October 2007 - Haskell-Cafe

binary operator modifiers
by Tim Newsham 19 Nov '07

19 Nov '07

I would love to have the ability to define binary operator modifiers. For example: f \overline{op} g = liftM2 op f g f \overleftarrow{op} n = liftM2 op f (return n) n \overrightarrow{op} g = liftM2 op (return n) f \widehat{f} x = liftM f x so that for example you could define functions like: (*3) \overline{+} (/2) and 3 \overrightarrow{+} \widehat{read} getContents Obviously you could write this out the long way: liftM2 (3+) $ liftM read getContents or go through the trouble of defining a bunch of binops f <+> g = liftM2 (+) f g n +> g = return n <+> g f <+ n = f <+> return n read' = liftM read (*3) <+> (/2) 3 +> read' getContents but doing this for more than one or two operators gets tedious quickly... Is there any way in Haskell to modify binops in this way and still be able to use them infix? If not, has anyone considered supporting strange syntaxes like this? Tim Newsham http://www.thenewsh.com/~newsham/

4 5

[Haskell-cafe] Letting the darcs test fail, if QuickCheck tests fail
by Henning Thielemann 18 Nov '07

18 Nov '07

When following the description on http://www.haskell.org/haskellwiki/How_to_write_a_Haskell_program#Add_some_… then darcs will run the QuickCheck tests on each 'darcs record', but the new patch is also accepted by darcs if one of the tests fail. What is the most simple way to let 'darcs record' fail, when a QuickCheck test fails?

3 3

Weird ghci behaviour?
by Dan Piponi 15 Nov '07

15 Nov '07

On Unix-like OSes: If I run "ghc test.hs" and then run "ghci test.hs", ghci fails to load up my code. I have to touch test.hs and then run ghci. I can understand ghci refusing to recompile something it thinks it has already compiled. But it appears to refuse to load it into an interactive session - which is less useful. In fact, removing test.hi makes ghci work again. This is ghc 6.6. Anyone else seeing this? -- Dan

18 29

Using type-level programming to tag functions with time and space complexity
by Daniel McAllansmith 13 Nov '07

13 Nov '07

I was wondering if anyone had done work on tagging functions at the type level with their time or space complexity and, if it's even feasible, calculating the complexity of compound functions. Any pointers? Cheers Daniel

3 2

ANN: FileManip 0.1, an expressive file manipulation library
by Bryan O'Sullivan 09 Nov '07

09 Nov '07

The FileManip package provides expressive functions and combinators for searching, matching, and manipulating files. It provides three modules. System.FilePath.Find lets you search a filesystem hierarchy efficiently: find always (extension ==? ".rb") >>= mapM_ remove System.FilePath.GlobPattern lets you perform glob-style pattern matching, without going through a regexp engine: "foo.c" ~~ "*.c" ==> True System.FilePath.Manip lets you rename files procedurally, edit files in place, or save old copies as backups: modifyWithBackup (<.> "bak") (unlines . map (takeWhile (/= ',')) . lines) "myPoorFile.csv"

5 11

speeding up fibonacci with memoizing
by Thomas Hartman 06 Nov '07

06 Nov '07

I just thought this was interesting, so I would share it. Thanks to whoever it was on #haskell who helped me, sorry I can't remember who. -- horribly slow. try slow_fibs 30, not too much higher than that and it hangs slow_fibs = map slow_fib [1..] slow_fib 1 = 1 slow_fib 2 = 1 slow_fib n = ( slow_fib (n-2) ) + (slow_fib(n-1)) -- versus, try memoized_fibs !! 10000 memoized_fibs = map memoized_fib [1..] memoized_fib = ((map fib' [0 ..]) !!) where fib' 0 = 0 fib' 1 = 1 fib' n = memoized_fib (n - 1) + memoized_fib (n - 2)

14 16

Adding GLUT package to WinHugs
by Peter Verswyvelen 04 Nov '07

04 Nov '07

The precompiled WinHugs Sep 2006 does not seem to come with the GLUT package. Although I'm able to add packages to GHC (using Cabal), I failed doing the same for WinHugs. How can I add the GLUT package to WinHugs? I tried starting MSYS, typed configure, which worked fine, but then "make" fails. Also typing runhugs Setup.hs configure fails with runhugs: Error occurred ERROR "c:\program files\winhugs\packages\base\Text\ParserCombinators\ReadP.hs":156 - Syntax error in type expression (unexpected `.') Thanks for any help, Peter Verswyvelen

2 4

Re: Why can't Haskell be faster?
by Jeff.Harper＠handheld.com 02 Nov '07

02 Nov '07

Peter Hercek wrote: > * it is easy to mark stuff strict (even in function signatures > etc), so it is possible to save on unnecessary CAF creations Also, the Clean compiler has a strictness analyzer. The compiler will analyze code and find many (but not all) cases where a function argument can be made strict without changing the behavior of the program.

28 47

Hiding side effects in a data structure
by C Rodrigues 02 Nov '07

02 Nov '07

While thinking about how to generate unique integer IDs on demand without using a state variable, I came up with an interesting design pattern. It's a way of doing side-effecting computation outside IO. Referential transparency is preserved by making the side effects spatial rather than temporal: by hiding side effects behind lazy thunks in a data structure, they can be disguised as the output of a single, apparently nondeterministic IO function used to data structure. This lets pure code use nondeterministic computation without the monadic plumbing required to maintain state. The getContents function works this way, but I came up with a more interesting example. The code below is a source of unique integer IDs that is modeled after the RandomGen class. It uses unsafeInterleaveIO under the hood, preserving referential transparency but not determinism. It seems to work. However, I'm not entirely sure how safe my use of unsafeInterleaveIO is. In particular, could the two branches of the tree get CSE'd? I'm also curious what people think about the general design pattern. module Unique where import Control.Monad(liftM) import Data.IORef import System.IO.Unsafe -- The goal is to produce an infinite tree of integers where each node in the -- tree has a unique value. type Unique = Int data Supply = Supply Unique Supply Supply -- The tree can be used in a stateful manner as a source of unique integers. getUnique :: Supply -> (Unique, Supply) getUnique (Supply u s1 _) = (u, s1) -- The tree can also be split into independent sources of unique integers. split :: Supply -> (Supply, Supply) split (Supply _ s1 s2) = (s1, s2) -- The catch is, the tree will probably be visited very sparsely, with most of -- it being skipped. Assigning every node its own integer is very bad, because -- that will waste most of the 2^32 available integers very quickly. In fact, -- it can get used up in just 32 calls to getUnique. -- -- Instead, we'll create a tree where integers magically appear only in places -- where they are actually used. -- First, we need an IO-bound supply of integers. newtype IOSupply = IOSupply (IORef Unique) newIOSupply :: IO IOSupply newIOSupply = liftM IOSupply $ newIORef 0 getUniqueIO :: IOSupply -> IO Unique getUniqueIO (IOSupply s) = do u <- readIORef s writeIORef s $ u+1 return u -- Now we'll use the IO-bound supply to create a tree having the desired -- properties. {-# NOINLINE getPureSupply #-} getPureSupply :: IOSupply -> IO Supply getPureSupply s = do s1 <- unsafeInterleaveIO $ getPureSupply s s2 <- unsafeInterleaveIO $ getPureSupply s n <- unsafeInterleaveIO $ getUniqueIO s return $ Supply n s1 s2 _________________________________________________________________ Climb to the top of the charts! Play Star Shuffle: the word scramble challenge with star power. http://club.live.com/star_shuffle.aspx?icid=starshuffle_wlmailtextlink_oct

14 21

Embedding the GHC API
by Joel Reymont 01 Nov '07

01 Nov '07

Has anyone tried to embed GHC as a library recently? What is the size of the resulting binary? I'm assuming a bare minimum of needed libraries. Thanks, Joel -- http://wagerlabs.com

4 3