December 2005 - Haskell-Cafe

RE: [Haskell-cafe] Project postmortem
by Simon Marlow 02 Dec '05

02 Dec '05

On 02 December 2005 12:11, Tomasz Zielonka wrote: > On Fri, Dec 02, 2005 at 11:20:54AM -0000, Simon Marlow wrote: >> If anyone is interested, this turned out to be a bug in the >> Network.BSD module, namely that getHostByName isn't thread safe >> because it is based on the C library function gethostbyname(), which >> returns data in a single static area. >> >> Workarounds are: >> >> - do your own mutual exclusion locking around getHostByName and any >> function that calls it (eg. connectTo). >> >> - use Network.Alt >> (http://www.cs.helsinki.fi/u/ekarttun/network-alt/) which has a >> thread-safe implementation of getHostByName. >> >> - wait for 6.4.2, which will contain a fix for this bug (we don't >> have a fix committed yet, Einar Karttunen has kindly offered to >> look into it). > > Do I understand correctly that another workaround is > - don't compile your programs with -threaded > ? No, the bug isn't related to -threaded. It still occurs without -threaded. Cheers, Simon

2 1

RE: [Haskell-cafe] Project postmortem
by Simon Marlow 02 Dec '05

02 Dec '05

On 21 November 2005 10:16, Joel Reymont wrote: > Is Wolfgang still around? > > Would you guys be willing to guide me through this? I could then > possibly become the next Mac OSX expert :-). > > I have the disassembler dumps, etc. I do not know how to approach > this problem. I read up a bit on the GHC internals, STG, code > generation, etc. If anyone is interested, this turned out to be a bug in the Network.BSD module, namely that getHostByName isn't thread safe because it is based on the C library function gethostbyname(), which returns data in a single static area. Workarounds are: - do your own mutual exclusion locking around getHostByName and any function that calls it (eg. connectTo). - use Network.Alt (http://www.cs.helsinki.fi/u/ekarttun/network-alt/) which has a thread-safe implementation of getHostByName. - wait for 6.4.2, which will contain a fix for this bug (we don't have a fix committed yet, Einar Karttunen has kindly offered to look into it). Cheers, Simon

3 2

RE: Re[2]: [Haskell-cafe] STM, orElse and timed read from a channel
by Simon Marlow 02 Dec '05

02 Dec '05

On 01 December 2005 13:27, Bulat Ziganshin wrote: > Hello Simon, > > Wednesday, November 30, 2005, 3:26:44 PM, you wrote: > >> It's a modification of GHC.Conc, so I can't supply a standalone >> implementation I'm afraid. > >>> I could use Tomasz's workaround in the meantime but isn't GHC 6.6 >>> supposed to be out in a year? > >> Rather less than a year, I hope. > > Joel, you can download GHC 6.6 beta builds (from > http://www.haskell.org/ghc/dist/head/dist ?), so you only > need to wait until Simon actually merges this update to the main code http://www.haskell.org/ghc/dist/current/dist/ and registerTimeout is already in. Cheers, Simon

1 0

Problem with continuations and typing
by oleg＠pobox.com 01 Dec '05

01 Dec '05

Jerzy Karczmarczuk wrote: > I tried to cook-up a simple version of nondeterministic computations > using the two-continuation model. I got stuck on an 'infinite type' > error, and my question is: have you seen a *concrete* (really > concrete) Haskell implementation of a similar model, I could inspire > myself on? (I know a few theoretical works on that). Yes, of course: the code for the LogicT paper http://pobox.com/~oleg/ftp/packages/LogicT.tar.gz (please see SFKT.hs) and http://www.haskell.org/pipermail/haskell/2005-October/016577.html http://www.haskell.org/pipermail/haskell/2005-October/016696.html As we can see, FR stream indeed takes success and the failure continuations. This is emphasized at the beginning of the first article. > But I would like to continue this exercice along these lines, without > too much exotism (no monads, yet...), for my students. Do you have any > simple work-around Introduce some algebraic constructors? Perhaps > higher-rank polymorphism could do something (but then I would have to > explain it to my folk...) You're correct on both counts. We can emulate equi-recursive types with iso-recursive ones: e.g., plain Haskell list. Or, you may choose to use higher-ranked type (the FR stream). There is a third solution: use a small typing hole in Haskell to sneak in the real equi-recursive type. I have a feeling though you might not wish to explain the latter method to your students. This present question is actually related to your previous inquiry about expressing a predecessor in a simply-typed lambda-calculus. It cannot be expressed. We need either a recursive or higher-ranked type: predecessor is 'car'.

1 0

returning lists
by raptor 01 Dec '05

01 Dec '05

hi, i'm having this problem, tring to implement split (for learning purposes). Here is one sample of my tries : split _ [] = [] split char (x:xs) | x /= char = x : split char xs | otherwise = [] it is ok i.e. it returns the first of the splits, but I want to get all of the results. str = "testingxsplitxtoxseexhowxitxwork" *Main> split 'x' str "testing" I want : ["testing","split",...] How do i do this.

3 4

RE: [Haskell-cafe] STM, orElse and timed read from a channel
by Simon Marlow 01 Dec '05

01 Dec '05

On 29 November 2005 13:01, Joel Reymont wrote: > Would you share your implementation for us advanced users? It's a modification of GHC.Conc, so I can't supply a standalone implementation I'm afraid. > I could use Tomasz's workaround in the meantime but isn't GHC 6.6 > supposed to be out in a year? Rather less than a year, I hope. Cheers, Simon

2 1

:t main
by Scherrer, Chad 01 Dec '05

01 Dec '05

I've been reading some of the articles about comonads, and I thought the idea of giving main the type OI () -> () was pretty interesting. So I was wondering, would it be possible to allow the type of main to be inferred? It seems like IO () OI () -> () OI () -> IO () all make sense (at least I think they do). One particularly nice side effect of this (pardon the pun) is that a good number of useful programs can be written as getContentsW :: OI () -> a doSomeStuff :: (Show b) => a -> b print :: (Show b) => b -> IO () main = print . doSomeStuff . getContentsW so much less understanding about monads et al is required to do some basic stuff. I think it would be much easier in this case to make the transition to Haskell. Also, has anyone given any thought to syntactic sugar for comonads? Since arrows and monads share the (<-) symbol, it seems like a nice unification could be possible... Maybe (-<) alone could be used similarly? Thanks, Chad Scherrer

2 1

RE: [Haskell-cafe] Detecting Cycles in Datastructures
by Simon Peyton-Jones 01 Dec '05

01 Dec '05

| Andrew Pimlott: | //about my highly spiritual essay on lazy computing of PI//: | | > In addition to being clever and terribly funny, the conclusion | > foreshadows (inspired?) later work on Enron [1]. | | Come on, it is improbable that Master Simon ever read my essay... I hadn't before, but I have now. Cool stuff. I once wrote a continued-fraction package with the something of the same flavour (since overtaken by David Lester's work, which to my deep sadness remains unpublished to this day). Your code clearly belongs in the nofib suite, as a splendid way to consume compute cycles and test compilers. Do send me a Main.hs to compute pi, with a changeable runtime (presumably just adjust 'take'). Simon

2 1

first post
by raptor 30 Nov '05

30 Nov '05

hi, I'm trying to figure out how to parse a text file, but these type-conversations are just killing me! Whats wrong with following code : module Main where main = putStrLn "hi" parse str = take 10 str load = do content <- readFile "parse.hs" return content when try : Prelude> parse load <interactive>:1:6: Couldn't match `[a]' against `IO String' Expected type: [a] Inferred type: IO String In the first argument of `parse', namely `load' In the definition of `it': it = parse load i tried different combinations, but still cant figure out... How do u use an "IO String" in normal "String" functions.. I want to particulary slurp file in a string and then make funny things with it...spliting to lines,parsing them and building hash like strucutre of it...etc. later i want to try spliting,regex-es..etc. It is very hard to figure out how to use these modules.. :"(

3 3

first post
by Chad Scherrer 30 Nov '05

30 Nov '05

Because the load function interacts with the outside world, its type involves the IO monad. You can get the result you're looking for using something like liftM parse load liftM takes an ordinary function and applies it to a monadic value (here the monad is IO), returning another monadic value. Or, if you're using ghci, you could just do s <- load parse s This is because ghci executes "in the IO monad" (maybe there are some details of this I'm missing, but that's how I think of it anyway. BTW, your load function could be equivalently written simply as load = readFile "parse.hs" -Chad Scherrer

1 0