- Haskell-Cafe - Haskell.org

Memoizing in parallel
by Pierre-Etienne Meunier 08 Nov '09

08 Nov '09

Hi, I'm designing an algorithm that uses dynamic programming. I've written it with an array, and it works, but it is still very slow and needs way too much memory. Then I realized that the array was very sparse (at most a O(\sqrt(n)) of its size is actually used). Now I want to rewrite it with a Data.Map, but since I do not know a priori what the keys are, I need a mutable ref somewhere. Which is fine, but... the array solution allowed me to explore the subsolutions in parallel with Control.Parallel, and use items already evaluated by other threads in the array, while an STRef forbids it. Can someone know how to do it (outside IO, and without using unsafePerformIO, of course) ? Thanks Pierre-Etienne Meunier

2 2

Names for properties of operators
by Neil Brown 08 Nov '09

08 Nov '09

Hi, We have names for properties of operators/functions. For example, if this holds: a % b = b % a for some operator %, we say that % is commutative. Similarly, if this holds: (a % b) % c = a % (b % c) we say that % is associative. Is there a name for this property, which I'm numbering 1, (where (%) :: a -> b -> b; i.e. the operator is potentially, but not necessarily, asymmetrically typed): 1: a % (b % c) = b % (a % c) For example, `Set.insert` obeys 1 for any values of a, b and c. (Any operator that is both associative and commutative automatically satisfies this property, but this property can be satisfied without the operator being either of those.) Given this property, we could prove useful follow-on results, such as: foldr (%) x ys = foldr (%) x (reverse ys) foldr (%) x ys = foldl (flip (%)) x ys The property 1 effectively states that the far-right hand element in a chain of such operators is special, but the ordering of everything to the left of it doesn't matter. One could conceive of a mirror property (where (%) :: a -> b -> a): 2: (a % b) % c = (a % c) % b If (%) obeys 1, flip (%) obeys 2 (and vice versa). I think these properties are useful -- I'd like to know if they have names already to describe them by. A similar property of two relations (where ((%), (~)) :: (a -> b -> b, c -> b -> b) ) would be: 3: a % (b ~ c) = b ~ (a % c) with mirror version (and adjusted types): 4: (a % b) ~ c = (a ~ c) % b Do these have a name? As an example, `Set.insert` and `Set.union` obey property 3 for all values of a, b and c. There are also symmetrically-typed examples of these operators, but the Set operations are easy and familiar. Thanks, Neil.

4 5

Fwd: [Haskell-cafe] Is () a 0-length tuple?
by Pasqualino "Titto" Assini 08 Nov '09

08 Nov '09

The syntax is similar, but what else is? In JavaScript there is a "null" value, that is the only value of the null type. Isn't () the same thing? The only value of the unary type? Best, titto 2009/11/6 John Dorsey <haskell(a)colquitt.org>: >> In what sense () is a 0-length tuple? > > In what sense isn't it? > > Data.Tuple is much to narrow to be of any use here. () is in at least most, > if not all, of the type classes that tuples are in. The syntax is > strikingly similar. > > If you ask me, it walks/quacks/smells like a duck, so it's a duck. > > Regards, > John > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe > -- Pasqualino "Titto" Assini, Ph.D. http://quicquid.org/

6 6

Nice addition to Foreign: castAny
by Maurício CA 08 Nov '09

08 Nov '09

This could be beside castPtr, castCharToCChar etc. ---- castAny :: (Storable a, Storable b) => a -> b castAny = unsafePerformIO . genericCast where genericCast :: (Storable a, Storable b) => a -> IO b genericCast v = return undefined >>= \r -> allocaBytes (max (sizeOf v) (sizeOf r)) $ \p -> poke p v >> if False then return r else peek (castPtr p) ---- GHCi: > let a = -1000 :: Int16 > castAny a :: Word16 --> 64536 > castAny a :: Ptr () 0xb4c2fc18 > castAny (castAny a :: Ptr ()) :: Int16 -1000 > let b = pi :: CLDouble > b 3.141592653589793 > castAny b :: CInt 1413754136 > castAny b :: Ptr () 0x54442d18 > castAny b :: CFloat 3.3702806e12 > castAny b :: Int8 24 At minimum, this is safer than 'unsafeCoerce'. What do you think? Best, Maurício

2 2

Master's thesis topic sought
by Matus Tejiscak 07 Nov '09

07 Nov '09

Hello, -Cafe, I'm looking for an interesting topic to hack on in my thesis. The thesis should be rather "theoretical"/abstract (writing a mail client in Haskell is not, for example), dealing with FP or related fields. I've had a few (blurry) ideas, ranging from investigating (possibilities for) Haskell extensions, to zygohistomorphic prepromorphisms, but nothing concrete, possibly because I'm not familiar with these areas enough to see what could be done -- which brings up a question whether it is a good idea to even try hacking on a topic like this. However, I'm eager to learn so if you have a topic you'd need somebody to work on, or just an interesting (or maybe even an uninteresting) idea, i'd be grateful for suggestions. :) Thanks, Matus

15 22

RE: Haskell as a first language?
by Simon Peyton-Jones 07 Nov '09

07 Nov '09

Haskell is a great language! Check out haskell.org. I'm ccing the Haskell Cafe which is read by many people better qualified to answer your question than me. (Since I've been working on Haskell for many years, I am not well qualified to say how it seems to a beginner.) S | -----Original Message----- | From: Charles Turner [mailto:charlie.h.turner@googlemail.com] | Sent: 11 July 2009 22:52 | To: Simon Peyton-Jones | Subject: Haskell as a first language? | | I'll make this short! Do you think Haskell is a good language to start | with? I am brand new to programming and have been using Scheme, some of | my peers suggest I should use Haskell. It seems "professional" to me. | Has features that a beginner should not worry about. What would you | suggest. (I'm not worried about bias) | | Thank you very much for your time. | | Charles Turner.

17 17

Re: [Haskell-cafe] Monad Comprehensions
by Günther Schmidt 07 Nov '09

07 Nov '09

Hello Ryan, thank you for your email, it's much appreciated. In summary I'd like to say that in the 2 years that I have been using Haskell I have managed to get a good working knowledge of the language which allows me to construct programs that are concise, correct and without greater redundancy. However I also realize more and more that in order to harness the full power of Haskell I need to gain a greater understanding of the underlying math and go beyond mechanical application. It is only lately that I am recognizing the significance of the often used term "abstraction". Certainly this is pointed out in the Haskell world often enough, but hey I'm a noob. For someone like me this is not immediately apparent. I mean there seems to be almost no end of what I still have to learn, barely scratching the surface, tip of the iceberg, that sort of thing. DSLs are apparently *the* way to go in the abstraction department. Oleg & Co's "Finally Tagless" style is my favorite when it comes to implementation. It bears this simplicity and elegance that I have come to recognize and what to me *is* "mathematical". Not all of my dealings in Haskell are goal-oriented. Well there is that of course, but mostly there are all these shiny beautiful things. I get distracted quite a lot too. Trying to dig deeper on the subject of Dan Piponis video for instance. BTW I found Phil Wadlers paper "Comprehending Monads" quite helpful here. I also find my focus of interest on the topics of articles on haskell-cafe shift. Before I would basically skip "deep stuff", ie. theoretical / abstract topics, and focus on technical details. Now I focus much more on "mathematical" issues despite the fact that I can barely understand them, it'll come to me eventually. Because now I consider the mathematical discussions far more fundamental and more "lasting". Technical details I can look up as the need arises and they will likely change over time, so no need to bother with them beforehand, they may be out-dated when you need them. Mathematical ones won't, they'll be good forever. Well that's my 2 cents for now. Günther Am 07.11.2009, 09:37 Uhr, schrieb Ryan Grant <rgrant(a)rgrant.org>: > 2009/11/1 Günther Schmidt <gue.schmidt(a)web.de>: >> In short, I'm truly lost. If anyone else, with roughly the same starting >> point as me, has found their way through this jungle I'd certainly >> appreciate some tips. In particular I wonder if someone has been able to >> follow what Dan demonstrates in this video, or was it a jaw-dropping >> experience for everyone else just as me? > > Günther, i have been lurking in H-cafe, and enjoying your pace of > exploration. AFAIK, you are trying to understand how to make > excellent languages using Haskell. you are asking some practical > questions that i find relevant to my own journey into Haskell. > > "best practices" for specifying a DSL are not well agreed upon, > partly due to recent advances. i sense this problem in your > question and in the newsworthiness of Dan's talk, (which i also > enjoyed with a jaw-drop). > > i encourage you to say exactly what you want from your DSL (do it > again, if you have already). > > i hope to follow your footsteps, maybe not in exact DSL needs, > but in path of discovery. >

1 0

Parsing Haskell
by Eric Macaulay 07 Nov '09

07 Nov '09

Hi all, I was hoping to use Language.Haskell.Parser to parse fragments of Haskell code. However, it appears that one can only use this module to parse complete Haskell programs. Is there anyway around this, or must I write a Haskell parser from scratch? Eric M.

6 6

Re: [Haskell-cafe] Re: Parsing Haskell
by Eric Macaulay 07 Nov '09

07 Nov '09

It has indeed reached me. Thanks! At 16:23 06/11/2009, Niklas Broberg wrote: > > You'd have to modify the parser in haskell-src(-exts) to do add the entry > > points, though. > >Actually haskell-src-exts already defines entry points for Module, >Stmt, Exp, Pat and Type (as well as [Module] in 1.3.x). > >Not sure if this message reaches the original poster though. > >Cheers, > >/Niklas >_______________________________________________ >Haskell-Cafe mailing list >Haskell-Cafe(a)haskell.org >http://www.haskell.org/mailman/listinfo/haskell-cafe

1 0

Haskell Weekly News: Issue 138 - November 07, 2009
by jfredett＠gmail.com 07 Nov '09

07 Nov '09

--------------------------------------------------------------------------- Haskell Weekly News http://sequence.complete.org/hwn/20091107 Issue 138 - November 07, 2009 --------------------------------------------------------------------------- Welcome to issue 138 of HWN, a newsletter covering developments in the [1]Haskell community. Lots of discussion about [2]Clean this week. As well there was a new DSL, feldspar, announced. It deals with digital signal processing applications. Announcements MonadRandom-0.1.4. Brent Yorgey [3]announced a new version of MonadRandom which adds applicative instances for Rand and RantT, so you can write your code in applicative style. Criterion 0.2, an improved Haskell benchmarking library. Bryan O'Sullivan [4]announced a new version of Criterion, all the details of this release are available on his [5]blog feldspar-language. Emil Axelsson [6]announced feldspar, a DSL for digital signal processing. feldspar-compiler. Emil Axelsson [7]announced the C code backend for the `feldspar` language. fdo-notify 0.1, a client for the Desktop Notifications protocol. Max Rabkin [8]announced a library for FreeDesktop.org's Desktop Notifications Protocol. language-python version 0.2 now available. Bernie Pope [9]announced a new version of the language-python package, which provides an AST and parser for Python 2.x-3.x (previously, only 3.x was supported). timeplot. Eugene Kirpichov [10]announced timeplot, which is useful visualizing log files. Singapore FP Users Group First Meeting. Max Cantor [11]announced the first meeting of the Singapore FP Users Group, it will be Monday, November 2nd at 6pm. Advgame 0.1.1. Tim Wawrzynczak [12]announced his port of Conrad Barski's 'Casting SPELs in Lisp' to Haskell. BlogLiterately-0.2. Robert Greayer [13]announced version 0.2 of BlogLiterately, a simple tool for uploading posts written in markdown and Literate Haskell to blogs. haskell-mode 2.6. Svein Ove Aas [14]announced a bugfix release of the Emac's Haskell mode. Discussion A Problem Defining a Monad instance. Petr Pudlak [15]asked about how to defined an instance of the Monad class for a monad whose argument is restricted by another typeclass. Point Free Case Expressions. Sebastiaan Visser [16]suggested that a new syntax be added for 'point-free' case expressions. Master's thesis topic sought. Matus Tejiscak [17]asked for suggestions for possible Master's Thesis topics. What's the deal with Clean? Deniz Dogan [18]asked about the recent discussion on the -cafe list about Clean, another Pure, Lazy, Strictly Typed language. Blog noise [19]Haskell news from the [20]blogosphere. Blog posts from people new to the Haskell community are marked with >>>, be sure to welcome them! * Neil Brown: [21]Text.Printf and monad transformers. * Bryan O'Sullivan: [22]Criterion 0.2, an improved Haskell benchmarking library. * Galois, Inc: [23]Tech Talk: Hoare-Logic â fiddly details and small print. * Gtk2HS: [24]Writing concurrent programs.. * Tom Schrijvers: [25]Postdoc/PhD Positions on the Monadic Constraint Programming project. * Darcs: [26]darcs weekly news #45. * Tom Schrijvers: [27]Haskell Type Constraints unleashed. * Luke Plant: [28]Building GHC is fun.... * Manuel M T Chakravarty: [29]Status Update of the Glasgow Haskell Compiler, October 2009. * Chris Smith: [30]Monads from Two Perspectives. * Neil Brown: [31]Concurrent Testing and Tracing: Useful Output for Test Failures. * Dan Piponi (sigfpe): [32]Buffon's Needle, the Easy Way. Buffon's needle is a popular probability problem. Rule lines on the floor a distance d apart and toss a needle of length l Quotes of the Week * blackdog: [About Hubris] I tell the Ruby guys that Haskell will help them speed up their Ruby code and keep their apps going, and I tell Haskell guys that it'll Trojan Horse those poor unsuspecting rubyists... * lispy: Great, I leave the channel for a few hours and suddenly Haskell has a new found work ethic. * roconnor: ivanm: I will keep the fail in the code * lament: just remember to ask, 'What are your questions', as opposed to 'Do you have any questions' * mauke: @unpl const (flip const) lambdabot: (\ _ c d -> d) About the Haskell Weekly News New editions are posted to [33]the Haskell mailing list as well as to [34]the Haskell Sequence and [35]Planet Haskell. [36]RSS is also available, and headlines appear on [37]haskell.org. To help create new editions of this newsletter, please see the information on [38]how to contribute. Send stories to jfredett . at . gmail . dot . com. The darcs repository is available at darcs get [39]http://patch-tag.com/r/jfredett/HWN2/pullrepo HWN2 . References 1. http://haskell.org/ 2. http://en.wikipedia.org/wiki/Clean_(programming_language) 3. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65844 4. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65813 5. http://www.serpentine.com/blog/2009/11/06/criterion-0-2-an-improved-haskell… 6. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65617 7. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65741 8. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65686 9. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65656 10. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65575 11. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65471 12. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65462 13. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65567 14. http://article.gmane.org/gmane.comp.lang.haskell.cafe/65487 15. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/65834 16. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/65750 17. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/65706 18. http://thread.gmane.org/gmane.comp.lang.haskell.cafe/65589 19. http://planet.haskell.org/ 20. http://haskell.org/haskellwiki/Blog_articles 21. http://chplib.wordpress.com/2009/11/06/text-printf-and-monad-transformers/ 22. http://www.serpentine.com/blog/2009/11/06/criterion-0-2-an-improved-haskell… 23. http://www.galois.com/blog/2009/11/04/chapman-hoare/ 24. http://haskell.org/gtk2hs/archives/2009/11/04/writing-concurrent-programs/ 25. http://tomschrijvers.blogspot.com/2009/11/postdocphd-positions-on-monadic.h… 26. http://blog.darcs.net/2009/11/darcs-weekly-news-45.html 27. http://tomschrijvers.blogspot.com/2009/11/haskell-type-constraints-unleashe… 28. http://lukeplant.me.uk/blog.php?id=1107301703 29. http://justtesting.org/post/232566357 30. http://cdsmith.wordpress.com/2009/11/02/monads-from-two-perspectives/ 31. http://chplib.wordpress.com/2009/11/02/concurrent-testing-and-tracing-usefu… 32. http://blog.sigfpe.com/2009/10/buffons-needle-easy-way.html 33. http://www.haskell.org/mailman/listinfo/haskell 34. http://sequence.complete.org/ 35. http://planet.haskell.org/ 36. http://sequence.complete.org/node/feed 37. http://haskell.org/ 38. http://haskell.org/haskellwiki/HWN 39. http://patch-tag.com/r/jfredett/HWN2/pullrepo%20HWN2

3 2

[Haskell-beginners] Re: Is Haskell for me?
by Felipe Lessa 07 Nov '09

07 Nov '09

Sorry for not sending to the list. ----- Forwarded message from Felipe Lessa <felipe.lessa(a)gmail.com> ----- On Fri, Nov 06, 2009 at 07:58:30PM -0200, Felipe Lessa wrote: > On 11/6/09, Gaius Hammond <gaius(a)gaius.org.uk> wrote: > > To be fair, Python offloads its heavy lifting to C libraries - NumPy > > and SciPy run at very close to full C speed on large datasets. This is > > also how Matlab works. Unladen Swallow is an upcoming JIT compiler for > > Python. > > > > Where Haskell shines for computation is when you can leverage lazy > > evaluation. > > *If* you can offload most of your work to SciPy. Depending on what you > do this is at least difficult. I don't know how much of a neural > network can be represented as big fat matrix :). I would also like to add that Haskell shines for parallel computation, being easy and rewarding to use multiple cores or processors in a SMP machine. Parallel computing in Python is at least ugly because you need to use multiple processes, while the same task in C++ needs to be dealt with manually using concurrency. -- Felipe. ----- End forwarded message -----

1 0

What's the new type ?
by zaxis 07 Nov '09

07 Nov '09

type Parser a = GenParser Char () a newtype GenParser tok st a = Parser (State tok st -> Consumed (Reply tok st a)) As i know, the Parser type is just an alias of GenParser. Then can the second line be replaced as below? newtype GenParser tok st a = GenParser Char () (State tok st -> Consumed (Reply tok st a)) If it can , then what is the new type ? Sincerely! ----- fac n = foldr (*) 1 [1..n] -- View this message in context: http://old.nabble.com/What%27s-the-new-type---tp26208600p26208600.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

4 5

Is () a 0-length tuple?
by Pasqualino "Titto" Assini 06 Nov '09

06 Nov '09

Hi, I just noticed that the json library maps () to a JSON empty array: []. This makes sense if () is a 0-length tuple, the (2,"hi") tuple for example maps to [2,"hi"] But is it so in Haskell? In what sense () is a 0-length tuple? Thanks, titto

3 2

ANN: MonadRandom-0.1.4
by Brent Yorgey 06 Nov '09

06 Nov '09

Just a quick note: I've uploaded a new version of the MonadRandom which adds Applicative instances for Rand and RandT. Now you can write your random code in an applicative style! -Brent

1 0

ANNOUNCE: NetSNMP-0.1.6
by John Dorsey 06 Nov '09

06 Nov '09

I'm pleased to annonce the release of NetSNMP-0.1.6, a partial (but useful) binding to the net-snmp library.[1] NetSNMP is a simple and self-contained client interface for doing basic SNMP GET, NEXT, and WALK operations. As far as I know it should work with any version of the C library (net-snmp) from version 5. This release improves threaded performance, and the management of foreign memory objects. I've been using it successfully in a threaded polling daemon which I'm running on 32-bit and 64-bit Linux systems. If you try out NetSNMP, or if you decide it doesn't meet your need, please drop me a line! Regards, John Dorsey

1 0

Fwd: [Haskell-cafe] What's the deal with Clean?
by Tracy Wadleigh 06 Nov '09

06 Nov '09

> I had to implement a ring buffer, and I wanted the code using it to be > written in Haskell. I ended up implementing the buffer in C, and wrapping > it in FFI from Haskell because implementing a destructive array in Haskell > is kind of unwieldy to someone of my experience level. In Clean, it looks > like the uniqueness typing allows for destructive updates in a very > controlled manner. > The ST monad provides this functionality. The never-instantiated-in-a-visible-way state parameter of the ST monad provides the "uniqueness" required for doing destructive updates in a pure way.

13 30

Memory Leak - Artificial Neural Network
by Shelby Moore 06 Nov '09

06 Nov '09

wren ng thornton wrote: http://www.haskell.org/pipermail/haskell-cafe/2009-November/068728.html >...use matrices/arrays rather than lists... >...even just using an unboxed array for the matrices > will make it much faster to traverse through... Matrix multiplication is the sequential order inner or outer product of the columns/rows of one matrix and rows/columns of 2nd matrix: http://en.wikipedia.org/wiki/Matrix_multiplication#Relationship_with_the_in… Afaics, optimize matrix multiplication with lists. Random arrays incurs a huge performance penality in Haskell, unless you revert to non-lazy implementation, in which case you have voided multi-core speed ups for very large arrays (aren't ANNs huge matrices?).

1 0

ANN: ZuriHac: Haskell hackathon in Zurich, March 19-21
by Johan Tibell 06 Nov '09

06 Nov '09

Hi all! I am very pleased to officially announce ZuriHac, a Haskell hackathon/get-together to be held March 19-21 at the Google Office in Zurich, Switzerland. Thanks to everyone who has already expressed interest in attending. I want to stress that everyone is welcome---you do not have to be a Haskell guru to attend! Helping hack on someone else's project could be a great way to increase your Haskell-fu. If you plan on coming, please officially register [1], even if you already put your name on the wiki. Registration, travel, lodging and many other details will soon be available on the ZuriHac wiki [2]. WHEN Friday March 19 2:30pm to 6:30pm Saturday March 20 10am to 6pm Sunday March 21 10am to 6pm WHERE We will be in the TechTalk area of the Google Office at Brandschenkestrasse 110. Please see the wiki [3] for directions. ORGANIZERS Johan Tibell Christophe Poucet Hope to see you in Zurich! - The ZuriHac team [1] http://haskell.org/haskellwiki/ZuriHac/Register [2] http://haskell.org/haskellwiki/ZuriHac [3] http://haskell.org/haskellwiki/ZuriHac#Getting_to_the_Google_Office

1 0

[ANN] Criterion 0.2, an improved Haskell benchmarking library
by Bryan O'Sullivan 06 Nov '09

06 Nov '09

I’m pleased to announce the availability of version 0.2 of my criterion library <http://hackage.haskell.org/package/criterion> for Haskell performance evaluation. For details, please see below: - http://www.serpentine.com/blog/2009/11/06/criterion-0-2-an-improved-haskell…

1 0

Memory Leak - Artificial Neural Network
by Hector Guilarte 06 Nov '09

06 Nov '09

Hello everyone, I just implemented an Artificial Neural Network but I'm having a serious memory leak. I was very careful of using tail recursion all over my code, but for some reason (a.k.a lazyness) my program is misusing incredible ammounts of RAM. I read the whole chapter 25 of Real World Haskell trying to find a solution with no luck. Maybe somebody can take a look at the code to help me out with this problem, I would really appreciate it. Thanks A LOT in advance, Hector Guilarte Ps: The file is also attached Ps2: The code is written in Spanglish, sorry for that, I'm working on that bad habbit... module Main where import Control.Monad import System.IO import qualified Random import System.IO.Unsafe import System.Environment import Data.List data ANN = ANN Layer Layer Layer -- ^ Red Neuronal de 3 capas deriving (Eq, Show) type Layer = [Neuron] -- ^ Lista de Neuronas que conforman la capa data Neuron = Neuron [(Float,Float)] Float -- ^ Lista de (pesos,xs) y umbral asociado deriving (Eq, Show) neurona:: Neuron -> -- ^ [(Pesos,Xs)] y Umbral Float neurona (Neuron entrada umbral) = let entradaTupla = unzip entrada pesos = fst entradaTupla xs = snd entradaTupla suma = foldl' (+) (-umbral) (zipWith (*) xs pesos) in sigmoidal suma neurona2:: [(Float,Float)] -> -- ^ [(Pesos,Xs)] Float -> -- ^ Umbral Float neurona2 valores umbral = let entradaTupla = unzip valores pesos = fst entradaTupla xs = snd entradaTupla suma = foldl' (+) umbral (zipWith (*) xs pesos) in sigmoidal suma -- ANN [] [Neuron [(4.7621,0.9993291),(4.7618,0.94501287)] 7.3061,Neuron [(6.3917,0.9993291),(6.3917,0.94501287)] 2.8441] [Neuron [(-10.3788,0.9993291),(9.7691,0.94501287)] 4.5589] sigmoidal:: Float -> Float sigmoidal x = 1 / (1 + (exp (-x))) main:: IO() main = do -- nombreArchivo <- getArgs -- archivo <- readFile (head nombreArchivo) pesos <- pesosIniciales 10000 randomXs <- generarRandomXs 5000 randomYs <- generarRandomYs 5000 let conjunto = generar 200 0 0 randomXs randomYs [] --print conjunto -- let lista = parsearString archivo [[]] -- let splitted = split lista [] let (a,b,c) = (unzip3 (take 200 conjunto)) --let (a,b,c) = ([0,1,0,1],[0,0,1,1],[0,1,1,0]) let ejemplos = zipWith (ajustarEjemplos) a b -- print ejemplos let nuevaRed = armarRed 2 8 1 pesos let entrenada = train nuevaRed ejemplos c let redInicializada = map (iniciarXsRed entrenada) ejemplos let resultados = map resultadoRed1Output (map evaluarRed redInicializada) print nuevaRed print entrenada print resultados return () ajustarEjemplos:: Float -> Float -> [Float] ajustarEjemplos a b = [a,b] train:: ANN -> [[Float]] -> [Float] -> ANN train red ejemplosTodos esperadosTodos = let entrenado = entrenamiento red ejemplosTodos esperadosTodos [] 200 squaredErrors = snd entrenado in if squaredErrors < 3 then fst entrenado else train (fst entrenado) ejemplosTodos esperadosTodos -- ENTRENAMIENTO entrenamiento:: ANN -> [[Float]] -> [Float] -> [Float] -> Int -> (ANN,Float) entrenamiento red _ _ accum 0 = let squaredErrors = foldl' (+) 0 (map (**2) accum) in (red,squaredErrors) entrenamiento red ejemplos esperados accum epoch = let redInicializada = iniciarXsRed red (head ejemplos) redEvaluada = evaluarRed redInicializada redAjustada = ajustarPesos redEvaluada (head esperados) error = (head esperados) - (resultadoRed1Output redAjustada) in entrenamiento redAjustada (tail ejemplos) (tail esperados) (accum ++ [error]) (epoch-1) resultadoRed1Output:: ANN -> Float resultadoRed1Output (ANN _ _ [(Neuron ((_,xs):_) _)]) = xs iniciarXsRed:: ANN -> [Float] -> ANN iniciarXsRed (ANN inputLayer hiddenLayer outputLayer) valores = let inputNueva = zipWith ajustarXsInput inputLayer valores in (ANN inputNueva hiddenLayer outputLayer) ajustarXsInput:: Neuron -> Float -> Neuron ajustarXsInput (Neuron listaNeurona threshold) xsInput = let listaNueva = map (ajustarXs xsInput) listaNeurona in (Neuron listaNueva threshold) -- FIN ENTRENAMIENTO pesosIniciales :: Int -> IO [Float] pesosIniciales n = do (replicateM n (Random.getStdRandom intervalo)) where intervalo = Random.randomR (-0.5,0.5) parsearString:: String -> [String] -> [String] parsearString [] lista = (tail lista) parsearString (x:xs) lista = if x == '\n' then parsearString xs ([]:lista) else parsearString xs (((head lista) ++ [x]):(tail lista)) split:: [String] -> [(Float,Float,Float)] -> [(Float,Float,Float)] split [] accum = accum split (x:xs) accum = let first = readNum x "" fstNum = read $ fst first second = readNum (snd first) "" sndNum = read $ fst second third = readNum (snd second) "" thrdNum = if (head $ fst third) == 'A' then 0 else 1 in split xs ([(fstNum,sndNum,thrdNum)]++accum) readNum:: String -> String -> (String,String) readNum [] num = ([(head num)],num) readNum (x:xs) num = if x == ' ' then (num,xs) else (if x == '\n' then (num,xs) else readNum xs (num ++ [x]) ) generar:: Int -> Int -> Int -> [Float] -> [Float] -> [(Float,Float,Float)] -> [(Float,Float,Float)] generar total dentro fuera randomXs randomYs accum | total == dentro + fuera = accum | dentro == total `div` 2 = let x = head randomXs y = head randomYs isDentro = ((x-15)**2) + ((y-6)**2) <= 9 in if isDentro then generar total dentro fuera (tail randomXs) (tail randomYs) accum else generar total dentro (fuera+1) (tail randomXs) (tail randomYs) (accum ++ [(x,y,0)]) | fuera == total `div` 2 = let x = head randomXs y = head randomYs isDentro = ((x-15)**2) + ((y-6)**2) <= 9 in if isDentro then generar total (dentro+1) fuera (tail randomXs) (tail randomYs) (accum ++ [(x,y,1)]) else generar total dentro fuera (tail randomXs) (tail randomYs) accum | otherwise = let x = head randomXs y = head randomYs isDentro = ((x-15)**2) + ((y-6)**2) <= 9 in if isDentro then generar total (dentro+1) fuera (tail randomXs) (tail randomYs) (accum ++ [(x,y,1)]) else generar total dentro (fuera+1) (tail randomXs) (tail randomYs) (accum ++ [(x,y,0)]) generarRandomXs :: Int -> IO [Float] generarRandomXs n = do (replicateM n (Random.getStdRandom intervalo)) where intervalo = Random.randomR (0.0,20.0) generarRandomYs :: Int -> IO [Float] generarRandomYs n = do (replicateM n (Random.getStdRandom intervalo)) where intervalo = Random.randomR (0.0,12.0) -- ARMAR RED armarRed:: Int -> Int -> Int -> [Float] -> ANN armarRed numNeuronasInput numNeuronasHidden numNeuronasOutput randoms = let layerInput = armarLayerInput numNeuronasInput numNeuronasHidden randoms [] layerHidden = armarLayerHidden numNeuronasHidden numNeuronasOutput (snd layerInput) [] layerOutput = armarLayerOutput numNeuronasOutput (snd layerHidden) [] in (ANN (fst layerInput) (fst layerHidden) layerOutput) armarLayerInput:: Int -> Int -> [Float] -> Layer -> (Layer,[Float]) armarLayerInput 0 _ randoms accum = (accum,randoms) armarLayerInput numNeuronasInput numNeuronasHidden randoms accum = let listaNeurona = armarListaNeuronasInput numNeuronasHidden randoms [] newRandoms = snd listaNeurona neurona = [(Neuron (fst listaNeurona) 0)] in armarLayerInput (numNeuronasInput-1) numNeuronasHidden newRandoms (accum ++ neurona) armarLayerHidden:: Int-> Int -> [Float] -> Layer -> (Layer,[Float]) armarLayerHidden 0 _ randoms accum = (accum,randoms) armarLayerHidden numNeuronasHidden numNeuronasOutput randoms accum = let listaNeurona = armarListaNeuronasHidden numNeuronasOutput randoms [] neurona = [(Neuron (fst listaNeurona) (head $ snd listaNeurona))] in armarLayerHidden (numNeuronasHidden-1) numNeuronasOutput (tail $ snd listaNeurona) (accum ++ neurona) armarListaNeuronasHidden:: Int -> [Float] -> [(Float,Float)] -> ([(Float,Float)],[Float]) armarListaNeuronasHidden 0 randoms accum = (accum,randoms) armarListaNeuronasHidden numElems randoms accum = let pesosYxs = [((head randoms),(head $ tail randoms))] in armarListaNeuronasHidden (numElems-1) (tail $ tail randoms) (accum ++ pesosYxs) armarListaNeuronasInput:: Int -> [Float] -> [(Float,Float)] -> ([(Float,Float)],[Float]) armarListaNeuronasInput 0 randoms accum = (accum,randoms) armarListaNeuronasInput numElems randoms accum = let pesosYxs = [((head randoms),0)] in armarListaNeuronasInput (numElems-1) (tail randoms) (accum ++ pesosYxs) armarLayerOutput:: Int -> [Float] -> Layer -> Layer armarLayerOutput 0 _ accum = accum armarLayerOutput numNeuronasHidden randoms accum = let neurona = [(Neuron [(0,(head randoms))] (head $ tail randoms))] in armarLayerOutput (numNeuronasHidden-1) (tail $ tail randoms) (accum ++ neurona) -- FIN ARMAR RED -- EVALUAR RED evaluarRed:: ANN -> ANN evaluarRed (ANN inputLayer hiddenLayer outputLayer) = let newHidden = ajustarLayer inputLayer hiddenLayer [] 0 newOutput = ajustarLayer newHidden outputLayer [] 0 in (ANN inputLayer newHidden newOutput) ajustarLayer:: Layer -> Layer -> Layer -> Int -> Layer ajustarLayer _ [] accum numNeurona = accum ajustarLayer leftLayer ((Neuron listaNeurona threshold):rightLayer) accum numNeurona = let valorLayer = evaluarLayer leftLayer threshold numNeurona listaNeuronaNew = map (ajustarXs valorLayer) listaNeurona in ajustarLayer leftLayer rightLayer (accum ++ [(Neuron listaNeuronaNew threshold)]) (numNeurona+1) ajustarXs:: Float -> (Float,Float) -> (Float,Float) ajustarXs xs (peso,_) = (peso,xs) evaluarLayer:: Layer -> Float -> Int -> Float evaluarLayer layer threshold numNeurona = let listaTuplas = extraerTuplaLayer layer numNeurona [] valor = neurona2 listaTuplas threshold in valor extraerTuplaLayer:: Layer -> Int -> [(Float,Float)] -> [(Float,Float)] extraerTuplaLayer [] _ accum = accum extraerTuplaLayer ((Neuron tupla _):resto) numNeurona accum = extraerTuplaLayer resto numNeurona (accum ++ [(tupla !! numNeurona)]) -- FIN EVALUAR RED -- AJUSTAR RED ajustarPesos:: ANN -> Float -> ANN ajustarPesos salida@(ANN inputLayer hiddenLayer outputLayer) esperado = let outputNuevo = map (ajustarPesoOutput esperado) outputLayer gradientes = snd $ unzip outputNuevo hiddenNuevo = map (ajustarPesoHidden gradientes) hiddenLayer gradientes2 = snd $ unzip hiddenNuevo inputNuevo = map (ajustarPesoInput gradientes2) inputLayer in (ANN inputNuevo (fst $ unzip hiddenNuevo) (fst $ unzip outputNuevo)) ajustarPesoOutput:: Float -> Neuron -> (Neuron,Float) ajustarPesoOutput esperado (Neuron [(peso,obtenido)] threshold) = let error = esperado-obtenido gradiente = obtenido*(1-obtenido)*error deltaTheta = tasaAprendizaje*(-1)*gradiente thresholdNuevo = threshold + deltaTheta in ((Neuron [(peso,obtenido)] thresholdNuevo),gradiente) ajustarPesoHidden:: [Float] -> Neuron -> (Neuron,Float) ajustarPesoHidden gradientes (Neuron listaNeurona threshold) = let (pesosViejos,xsViejos) = unzip listaNeurona pesosAjustados = zipWith ajustarPesosHidden listaNeurona gradientes sumatoriaGradientes = foldl' (+) 0 (zipWith (*) gradientes pesosViejos) gradiente = (head xsViejos)*(1-(head xsViejos))*sumatoriaGradientes thresholdNuevo = tasaAprendizaje*(-1)*gradiente in ((Neuron pesosAjustados thresholdNuevo),gradiente) ajustarPesoInput:: [Float] -> Neuron -> Neuron ajustarPesoInput gradientes (Neuron listaNeurona threshold) = let (pesosViejos,xsViejos) = unzip listaNeurona pesosAjustados = zipWith (+) pesosViejos (map (*tasaAprendizaje) (zipWith (*) gradientes xsViejos)) listaNeuronaNueva = zip pesosAjustados xsViejos in (Neuron listaNeuronaNueva threshold) ajustarPesosHidden:: (Float,Float) -> Float -> (Float,Float) ajustarPesosHidden (pesoViejo,xs) gradiente = let deltaW = tasaAprendizaje*xs*gradiente pesoNuevo = pesoViejo + deltaW in (pesoNuevo,xs) -- FIN AJUSTAR RED tasaAprendizaje = 0.1

4 6

Re: Simple FAST lazy functional primes
by Steve 06 Nov '09

06 Nov '09

On Tue, 2009-11-03 at 10:52 -0500, Will wrote: > I've just tried it and it was twice slower than mine. (?) I didn't use > the [Int] > signature in both. Maybe it's not real issues we're dealing with here, > but > compiler/OS/CPU issues? (or have you've forgotten to put the [Int] > signature > into my code too, when tested? It runs twice as fast with it). > Although your > code has an advantage that it is very easy to add the wheel > optimization to it. I used [Int] for both. If you didn't use [Int] then it defaults to [Integer] and you are probably testing the speed of GMP integer operations (rather than the speed of Haskell Int operations) which could give differing conclusions. I had looked into using a wheel before. Its nice in theory, but not so useful in practice. At least that's my experience where using a wheel made the primes function slower. > What I have now, is this: > > qprimes = 2: 3: sieve emptyQ primes' 5 where > primes' = tail qprimes > sieve q (p:ps) x > = h ++ sieve (addQ q' (2*p) (p*p+2*p)) ps (p*p+2) > where > (h,q') = noComps q [x,x+2..p*p-2] > ...... Adding your code and conclusions to "Prime numbers" on the Haskell wiki, could be useful. http://www.haskell.org/haskellwiki/Prime_numbers I've just updated the page to split it into "Finding Primes" and "Testing Primality". Steve

2 1

Code generated by pointfree does not compile
by Boris Lykah 05 Nov '09

05 Nov '09

Hi all! I was playing with pointfree tool from hackage and found that it produces wrong code for some functions which use list comprehension. Here are several examples: :main f x = concat [x:f x| x<-[0..x-1]] f = fix ((join .) . flip flip [] . ((:) .) . ap (:) . (`ap` (enumFromTo 0 . subtract 1)) . ((<-) .) . ((|) =<<)) This does not compile because (<-) and (|) are used as functions whereas they are not. :main map (+1) [x| x<-[0..x]] [x | x <- [0..x] + 1] This does not compile either because the result code tries to add one to list. Why does this happen? This looks like a bug. Thanks, Boris Lykah -- View this message in context: http://old.nabble.com/Code-generated-by-pointfree-does-not-compile-tp262117… Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

2 1

Are all arrows functors?
by Neil Brown 05 Nov '09

05 Nov '09

Hi, I was thinking about some of my code today, and I realised that where I have an arrow in my code, A b c, the type (A b) is also a functor. The definition is (see http://www.haskell.org/ghc/docs/latest/html/libraries/base/Control-Arrow.ht…) fmap = (^<<) -- Or, in long form: fmap f x = arr f <<< x Out of curiosity, and since this is a typical haskell-cafe question, does this definition of fmap hold for all arrows? And is there a wiki page somewhere that has a table of all of these Haskell type-classes (Functor, Monad, Category, Arrow, Applicative and so on), and says that if you are an instance of class A you must have some corresponding instance of B? (e.g. all Monads are Functors and Applicatives) I'm fairly certain my arrow isn't a Monad or Applicative, although of course it must be a Category, given the type-class dependency, but it would be nice when using one of these things to see what other instances you should automatically supply. Thanks, Neil.

6 5

Storable Vector as a Storable record?
by Hemanth Kapila 05 Nov '09

05 Nov '09

Hi, Is it possible to somehow make a StorableVector of a StorableVector via store-record or something? If yes, could some one please provide me with some hint? I need a very fast and efficient array of a large number of $ arrays of *Int *s. And the storableVector seems to be extremely nice. Am trying to understand the way a tuple is made a storable record in the synthesizer package on hackage but I thought I will ask it here too. Thanks in advance Hemanth K

2 1

proposal: point free case expressions
by Sebastiaan Visser 05 Nov '09

05 Nov '09

Hello all, Wouldn't it be nice if we could write point free case statements? I regularly find myself writing down something like this: > myFunc = anotherFunc $ \x -> case x of > Left err -> print err > Right msg -> putStrLn msg We could really use a case statement in which we skip the scrutinee and make `(case of {})' be syntactic sugar for `(\x -> case x of {})'. So we could write: > myFunc = anotherFunc $ case of > Left err -> print err > Right msg -> putStrLn msg A minor syntactical addition, a big win! Cheers, -- Sebastiaan Visser

14 21

Re: Fair diagonals
by Louis Wasserman 05 Nov '09

05 Nov '09

+1 on Control.Monad.Omega. In point of fact, your diagN function is simply diagN = runOmega . mapM Omega You'll find it an interesting exercise to grok the source of Control.Monad.Omega, obviously, but essentially, you're replacing concatMap with a fair (diagonal) traversal order version. Louis Wasserman wasserman.louis(a)gmail.com http://profiles.google.com/wasserman.louis

2 2

Strict Monad
by Álvaro García Pérez 05 Nov '09

05 Nov '09

Hi, I'm trying to characterise some strict monads to work with a particular lambda-calculus evaluator, as an alternative to CPS monad. In the thread "Stupid question #852: Strict monad" the implementation of some strict monads (and pseudo-monads, not meeting the identity laws) is discussed. It's clear form that thread that using pattern-matching in the (>>=) operation will force evaluation, then the Id monad defined with pattern-matching is strict (and it's indeed a monad): > newtype Id a = Id a > > instance Monad Id where > return = Id > (Id a) >>= f = f a But it's impossible to derive a monad transformer from it, because you don't know the constructors of the monads you're transforming. I then tried to use strict application ($!). My first attempt was > newtype Strict a = InSt { outSt :: a } > > instance Monad Strict where > return = InSt > m >>= f = (\x -> f x) $! (outSt m) which is not a monad (doesn't meet the left identity law). (return undefined) >>= (\x -> const (return 1) x) =/= (return 1) Then I wondered if it was enough to force the evaluation of the whole monad, instead of the value inside the monad, yielding the second attempt: > newtype Strict a = InSt { outSt :: a } > > instance Monad Strict where > return = InSt > m >>= f = (\x -> f (outSt x)) $! m I placed the outSt inside the anonymous function, leaving the monad on the right of the ($!). This meets the identity laws and surprisingly (I was expecting a lazy behaviour) implements strict semantics (it behaves like CPS, which is strict as well). A transformer from this monad can be easily written: > newtype StrictT m a = InStT { outStT :: m a } > > instance Monad m => Monad (StrictT m) where > return = InStT . return > m >>= t = InStT $ (\x -> x >>= (\a -> outStT $ t a)) $! (outStT m) > > instance MonadTrans StrictT where > lift = InStT Is it common practice to use this monad and this transformer? Why is it not in the standard library? I looked for this monad in the literature but I didn't find anything similar. It seems naive to me that this has never been previously described. Am I doing something wrong and this is not a monad at all? Alvaro.

2 2

About Regex question
by zaxis 05 Nov '09

05 Nov '09

import Text.Regex.Posix import Data.Char import Data.List guessKeys line = concat $ intersperse "-" (modifiers ++ [map toLower key]) where modifiers = map (!!1) (line =~ "(mod|shift|control)Mask") (_, _, _, [key]) = line =~ "xK_(\\w+)" :: (String, String, String, [String]) aaa line = map (!!1) (line =~ "(mod|shift|control)Mask") bbb line = line =~ "xK_(\\w+)" :: (String, String, String, [String]) In ghci >guessKeys "modF1" "*** Exception: Test.hs:8:10-85: Irrefutable pattern failed for pattern (_, _, _ , [key]) > aaa "modF1" <interactive>:1:0: No instance for (Text.Regex.Base.RegexLike.RegexContext Regex [Char] [[a]]) arising from a use of `aaa' at <interactive>:1:0-10 > bbb "modF1" ("modF1","","",[]) I just extract the functions from where clause of guessKeys. But why are there different running result ? Sincerely! ----- fac n = foldr (*) 1 [1..n] -- View this message in context: http://old.nabble.com/About-Regex-question-tp26211696p26211696.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

2 1

Re: What's the deal with Clean?
by John Lato 05 Nov '09

05 Nov '09

Hi Bulat, I hope you don't mind me asking this, but when you state that "idiomatic haskell is much slower than idiomatic C", what do you mean by "much slower"? 2 times? 3 times? 20 times? Order of magnitude? I'm not going to disagree with you, however it seems to me the gap is closing. My opinion may be colored by having just examined Masayuki Takagi's SPH code, which I considered very idiomatic Haskell and for me performed equivalently to his C++ code with the changes I detailed in an earlier email (none of which involved any substantial changes to his code or style in my opinion). Best, John > From: Bulat Ziganshin <bulat.ziganshin(a)gmail.com> > > i don't say that haskell is slow for real programs. i say that > idiomatic haskell is much slower than idiomatic C and these papers > demonstrates how much, for simple computation-intensive loops > > -- > Best regards, > Bulat mailto:Bulat.Ziganshin@gmail.com >

2 2

Base classes can be _ELIMINATED_ with interfaces
by Shelby Moore 05 Nov '09

05 Nov '09

I was correct before, except I conflated the word "extended" with "eliminated" in my mind: http://lambda-the-ultimate.org/node/1277#comment-51723 The most robust solution to Tim Sweeney's problem is to rethink what a "class" should be: http://www.haskell.org/haskellwiki/?title=Why_Haskell_matters&oldid=24286#H… In Haskel, subtyping is done with Module. A "type class" is a polymorphic (relative to data type) interface, and the polymorphism is strictly parameterized for the client/consumer of the interface, i.e. the data type is known to the function that inputs the interface AT COMPILE TIME: http://www.haskell.org/~pairwise/intro/section2.html#part3 If all the functions/methods of package Engine input exclusively interfaces (i.e. Engine.IActor), then client package GearsOfWar can create a new subtype GearsOfWar.Actor which also implements the interface Engine.IActor (and perhaps even more interfaces). A problem with virtual (runtime pointer) inheritance is that it hides the subclass from the compiler. Polymorphic interfaces are missing from C++ and one can sort of hack an emulation with abstract classes, private delegates, and templates: http://lambda-the-ultimate.org/node/1277#comment-51806 In Haskell, parameterized polymorphic interfaces can be inherited, and this is knowable to the client function at compile time. Fundamental theorems tell us that any membership rule for a set can not be an absolute point, thus we should strive for typing and classing architecture which is the most granular (these are my summaries): * Russell's Paradox: there is no rule for a set that does not cause it to contain itself, thus all sets are infinitely recursive. * Liskov Substition Principle: it is an undecidable problem that subsets inherit. * Linsky Referencing: it is undecidable what something is when it is described or perceived. * Coase Theorem: there is no external reference point, any such barrier will fail. * Godel's Theorem: any formal theory, in which all arithmetic truths can be proved, is inconsistent. * 1856 Thermo Law: entire universe (a closed system, i.e. everything) trends to maximum disorder.

2 6

ANNOUNCE: feldspar-compiler
by Emil Axelsson 05 Nov '09

05 Nov '09

On behalf of my project colleagues at ELTE University, I'm happy to announce the release of feldspar-compiler: http://hackage.haskell.org/package/feldspar-compiler This package contains the C code back-end for the Feldspar EDSL, announced yesterday: http://hackage.haskell.org/package/feldspar-language For more information, see http://feldspar.sourceforge.net/ / Emil

1 0

representing Haskell objects in a uniform way
by Pasqualino "Titto" Assini 05 Nov '09

05 Nov '09

Hi, I am writing a little IPC system to make Haskell values and functions remotely invokable. To do so, I need (or so I believe) to make my objects accessible via a generic interface as in: class AFun f where afun :: Data a => f -> ([Dynamic] -> a) So my generic object is something that takes an array of parameters, that being Dynamic can be anything, and returns a Data, that I can easily serialise and send back on the wire. I start by defining an instance for functions: instance (Typeable a,AFun b) => AFun (a->b) where afun f (p:ps) = let Just v = fromDynamic p in afun (f v) ps afun _ _ = error "Too few arguments" So far so good, but when I try to define an instance for values: instance Data v => AFun v where afun f [] = f afun _ _ = error "Too many arguments" I get: Couldn't match expected type `a' against inferred type `v' `a' is a rigid type variable bound by the type signature for `afun' at /home/titto/.quid2/state/ubuntu.local.8080/wikidata/haskell/package/haskelld/src/HaskellD/Test.hs:7:17 `v' is a rigid type variable bound by the instance declaration .... Why is that? a and v are both declared to be a Data, why should they not match? The full code follows: {-# LANGUAGE FlexibleInstances ,UndecidableInstances ,OverlappingInstances #-} import Data.Data import Data.Dynamic class AFun f where afun :: Data a => f -> ([Dynamic] -> a) instance (Typeable a,AFun b) => AFun (a->b) where afun f (p:ps) = let Just v = fromDynamic p in afun (f v) ps afun _ _ = error "Too few arguments" instance Data v => AFun v where afun f [] = f afun _ _ = error "Too many arguments" Thanks in advance, titto

3 3

debugging
by Mohamed Ayed 05 Nov '09

05 Nov '09

can you please let me know how to debug Haskell programs? I got the observe module but it is complainnig about some "concurrent " moduel that it can not find . Also, how to change the path that Hugs uses to search for files ?

2 1

question
by Mohamed Ayed 05 Nov '09

05 Nov '09

Hi, I want to write a program that will take an AST and a cost vector that represents a cost for each operation and produce a simplified version of the tree based on cost reduction. Can any one give me some ideas ? Thanks

2 2

Re: How to optimize the performance of a code in Haskell?
by John Lato 05 Nov '09

05 Nov '09

For me, the Haskell version is about 1.5x slower than the C++ version, 19.6 seconds compared to 12.9. There wasn't an appreciable difference between -O and -O2. Which compiler (and version) are you using? I got: john$ g++ -O2 -o sph sph.cpp john$ time ./sph 300 real 0m12.914s user 0m11.015s sys 0m0.326s john$ g++ --version i686-apple-darwin9-g++-4.0.1 (GCC) 4.0.1 (Apple Inc. build 5490) Copyright (C) 2005 Free Software Foundation, Inc. and john$ ghc --make -O2 sph.hs john$ time ./sph 300 real 0m19.606s user 0m18.933s sys 0m0.304s john$ ghc --version The Glorious Glasgow Haskell Compilation System, version 6.10.4 on a recent model MacBook, OSX 10.5.8. Cheers, John > > Message: 4 > Date: Wed, 04 Nov 2009 17:53:49 +0100 > From: Kalman Noel <noel.kalman(a)googlemail.com> > Subject: [Haskell-cafe] Re: How to optimize the performance of a code > in Haskell? > To: Masayuki Takagi <kamonama2(a)gmail.com> > Cc: Haskell Cafe <haskell-cafe(a)haskell.org> > Message-ID: <4AF1B19D.2010403(a)googlemail.com> > Content-Type: text/plain; charset=ISO-8859-15; format=flowed > > (I take it you accidently wrote to fa.haskell, which is just a mirror of -cafe > and -beginners, so I'm cc-ing the Café with a full quote.) > > Masayuki Takagi: >> I'm writing fluid simulation programs with SPH(Smoothed particle >> hydrodynamics) in Haskell and C++. (The purpose that I write in two >> languages is to make a workflow that first i write in Haskell for >> rapid prototyping and accuracy then rewrite in C++ for performance.) >> >> I've compared them in performance. Then, although I have already done >> optimization with profiler, the Haskell code is 20 times slower than >> the C++ code. >> >> I think the performance of the Haskell code is so slow that there can >> be room for optimization. I'm happy if the Haskell code work 3 times >> slower than the C++ code at worst. >> >> How can I make the Haskell code faster? >> What information should I refer? >> >> The codes are here: >> http://kamonama.sakura.ne.jp/sph/20091101/sph.hs.zip >> http://kamonama.sakura.ne.jp/sph/20091101/sph.cpp >> >> To run the code in Haskell: >> $ ghc --make -O sph.hs >> $ ./sph 300 >> (300 is the time step to be conputed) >> >> To run the code in C++: >> $ g++ -O2 -o sph sph.cpp >> $ ./sph 300 >> (300 is the time step to be conputed) >> >> thanks >> > > I've not looked at the code, but you'll want ghc to do better optimizations > than -O. -O2 is what you should use in general. Also, number-crunching often > profits from -fexcess-precision. >

1 1

Interactive chatbot
by Torsten Otto 04 Nov '09

04 Nov '09

Hi! My students have the task to program an interactive chatbot. We have run into a problem that I can't solve either: When we read the user's input through > t <- getLine it is not possible to delete typos before hitting enter and thereby sending the input off to the system (at least in OS X, bash). I didn't find that terribly problematic, but of course it is a bit of a show stopper from their point of view. The input is then used to generate a reply in purely functional code, and the reply sent to the command line via putStr. Is there a more clever way to interact with the user that would allow editing ones text before sending it to the bot? I guess we could try with a website, but don't know off hand how to do that, either, although I've seen beautiful webservers made in Haskell... Regards, Torsten Otto

5 4

Announce: language-python version 0.2 now available
by Bernie Pope 04 Nov '09

04 Nov '09

I'm pleased to announce that version 0.2 of the language-python package is now available on hackage: http://hackage.haskell.org/package/language-python language-python provides lexical analysis and parsing for Python. Major features of this release: - Support for versions 2.x and 3.x of Python (previously only 3.x was supported). - Lexical tokens and AST nodes are annotated with accurate source span information. - Comments are retained as tokens, and are collected by the parser. Main shortcomings of this release: - Support for Unicode is limited (waiting on Unicode support in Alex). - It has only undergone minimal testing (testing infrastructure is still being built). I've also written a small client of the package, called language-python-colour, which renders Python source code as XHTML for colouring etc. The main purpose of this is to demonstrate how to use language-python, and the utility of accurate source spans. http://hackage.haskell.org/package/language-python-colour Example output: http://www.cs.mu.oz.au/~bjpop/code/lsystem.py.html Cheers, Bernie.

3 4

Emacs: Haskell snippets for YASnippet
by Daniel Schüssler 04 Nov '09

04 Nov '09

Hi List, this is rather trivial, but maybe someone else finds these useful: darcs get http://code.haskell.org/~daniels/haskell-snippets/ Especially the LANGUAGE ones have saved me quite some typing :) Additions welcome. Usage: If not already installed, get YASnippet: http://code.google.com/p/yasnippet/ and put this into your .emacs: (load-file "some-path/haskell-snippets.el") to expand a snippet, just enter the macro string (these are listed in the haskell-snippets.el file) and press <tab>. If the snippet has holes, press <tab> again to jump to the next hole. Greetings, Daniel

3 4

ANN: fdo-notify 0.1, a client for the Desktop Notifications protocol
by Max Rabkin 04 Nov '09

04 Nov '09

Haskellers, I present to you fdo-notify, a client library for FreeDesktop.org's Desktop Notifications protocol. This is the DBUS protocol served by NotifyOSD and other notifications systems, which allows a wide variety of applications to present notifications to the user in a uniform way. The library should not require knowledge of DBus or the Desktop Notifications protocol itself, at least once fully documented (there should be enough haddocumentation for basic use already). Hackage: http://hackage.haskell.org/package/fdo-notify Mercurial: http://bitbucket.org/taejo/fdo-notify/ Basic notifications and updated/replacement notifications are supported. Images are not yet supported (adding support should be easy, but which Image type? Is there a de facto standard imaging library like the Python Imaging Library?) and nor are signals. Signals allow an application to be informed when some action is taken on the notification (NotifyOSD doesn't support actions, but other implementations might). The library is released under the 3-clause BSD license, but depends on the dbus-core and dbus-client libraries which are GPLed. I hope you find this library useful. Regards, Max

1 0

Fwd: [Haskell-cafe] What's the deal with Clean?
by Alberto G. Corona 04 Nov '09

04 Nov '09

Artyom. I know what uniqueness means. What I meant is that the context in which uniqueness is used, for imperative sequences: (y, s')= proc1 s x (z, s'')= proc2 s' y ..... is essentially the same sequence as if we rewrite an state monad to make the state explicit. When the state is the "world" state, then it is similar to the IO monad. An state monad forces a single use of the implicit state variable too (unless you pass it trough the next step without changes. That can be done in Clean too. 2009/11/4 Artyom Shalkhakov <artyom.shalkhakov(a)gmail.com> Hello, > > 2009/11/4 Alberto G. Corona <agocorona(a)gmail.com>: > > The code executed by uniqueness types is somehow similar to the internal > > code executed in a state monad (or in the case of IO, the IO monad). The > > main difference is that the pairs of results (state, value) are > explicitly > > written in Clean by the programmer and the type sytem assures that the > > order of executions makes sense at compile time, whereas in the case of > the > > state monad the sequence of instructions is lazily assembled at runtime > in > > the first step and executed in a second step. So there is a little more > > overhead in haskell but the code is higher level. > > Am I right? > > I would rather say: code with uniqueness types allows for safe > destructive updates. > > In Clean, a variable of unique type is ensured to have only one > reference to it, at any time (that's why it's called "uniqueness > typing"). So you can't write the code like this > > > f(x) + f(x) > > where f : *a -> int (x is of unique type), because x is clearly > referenced two times here. What to do? Let f yield another reference > to x! That also means that the old reference is not usable any more, > since you have new one. f becomes: > > > f : *a -> (int, *a) > > and the code looks very familiar: > > > let (a, x') = f(x) > > (b, x'') = f(x') > > in a + b > > The function f can use destructive updates under the hood though it > doesn't violate referential transparency. I bet you can you see why. > > I'd say that call-by-need is orthogonal to uniqueness typing. > > Cheers, > Artyom Shalkhakov. >

3 3

Re: How to optimize the performance of a code in Haskell?
by Kalman Noel 04 Nov '09

04 Nov '09

(I take it you accidently wrote to fa.haskell, which is just a mirror of -cafe and -beginners, so I'm cc-ing the Café with a full quote.) Masayuki Takagi: > I'm writing fluid simulation programs with SPH(Smoothed particle > hydrodynamics) in Haskell and C++. (The purpose that I write in two > languages is to make a workflow that first i write in Haskell for > rapid prototyping and accuracy then rewrite in C++ for performance.) > > I've compared them in performance. Then, although I have already done > optimization with profiler, the Haskell code is 20 times slower than > the C++ code. > > I think the performance of the Haskell code is so slow that there can > be room for optimization. I'm happy if the Haskell code work 3 times > slower than the C++ code at worst. > > How can I make the Haskell code faster? > What information should I refer? > > The codes are here: > http://kamonama.sakura.ne.jp/sph/20091101/sph.hs.zip > http://kamonama.sakura.ne.jp/sph/20091101/sph.cpp > > To run the code in Haskell: > $ ghc --make -O sph.hs > $ ./sph 300 > (300 is the time step to be conputed) > > To run the code in C++: > $ g++ -O2 -o sph sph.cpp > $ ./sph 300 > (300 is the time step to be conputed) > > thanks > I've not looked at the code, but you'll want ghc to do better optimizations than -O. -O2 is what you should use in general. Also, number-crunching often profits from -fexcess-precision.

2 1

multi-line regex
by Michael Mossey 04 Nov '09

04 Nov '09

I have some very simple regex-matching needs, and Text.Regex.Posix will work fine, EXCEPT I need to match multi-line patterns, and/or find all occurrences of text that may occur several times on different lines. So I need to turn on some kind of flag. Can someone show me how to do that? I have worked the examples in RWH so I basically know how to run the thing.

4 5

ANN: haskell-src-exts-1.3.0
by Niklas Broberg 04 Nov '09

04 Nov '09

Fellow Haskelleers, I'm pleased to announce the release of haskell-src-exts-1.3.0! * On hackage: http://hackage.haskell.org/package/haskell-src-exts * Via cabal: cabal install haskell-src-exts * Darcs repo: http://code.haskell.org/haskell-src-exts Version 1.3.0 is a new major release, following the PVP, as it contains a few backwards-incompatible changes, making the 1.2.x branch a very short parenthesis in history. There are two main new things in 1.3.0: fixity application is now handled uniformly regardless of which AST you use, simple or annotated, and the parser now supports multiple modules in the same source file. haskell-src-exts-1.3.0: ==================== * The *.Fixity modules now export the same Fixity type (incidentally the one previously used by the simple un-annotated parsing) and the same helper functions. If you've done all your Fixity handling using the helper functions (infix_, infixl_, infixr_), or if you've only used fixities for the un-annotated AST, you're safe. Only if you've manually created values of the annotated Fixity type will you need to pay attention. * A new entry point in the parser allows multiple modules to be parsed from the same source file ( parseModules[With[Mode|Comments]] ). * SpecialCon is now an instance of Pretty (don't ask me why it wasn't before). * fromParseResult now displays the filename in case of a failed parse. * A few bug fixes to the exact-printer (that not many seem to be using just yet). Please help me test and report! Grab a darcs version, put your source files in the Test/examples dir, and go cabal test (in the top dir). Any failing cases not due to CPP or literate source files (for the exact-printer), please report to the trac: http://trac.haskell.org/haskell-src-exts Cheers, /Niklas

2 1

IORefs and weak pointers
by Patai Gergely 04 Nov '09

04 Nov '09

Hello all, I wanted to create a weak pointer with an IORef as the key and something else as the value, but I saw no way to do it through the API provided. After some experimentation I came up with the following abomination for a solution: myWeakRef (IORef (STRef r)) v f = IO $ \s -> case mkWeak# r v f s of (# s', w #) -> (# s', Weak w #) This works perfectly when the code is compiled both with and without optimisations, but ghci chokes on it with an internal error. So my question is if I can expect this to work at least this much in the long run, or is it a hopelessly fragile hack? Gergely -- http://www.fastmail.fm - Same, same, but different...

3 5

Re: [Haskell-cafe] Best Editor In Windows
by Tom.Amundsen 04 Nov '09

04 Nov '09

Nathan P. Campos wrote: > > Hello, > I'm very new at Haskell, i'm reading a book and starting, but i want to > know which is the best editor for development under Windows, because now > i'm using Notepad++(That i use to develop in C++). > > Thanks, > Nathan Paullino Campos > If you use emacs, check out http://www.haskell.org/haskell-mode/ -- View this message in context: http://www.nabble.com/Best-Editor-In-Windows-tp25908187p25917670.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

11 18

Gauss Elimination -> More Clean2Haskell
by Philippos Apolinarius 04 Nov '09

04 Nov '09

I am keeping with my project of translating programs from Clean to Haskell. As far as arrays go, I don't understand well how to use them in Haskell. Therefore, I will appreciate if somebody can find time to check the program below, and make suggestions to improve it. My Haskell program is about 4 times slower than the Clean version. It would be great if one could reduce the execution time by half, approaching to the speed of Clean and Scheme. Here are the constraints: 1 --- The program must be implemented using arrays. Update must be done in place, in order to minimize use of the garbage collector. I have used Data.Array.IO, but I guess that Data.Array.ST is better. Is it easy to rewrite the program in order to use DataArray.ST? 2 -- I liked very much the forM_ monad. I wonder if there is an accumulating monad that play the role of a program like leftSide. 3 -- Clean program almost double its speed, if one uses strictness annotations. Is it possible to use similar anotations for Haskell? Here is how I compiled the program: ghc -O2 gel.hs --make In order to run the program with 2000 equations, type gel.exe 2000 +RTS -sstderr The program will create a linear system with 2000 equations so that all elements of the solution is equal to 1. It prints 20 elements of the solution. Here is the program: {- File: gel.hs Compilation: ghc -O2 gel.hs --make Run: time gel.exe 2000 -} import Control.Monad import Data.Array.IO import System.IO import System.Random import System (getArgs) prtSol i n1 arr | i < 1= return () prtSol i n1 arr= do b <- readArray arr (i, n1) putStr ((show b)++" ") prtSol (i-1) n1 arr fillArray xs s (i, n) (j, m) arr | i > n= return () fillArray xs s (i,n) (j, m) arr | i==n && j>m= do writeArray arr (i, j) s return () fillArray xs s (i, n) (j, m) arr | j > m = do writeArray arr (i, j) s fillArray xs 0.0 (i+1, n) (1, m) arr fillArray (val:xs) s (i, n) (j, m) arr= do writeArray arr (i, j) val fillArray xs (s+val) (i, n) (j+1, m) arr sLU arr n= sIJ 2 1 2 n arr sIJ i j k n arr | i > n = return () sIJ i j k n arr | k > n = sIJ (i+1) i (i+1) n arr sIJ i j k n arr = do im <- pmax (j+1) j swap j im 1 a <- readArray arr (k, j) forM_ [j..n+1] $ \l -> do ajj <- readArray arr (j, j) ajl <- readArray arr (j, l) akl <- readArray arr (k, l) writeArray arr (k, l) (akl-a*(ajl/ajj)) sIJ i j (k+1) n arr where pmax line imax | line > n = return imax pmax line imax = do alj <- readArray arr (line, j) aij <- readArray arr (imax, j) if (abs alj)> (abs aij) then pmax (line+1) line else pmax (line+1) imax swap r s q | q>n+1 = return () swap r s q | r==s = return () swap r s q = do arq <- readArray arr (r,q) asq <- readArray arr (s,q) writeArray arr (s,q) arq writeArray arr (r,q) asq swap r s (q+1) leftSide acc i j n arr | j>n= return acc leftSide acc i j n arr = do v <- readArray arr (j, n+1) a <- readArray arr (i, j) leftSide (acc-v*a) i (j+1) n arr solvit i n arr | i<1= return () solvit i n arr= do a <- readArray arr (i, i) acc <- readArray arr (i, n+1) v <- leftSide acc i (i+1) n arr writeArray arr (i, n+1) (v/a) solvit (i-1) n arr rnList :: (Double, Double) -> IO [Double] rnList r=getStdGen>>=(\x->return(randomRs r x)) dims [input] = (read input, read input) dims _ = (1000, 1000) main = do xs <- rnList (1.0,1000.0) args <- getArgs let (n, m)= dims args arr <- newArray_ ((1,1),(n,m+1)) :: IO (IOUArray (Int, Int) Double) fillArray xs 0.0 (1,n) (1,m) arr sLU arr n solvit n n arr prtSol (min 20 n) (n+1) arr print "Done" __________________________________________________________________ Be smarter than spam. See how smart SpamGuard is at giving junk email the boot with the All-new Yahoo! Mail. Click on Options in Mail and switch to New Mail today or register for free at http://mail.yahoo.ca

4 3

Fwd: [Haskell-cafe] What's the deal with Clean?
by Alberto G. Corona 04 Nov '09

04 Nov '09

The code executed by uniqueness types is somehow similar to the internal code executed in a state monad (or in the case of IO, the IO monad). The main difference is that the pairs of results (state, value) are explicitly written in Clean by the programmer and the type sytem assures that the order of executions makes sense at compile time, whereas in the case of the state monad the sequence of instructions is lazily assembled at runtime in the first step and executed in a second step. So there is a little more overhead in haskell but the code is higher level. Am I right? 2009/11/4 wren ng thornton <wren(a)freegeek.org> > Stephen Tetley wrote: > >> 2009/11/3 Andrew Coppin <andrewcoppin(a)btinternet.com>: >> >> As far as I can tell, Clean is to Haskell as C is to Pascal. I.e., Clean >>> is >>> notionally very similar to Haskell, but with lots of added clutter, >>> complexity and general ugliness - but it's probably somehow more >>> machine-efficient as a result. >>> >>> (All of which makes the name "Clean" rather ironic, IMHO.) >>> >> >> OUuch - you really could put it the other way round. >> > > Part of this really comes down to how one feels about the monads vs > uniqueness types argument. It's a silly argument to have since the ideas are > orthogonal and only really intersect at IO, but there's history there which > lead to the current state of things. > > Sometimes in Haskell I've thought about how uniqueness typing would make > something faster, but in general all the plumbing associated with it in > Clean makes me feel like I'm writing systems-level code (i.e. C, asm) > instead of using a high-level language. The extra plumbing really makes it > feel dirtier to work with. That doesn't mean Clean is bad, but I think it > does contribute to the "cluttered" feeling Haskellers get. > > But as I said, it's a silly argument and folks should use whichever gives > them warm fuzzies. I also have a vague unnameable distaste whenever working > with Python, and rather enjoy working with Perl. Nobody's perfect :) > > -- > Live well, > ~wren > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe >

2 1

Essence of Functional Programming for Imperative Programmers
by Shelby Moore 04 Nov '09

04 Nov '09

New concise guide I am creating: http://www.coolpage.com/commentary/economic/shelby/Functional_Programming_E… I was able to condense Category theory and implementation of Monads to one screen: http://www.coolpage.com/commentary/economic/shelby/Functional_Programming_E… Overall, I think I have a unique method for comparing and condensing the explanation of the transistion from imperative to pure functional. What you think? It is a work-in-progress, so corrections, feedback, and flames are welcome. I will do the OOP section next and incorporate the explanations from these posts: http://www.haskell.org/pipermail/haskell-cafe/2009-November/068440.html (data vs. Module) http://www.haskell.org/pipermail/haskell-cafe/2009-November/068432.html (interface vs. virtual) P.S. The link will not change if you want to link to it now. If you mirror it, please update your mirrors periodically. There is no copyright claimed, I don't believe in copyrights any more. I intend to publish everything as PUBLIC DOMAIN (i.e. no license at all, because licenses impact composability). If I want to charge, I will put functionality behind an unpublished interface (i.e. Module). P.P.S. I only started learning functional programming about a week ago.

1 0

Word128, Word256
by Maurício CA 04 Nov '09

04 Nov '09

Hi, Do you think we could have the range of sizes for Int* and Word* expanded to allow also 128 and 256 bits sizes? My reason is that I have a long standing issue trying to bind to C numerical libraries using complex numbers, as those are usually structs passed by value. See this from GNU GSL: typedef struct { double dat[2]; } gsl_complex; I imagine I could do: type GslComplex = Word128 -- size would depend on architecture and then write foreign declarations to have: get_real :: GslComplex -> CDouble get_imag :: GslComplex -> CDouble gsl_complex_sin :: GslComplex -> GslComplex gsl_complex_cos :: GslComplex -> GslComplex Do you think this is a reasonable request? Thanks for your attention, Maurício

4 8

Hugs Trex for GHC / JHC / UHC / ... ?
by Peter Verswyvelen 04 Nov '09

04 Nov '09

I briefly read about Hugs record system Trex, and at first sight this was really nice! I know this question was asked a long time ago already, but are there any plans to add this extension to GHC? What about the other compilers, like JHC, UHC, etc? Is it possible to emulate Trex (lots of other record systems exist), *and* get the same easy to understand errors? What were the problems with Trex (if any)? Thanks, Peter Verswyvelen

2 1