample (tail as) bs | otherwise = (bx1, (y_at (ax0, ay0) (ax1, ay1) bx1, by1)) : resample as (tail bs) y_at (x0, y0) (x1, y1) x = (y1 - y0) / (x1 - x0) * (x - x0) + y0 So there are a number of things here that don't seem to work so well with a vector interface. The first is that this returns [(Val, (Val, Val))]. I could either write an instance of Storable for (Double, (Double, Double)), or I could pass a function (Val, Val) -> Val. It looks like uvector handles zips more nicely though. The second is that this, like most such functions, wants a pair of samples to do linear interpolation. With lists I can do pattern matching as above or 'zip xs (drop 1 xs)', and with vectors the slightly less natural mapAccumL. Ok, then the third problem is that this progresses at a variable rate down two vectors, and I don't know any way to express that. Also, as a fourth problem, many functions taking samples will wind up generating more samples, which requires a concatMap kind of thing, which uvector doesn't seem to have, but in storablevector and lazy bytestring involves converting the whole thing to a list and back. What I'd ideally like is a lazy stream of unpacked chunks ala ByteString.Lazy. It seems like it might be possible to do something with ST and mutable arrays, and somehow come up with an abstraction that hides the packing and chunks and freezing but is still flexible enough to write something like 'resample', but I haven't thought about what that would look like. And it would be much nicer to try to fit something into the array fusion kind of interface. Is it possible? Is concatMap and "complicated zip" incompatible with the listlike array interface and fusion? Is there a better way to express this that avoids the problems? Initially this seemed like the perfect application for storablevector or uvector, but maybe it's not actually.

2 1