May 2007 - Haskell-Cafe

Intermediate Haskell Books?
by Adrian Neumann 07 May '07

07 May '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: RIPEMD160 Are there any good books about intermediate to advanced Haskell? The descriptions here http://haskell.org/haskellwiki/Books_and_tutorials aren't very helpful. Adrian -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGPZCo11V8mqIQMRsRA8nVAJ9fs4c013qKOuEjX//uLW3hLTYtxACfTS/t 5D6I+/OGifs3ltYhx6rwFZA= =sNJP -----END PGP SIGNATURE-----

5 4

c2hs errors when compiling hsGnuTls
by David House 07 May '07

07 May '07

Hey there, I'm getting the following errors when I try to compile hsGnuTls [1]: ~/hs/sandbox/hsgnutls $ c2hs --version C->Haskell Compiler, version 0.14.5 Travelling Lightly, 12 Dec 2005 build platform is "i486-pc-linux-gnu" <1, True, True, 1> ~/hs/sandbox/hsgnutls $ runhaskell Setup.lhs build Setup.lhs: Warning: The field "hs-source-dir" is deprecated, please use hs-source-dirs. Preprocessing library hsgnutls-0.2.3... c2hs: Error in C header file. /usr/include/bits/pthreadtypes.h:69: (column 6) [FATAL] >>> Syntax error! The symbol `;' does not fit here. Setup.lhs: got error code while preprocessing: Network.GnuTLS.GnuTLS c2hs version: I've attached the file it references in case that's relevant. Any tips on how I might address this? Thanks in advance, -David House, dmhouse(a)gmail.com

2 2

The Monad.Reader - Call for copy
by Wouter Swierstra 07 May '07

07 May '07

Call for Copy The Monad.Reader - Issue 8 I would like to welcome articles for the next issue of The Monad.Reader. * The Monad.Reader * The Monad.Reader is a electronic magazine about all things Haskell. It is less-formal than journal, but somehow more enduring than a wiki- page. There have been a wide variety of articles, including: exciting code fragments, intriguing puzzles, book reviews, tutorials, and even half-baked research ideas. * Submission Details * Get in touch with me if you intend to submit something -- the sooner you let me know what you're up to, the better. Please submit articles for the next issue by e-mail (wss at cs.nott.ac.uk) to me before ** July 27, 2007 ** Articles should be written according to the guidelines available from http://www.haskell.org/haskellwiki/The_Monad.Reader Please submit your article in PDF, together with any source files you used. The sources will be released together with the magazine under a BSD license. If you would like to submit an article, but have trouble with LaTeX please let me know and we'll sort something out. Looking forward to your submission, Wouter This message has been checked for viruses but the contents of an attachment may still contain software viruses, which could damage your computer system: you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation.

1 0

Kleene star in types
by Matthew Sackman 07 May '07

07 May '07

Apologies for sending this twice. Mutt obviously outsmarted mailman and put a useful mime type header in. Not so this one! This email is a .lhs. > {-# OPTIONS -fglasgow-exts #-} > {-# OPTIONS -fallow-undecidable-instances #-} > {-# OPTIONS -fallow-overlapping-instances #-} Let's implement the Kleene * in the type system. First some phantom types: > data At n > data Bt n > data Ct n > data Endt Now a chain to hold it together: > data Chain :: * -> * where > Ac :: (Chain n) -> Chain (At n) > Bc :: (Chain n) -> Chain (Bt n) > Cc :: (Chain n) -> Chain (Ct n) > Endc :: Chain Endt > > instance Show (Chain n) where > show (Ac n) = "Ac ~> " ++ (show n) > show (Bc n) = "Bc ~> " ++ (show n) > show (Cc n) = "Cc ~> " ++ (show n) > show (Endc) = "Endc" Now, first things first, implement identity. I don't really care about the actual functions right now, but let's put it in anyway. > class IdOp a b | a -> b, b -> a where > idOp :: a -> b > instance IdOp (Chain Endt) (Chain Endt) where > idOp x = x > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (At l)) (Chain (At r)) where > idOp (Ac n) = Ac (idOp n) > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (Bt l)) (Chain (Bt r)) where > idOp (Bc n) = Bc (idOp n) > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (Ct l)) (Chain (Ct r)) where > idOp (Cc n) = Cc (idOp n) Now, I'm sure many of you are wondering why I didn't just write: instance IdOp a a where idOp = id I'll get back to that. Now, if I go zero steps, how hard is it to go zero or one steps? > class ZeroOneStep a b where > step :: a -> b > instance ZeroOneStep (Chain (At l)) (Chain l) where > step (Ac n) = n > instance ZeroOneStep (Chain (Bt l)) (Chain l) where > step (Bc n) = n > instance ZeroOneStep (Chain (Ct l)) (Chain l) where > step (Cc n) = n > instance (IdOp (Chain l) (Chain r)) => > ZeroOneStep (Chain l) (Chain r) where > step = idOp That's quite nifty. I now need the allow-overlapping-instances, but it works fine: > fooZeroOne :: (ZeroOneStep a b) => a -> b -> Bool > fooZeroOne _ _ = True $ Main> fooZeroOne (Ac $ Bc Endc) (Ac $ Bc Endc) $ True $ Main> fooZeroOne (Ac $ Bc Endc) (Bc Endc) $ True $ Main> fooZeroOne (Ac $ Bc Endc) (Endc) $ ...big exposion... $ Main> fooZeroOne (Endc) (Bc Endc) $ ...big exposion... $ Main> fooZeroOne (Endc) (Endc) $ True So now I want zero or more steps, i.e. the Kleene *: > class ZeroMoreSteps a b where > stepN :: a -> b > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (At l)) (Chain r) where > stepN (Ac n) = stepN n > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (Bt l)) (Chain r) where > stepN (Bc n) = stepN n > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (Ct l)) (Chain r) where > stepN (Cc n) = stepN n I.e. if we can already get from l to r then it's not very hard to go one step further. Now I need to bring in single step case: > instance (ZeroOneStep (Chain l) (Chain r)) => > ZeroMoreSteps (Chain l) (Chain r) where > stepN = step So now it should be done right? Because ZeroOneStep allows for zero steps (via IdOp) so that should be it right? > fooZeroMore :: (ZeroMoreSteps a b) => a -> b -> Bool > fooZeroMore _ _ = True $ Main > fooZeroMore Endc Endc $ True $ Main > fooZeroMore (Ac Endc) Endc $ True $ Main > fooZeroMore (Ac Endc) (Ac Endc) $ ...big exposion... Now what I think is happening is this: the rules for overlapping instances say that the instance which is chosen is the one that specifies the most information matching the particular context. In this case, I think it's choosing and trying to use the "wrong" instances because of this rule. Even if I try: > instance ZeroMoreSteps (Chain l) (Chain l) where > stepN = id It still blows up. So I have to do it really manually: > instance ZeroMoreSteps (Chain Endt) (Chain Endt) where > stepN = id > instance ZeroMoreSteps (Chain (At l)) (Chain (At l)) where > stepN = id > instance ZeroMoreSteps (Chain (Bt l)) (Chain (Bt l)) where > stepN = id > instance ZeroMoreSteps (Chain (Ct l)) (Chain (Ct l)) where > stepN = id $ Main > fooZeroMore (Ac Endc) (Ac Endc) $ True Now it works. So am I right - is the reason for this the selection of the instance to use given overlapping instances? If so, can anything be done to make it "better"? I can think of two ideas, which I suspect should both be laughed out of town asap: i) backtrack - if you can't make the instance work then go back to where you last had a choice and try the next instance; ii) whilst selecting the instance which is most specific seems like the "sensible" choice, there are cases, eg guards, where the vertical order in the program determines the ordering of things to try. Maybe allow the programmer to specify an order to try things in? Note however how cool this now is: > data Cell :: * -> * where > Cell :: x -> Cell a -- muh ha Ha HAA! > data Section :: * -> * -> * where > SectionTo :: (ZeroMoreSteps a b) => > (Cell a) -> b -> Section a b > SectionFrom :: (ZeroMoreSteps a b) => > a -> (Cell b) -> Section a b > fooCoolness :: (ZeroMoreSteps a b) => > a -> Section a b -> Section a b -> b > fooCoolness chain _ _ = stepN chain $ Main > fooCoolness (Bc . Ac $ Endc) (SectionTo (Cell True) (Ac $ Endc)) (SectionFrom (Bc . Ac $ Endc) (Cell 1)) $ Ac ~> Endc And stepN just goes the "right" number of steps. The point is that I don't need to use pattern matching to decompose the various constructorsof Section in order to find (and perhaps fail) the unused section of the chain. Cheers, Matthew

1 0

Kleene star in types
by Matthew Sackman 07 May '07

07 May '07

This email is a .lhs. > {-# OPTIONS -fglasgow-exts #-} > {-# OPTIONS -fallow-undecidable-instances #-} > {-# OPTIONS -fallow-overlapping-instances #-} Let's implement the Kleene * in the type system. First some phantom types: > data At n > data Bt n > data Ct n > data Endt Now a chain to hold it together: > data Chain :: * -> * where > Ac :: (Chain n) -> Chain (At n) > Bc :: (Chain n) -> Chain (Bt n) > Cc :: (Chain n) -> Chain (Ct n) > Endc :: Chain Endt > > instance Show (Chain n) where > show (Ac n) = "Ac ~> " ++ (show n) > show (Bc n) = "Bc ~> " ++ (show n) > show (Cc n) = "Cc ~> " ++ (show n) > show (Endc) = "Endc" Now, first things first, implement identity. I don't really care about the actual functions right now, but let's put it in anyway. > class IdOp a b | a -> b, b -> a where > idOp :: a -> b > instance IdOp (Chain Endt) (Chain Endt) where > idOp x = x > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (At l)) (Chain (At r)) where > idOp (Ac n) = Ac (idOp n) > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (Bt l)) (Chain (Bt r)) where > idOp (Bc n) = Bc (idOp n) > instance (IdOp (Chain l) (Chain r)) => > IdOp (Chain (Ct l)) (Chain (Ct r)) where > idOp (Cc n) = Cc (idOp n) Now, I'm sure many of you are wondering why I didn't just write: instance IdOp a a where idOp = id I'll get back to that. Now, if I go zero steps, how hard is it to go zero or one steps? > class ZeroOneStep a b where > step :: a -> b > instance ZeroOneStep (Chain (At l)) (Chain l) where > step (Ac n) = n > instance ZeroOneStep (Chain (Bt l)) (Chain l) where > step (Bc n) = n > instance ZeroOneStep (Chain (Ct l)) (Chain l) where > step (Cc n) = n > instance (IdOp (Chain l) (Chain r)) => > ZeroOneStep (Chain l) (Chain r) where > step = idOp That's quite nifty. I now need the allow-overlapping-instances, but it works fine: > fooZeroOne :: (ZeroOneStep a b) => a -> b -> Bool > fooZeroOne _ _ = True $ Main> fooZeroOne (Ac $ Bc Endc) (Ac $ Bc Endc) $ True $ Main> fooZeroOne (Ac $ Bc Endc) (Bc Endc) $ True $ Main> fooZeroOne (Ac $ Bc Endc) (Endc) $ ...big exposion... $ Main> fooZeroOne (Endc) (Bc Endc) $ ...big exposion... $ Main> fooZeroOne (Endc) (Endc) $ True So now I want zero or more steps, i.e. the Kleene *: > class ZeroMoreSteps a b where > stepN :: a -> b > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (At l)) (Chain r) where > stepN (Ac n) = stepN n > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (Bt l)) (Chain r) where > stepN (Bc n) = stepN n > instance (ZeroMoreSteps (Chain l) (Chain r)) => > ZeroMoreSteps (Chain (Ct l)) (Chain r) where > stepN (Cc n) = stepN n I.e. if we can already get from l to r then it's not very hard to go one step further. Now I need to bring in single step case: > instance (ZeroOneStep (Chain l) (Chain r)) => > ZeroMoreSteps (Chain l) (Chain r) where > stepN = step So now it should be done right? Because ZeroOneStep allows for zero steps (via IdOp) so that should be it right? > fooZeroMore :: (ZeroMoreSteps a b) => a -> b -> Bool > fooZeroMore _ _ = True $ Main > fooZeroMore Endc Endc $ True $ Main > fooZeroMore (Ac Endc) Endc $ True $ Main > fooZeroMore (Ac Endc) (Ac Endc) $ ...big exposion... Now what I think is happening is this: the rules for overlapping instances say that the instance which is chosen is the one that specifies the most information matching the particular context. In this case, I think it's choosing and trying to use the "wrong" instances because of this rule. Even if I try: > instance ZeroMoreSteps (Chain l) (Chain l) where > stepN = id It still blows up. So I have to do it really manually: > instance ZeroMoreSteps (Chain Endt) (Chain Endt) where > stepN = id > instance ZeroMoreSteps (Chain (At l)) (Chain (At l)) where > stepN = id > instance ZeroMoreSteps (Chain (Bt l)) (Chain (Bt l)) where > stepN = id > instance ZeroMoreSteps (Chain (Ct l)) (Chain (Ct l)) where > stepN = id $ Main > fooZeroMore (Ac Endc) (Ac Endc) $ True Now it works. So am I right - is the reason for this the selection of the instance to use given overlapping instances? If so, can anything be done to make it "better"? I can think of two ideas, which I suspect should both be laughed out of town asap: i) backtrack - if you can't make the instance work then go back to where you last had a choice and try the next instance; ii) whilst selecting the instance which is most specific seems like the "sensible" choice, there are cases, eg guards, where the vertical order in the program determines the ordering of things to try. Maybe allow the programmer to specify an order to try things in? Note however how cool this now is: > data Cell :: * -> * where > Cell :: x -> Cell a -- muh ha Ha HAA! > data Section :: * -> * -> * where > SectionTo :: (ZeroMoreSteps a b) => > (Cell a) -> b -> Section a b > SectionFrom :: (ZeroMoreSteps a b) => > a -> (Cell b) -> Section a b > fooCoolness :: (ZeroMoreSteps a b) => > a -> Section a b -> Section a b -> b > fooCoolness chain _ _ = stepN chain $ Main > fooCoolness (Bc . Ac $ Endc) (SectionTo (Cell True) (Ac $ Endc)) (SectionFrom (Bc . Ac $ Endc) (Cell 1)) $ Ac ~> Endc And stepN just goes the "right" number of steps. The point is that I don't need to use pattern matching to decompose the various constructorsof Section in order to find (and perhaps fail) the unused section of the chain. Cheers, Matthew

1 0

RE: [Haskell] Monad and Monad transformer???
by Simon Peyton-Jones 07 May '07

07 May '07

As someone mentioned, don't worry about being new to Haskell, but DO send your messages to haskell-cafe(a)haskell.org, not the main Haskell list. I'm redirecting this so that the Helpful Denizens of the Haskell Cafe can help you. Simon | -----Original Message----- | From: haskell-bounces(a)haskell.org [mailto:haskell-bounces@haskell.org] On Behalf Of Duong Thai Ha | Sent: 06 May 2007 17:03 | To: haskell(a)haskell.org | Subject: [Haskell] Monad and Monad transformer??? | | hi | I am a newbee to haskell so please forgive me if the question is not | appropreate. | | I am trying to get the idea of using Monad and have to understand the | following code. | | Can you please tell me is how different Monad and Monad transformer is? | And what is mean by | | > newtype State s a = State { runState :: s -> (a, s) } | | --------------------------------- | | Define (State s) as a State monad | | > instance Monad (State s) where | > return a = State $ \s -> (a,s) | > m >>= k = State $ \s -> let (a,s') = runState m s in runState (k | a) s' | | Thanks lot for your help | I really do appreciate that | | Best regard | _______________________________________________ | Haskell mailing list | Haskell(a)haskell.org | http://www.haskell.org/mailman/listinfo/haskell

1 0

GADTs, type classes, existentials
by Mike Hamburg 07 May '07

07 May '07

Hello all, I'm trying to build a variation on Maps which supports a fast "concat-like" operation, for a library I'm writing. I'd rather not re-implement Data.Map, so I'm having a try with GADTs. The relevant part of my source file: ------------------------------------------------------------------------- data MetaMap k v where Map :: Ord k => Map.Map k v -> MetaMap k v Cat :: MetaMap k1 (MetaMap k2 v) -> MetaMap (k1,k2) v {- other, similar constructors -} lookup :: Monad m => k -> MetaMap k v -> m v lookup k (Map m) = Map.lookup k m lookup k (Cat m) = case k of (k1,k2) -> lookup k1 m >>= lookup k2 ------------------------------------------------------------------------- Unfortunately, this doesn't work: Data/MetaMap.hs:45:20: Could not deduce (Ord k) from the context (Monad m) arising from use of `Map.lookup' at Data/MetaMap.hs:45:20-33 Possible fix: add (Ord k) to the type signature(s) for `lookup' In the expression: Map.lookup k m In the definition of `lookup': lookup k (Map m) = Map.lookup k m This error happens because the Ord constraint isn't propagated into the function body. Of course, I can't add Ord k to the type signature for lookup, because one can't deduce Ord k from Ord (k1,k2). Is there a clean way around this error? Thanks for your time, Mike Hamburg

4 4

Mounting haskell.org wiki under WikipediaFS?
by dons＠cse.unsw.edu.au 07 May '07

07 May '07

Anyone tried editing haskell.org's wiki as text, using: http://wikipediafs.sourceforge.net/ "WikipediaFS is a mountable Linux virtual file system that enables you to deal with Wikipedia (or any Mediawiki-based site) articles as if they were real files. It is thus possible to use a real text editor to view and edit articles." Seems like this might make some of the larger edit passes a lot easier. -- Don

2 1

Is Excel a FP language?
by Tony Morris 07 May '07

07 May '07

In a debate I proposed "Excel is a functional language". It was refuted and I'd like to know what some of you clever Haskellers might think :) My opposition proposed (after some "weeding out") that there is a distinction between Excel, the application, the GUI and Excel, the language (which we eventually agreed (I think) manifested itself as a .xls file). Similarly, VB is both a language and a development environment and referring to VB is a potential ambiguity. I disagree with this analogy on the grounds that the very definition of Excel (proposed by Microsoft) makes no distinction. Further, it is impossible to draw a boundary around one and not the other. I also pointed to the paper by Simon Peyton-Jones titled, "Improving the world's most popular functional language: user-defined functions in Excel", which quite clearly refers to Excel as a [popular] functional language. The debate started when I referred to the fact that financial institutions change their functional language from Excel to something like OCaml or Haskell. Of course, there is no doubting that these companies can replace their entire use of Excel with a functional language, which I think is almost enough to fully support my position (emphasis on "almost"). -- Tony Morris http://tmorris.net/

14 15

Silly mail client
by Andrew Coppin 06 May '07

06 May '07

OK, this is hacking me off now... Does ANYBODY know how I can convince Thunderbird to send replies to Haskell Cafe rather than sending them to the original poster? This is really becoming tiresome...

7 7