November 2007 - Haskell-Cafe

ANN: infinity 0.3
by Austin Seipp 07 Nov '07

07 Nov '07

After quite a bit of work and other such stuff, I pretty much totally rewrote my previous IRC bot and instead updated it from v0.1 to v0.3. You may wonder where 0.2 went; so do I. This is basically an entire rewrite (I threw the first one out.) Now all code can be dynamically hot-swapped via the techniques described in Don's Yi paper. This can happen across multiple servers and as servers are taken down and the like. All code swaps near flawlessly, and even after the initial mechanism was working, the bot ran for quite a while with a good amount of success. I didn't plan to release it in this state, however, with it's dependencies on hs-plugins (which is ghc 6.6 only as of right now,) and my recent installation of GHC 6.8.1, I can't work on it because apparently although both versions appear to be installed, it seems as if ghc-6.6.1 is picking up configuration and package info from 6.8.1 (and therefore ghc-6.6.1 tells me it can't find packages like irc and plugins, although 'ghc-pkg-6.6.1 list' obviously shows they exist.) I also can't work on this on my OpenBSD box which only has 6.6.1, due to some arcane error with hs-plugins (can't find the __errno symbol, I believe.) So because of that, development is temporarily stalled and if you want to test it, you'll need 6.6.1 installed and working properly. I don't think this would have been *too* far from the actual, official release though. At this point I only planned minor tweaks and its release was pretty inevitable anyway. You can get a tarball from hackage: http://hackage.haskell.org/cgi-bin/hackage-scripts/package/infinity-0.3 The darcs repository is here: http://code.haskell.org/infinity/src/ Personally, it was real fun to work on this thing. I wrote almost all the base code for plugins, and for the Net and Monitor to use before the bot ever made its first connection, and a lot of it seemed to just piece together really easily. The code is undeniably cleaner and more extensible than the previous 0.1 code, but I still feel by most standards it's somewhat cluttered, so if anybody here would like to give any comments just from reading the source, that'd be pretty awesome. Until I can continue to work on it (whether hs-plugins is available for 6.8.1 or I get another box with 6.6.1 installed,) this will be considered the stable version by my definition, and all patches are of course welcome. darcs send austin at youareinferior dot net. Criticism is advised. -- - Austin

1 0

hoogle broken?
by Michael Vanier 07 Nov '07

07 Nov '07

It looks as if hoogle isn't working. I get 404s whenever I try to do any search on hoogle. Mike

4 4

Haskell home page not updated
by Paul Johnson 07 Nov '07

07 Nov '07

The Haskell home page hasn't been updated since 23rd September, even though a "Haskell Weekly News" came out on October 25th. Paul.

2 1

Problem linking with GHC 6.8.1
by david48 07 Nov '07

07 Nov '07

This is a program that works under ghci. GHC was installed using the tarball on hashell.org. When I try to compile the program, it can't link. It used to work with 6.6.1. ghc --version The Glorious Glasgow Haskell Compilation System, version 6.8.1 ghc --make -O2 -o edimail Main.hs [1 of 5] Compiling Decode ( Decode.hs, Decode.o ) [2 of 5] Compiling Net ( Net.hs, Net.o ) [3 of 5] Compiling Smtp ( Smtp.hs, Smtp.o ) [4 of 5] Compiling Pop3 ( Pop3.hs, Pop3.o ) [5 of 5] Compiling Main ( Main.hs, Main.o ) Linking edimail ... /usr/bin/ld: cannot find -lgmp collect2: ld a retourné 1 code d'état d'exécution

3 6

GHC 6.8 and Mac OS X Leopard
by Tim Clark 07 Nov '07

07 Nov '07

Has anyone got GHC 6.8.1 building on Leopard? Unfortunately macports isn't even letting me build GHC 6.6.1 since the bootstrap GHC isn't working. Does anyone have any ideas or am I just too impatient? Thanks, Tim Clark

2 2

Monte Carlo Pi calculation (newbie learnings)
by Alex Young 06 Nov '07

06 Nov '07

Hi all, I'm new to Haskell, and don't quite understand how IO and lazy evaluation work together yet. In order to improve my understanding, I thought I'd try to knock together a translation of a fairly simple problem I can code up in C, that of calculating pi by throwing random numbers about. The C code I started with is as follows: ////////////////////////////////////////////////// #include <stdio.h> #include <stdlib.h> #include <time.h> int main(int argc, char *argv[]) { if(argc != 2) { printf("usage: %s iteration_count", argv[0]); return 1; } int total; if(sscanf(argv[1], "%d", &total) != 1) return 1; int count = 0; int unit_radius = RAND_MAX * RAND_MAX; int i; srand(time(NULL)); for(i = 0; i < total; i++) { int x, y; x = rand(); y = rand(); count += x*x + y*y < unit_radius; } double pi = (count << 2) / (double)total; printf("Count: %d\nTotal: %d\n", count, total); printf("%f", pi); return 0; } /////////////////////////////////////////////////// All very simple - the key is that I'm counting the result of a conditional evaluated inside a loop. Ignore whether it's correct, or accurate, for now; it happily runs and gives an accurate-looking result when run with an argument of 10000000. My (thoroughly ugly, not currently working) Haskell version looks like this: {--------------------------------------------------} module Main where import Random import System.Environment import List import Monad randMax = 32767 unitRadius = randMax * randMax rand :: IO Int rand = getStdRandom (randomR (0, randMax)) randListTail accum 0 = accum randListTail accum n = randListTail (rand : accum) (n - 1) randList :: Int -> [IO Int] randList n = randListTail [] n randPairs :: Int -> [(IO Int, IO Int)] randPairs n = zip (randList n) (randList n) pairIsInside x y = if x*x + y*y < unitRadius then 1 else 0 doCountPair :: (IO Int, IO Int) -> IO Int doCountPair (a, b) = do x <- a y <- b return (pairIsInside x y) fSumListTail :: Int -> [(IO Int, IO Int)] -> IO Int fSumListTail total [] = do return total fSumListTail total (x:xs) = do y <- doCountPair x fSumListTail (total+y) xs fSumList :: [(IO Int, IO Int)] -> IO Int fSumList l = fSumListTail 0 l piAsDouble :: Int -> Int -> Double piAsDouble a b = (fromInteger (toInteger a)) / (fromInteger (toInteger b)) calculatePi total = do count <- fSumList (randPairs total) return (piAsDouble (4*count) total) main = do args <- getArgs (calculatePi (read (args !! 0))) >>= (putStr . show) {--------------------------------------------------} Now, I have two questions. The easy question is, how can I make this more idiomatic? I seem to be jumping through hoops rather a lot to achieve what should be rather simple operations. The piAsDouble and fSumListTail functions are perfect examples, but I'm not wildly keen on doCountPair either. The second question is performance-related. The Haskell code above overflows the stack when run with an argument of 1000000, so either somewhere a list I'm intending to be lazily evaluated and chucked away is being retained, or one of my recursive functions isn't tail-optimising. Is it obvious to a more trained eye where the problem is? If not, how should I start to diagnose this? I'm not really sure where I should look for more information. More information, if it's relevant: Haskell: Glasgow Haskell Compiler, Version 6.8.0.20070917, for Haskell 98, stage 3 booted by GHC version 6.6 OS: Windows Vista, 32bit Any guidance gratefully appreciated... Thanks, -- Alex

12 18

Re: [Haskell-cafe] Monte Carlo Pi calculation (newbie learnings)
by Yitzchak Gale 06 Nov '07

06 Nov '07

Hi Alex, > I think I would have avoided most of the problems > in the first place. I'm finding the documentation > to be rather undiscoverable at the moment, but I'm > sure it's just a question of learning my way around it. Make sure to keep http://haskell.org/ghc/docs/latest/html/libraries/ on your favorites list. >> Represent >> your big iteration as a list. Try to stick >> to fairly simple operations on the list. Then the >> compiler will figure out how to keep things >> lazy and also release memory as it goes along. > See, this is what I thought I was doing, but I > couldn't see where to simplify it further. Sorry, you're right, I was begging the question. Hmm. Well, for starters: o Try to keep your big list far away from the IO monad. o Whenever possible, use the following library functions rather than manual recursion for iterating over your big list: take, takeWhile, drop, dropWhile, map, concatMap, filter, zip, zipWith. Browse around the wiki in places like: http://haskell.org/haskellwiki/Stack_overflow http://haskell.org/haskellwiki/Performance/Laziness http://haskell.org/haskellwiki/Performance/Strictness http://haskell.org/haskellwiki/Category:Style http://haskell.org/haskellwiki/Category:Tutorials -Yitz

1 0

Re: Best Linux for Haskell?
by Ivan Lazar Miljenovic 06 Nov '07

06 Nov '07

Seth Gordon wrote: > Peter Verswyvelen wrote: > > Each year I give Linux a try. And usually I kick it off my harddrive > > after a month, and stick to Windows. However, it does get better each > > year, so > > > > > > > > So which kind Linux works best for running GHC (6.8.1) and related > > tools? (I want to give Yi a go, I cant get it to work on Windows). > > Debian? Fedora? Ubuntu? > > I don't know which Linux distribution is best *for Haskell*, but in my > experience, Ubuntu has been the best distribution for getting up and > running on Linux without fuss. The latest version lets you boot into > Linux from a CD-ROM, so you can try it out and confirm that all your > hardware is compatible before you touch anything on your hard drive. > > I haven't tried 6.8.1 yet, but I have downloaded and used the 6.6.x > binaries on an Ubuntu system back when the repositories were still using > an earlier version. I strongly suspect that if you got hold of 6.8.1 > packages that were built for Debian, they would also install on Ubuntu, > but I haven't tried doing that yet. Well, we're testing out ghc-6.8.1 on the Gentoo-Haskell overlay right now... but it's probably not as good a distribution for a newbie :p -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com

1 0

ghc -e with standard input
by Maurício 06 Nov '07

06 Nov '07

Hi, Is there a way to run 'ghc -e' taking input from standard input? I would like to use it in a pipe. Thanks, Maurício

6 10

Go parallel
by Andrew Coppin 06 Nov '07

06 Nov '07

Hi folks. Take a look at this: render :: IOArray Point Colour -> (Point -> Colour) -> IO () render framebuffer fn = mapM_ (\p -> writeArray framebuffer p (fn p)) all_points How do I alter this to compute fn in multiple threads in parallel? (As you can see from the type signature, the calculation is independent for each pixel.) Thanks.

7 13