January 2010 - Haskell-Cafe

space leaks and optimizations
by Alexei Kitaev 15 Jan '10

15 Jan '10

Dear Haskellers, Recently I have been looking for a programming language that would be suitable for small scientific and recreational projects and palatable to a picky person like me. (I do theoretical physics and some math; I do not program very often.) Haskell and Clean look attractive from a mathematician's point of view and allow for very elegant solutions in some cases. I tried Clean first because it has a more efficient implementation and better array support. However, I am leaning toward Haskell since it has more libraries, larger community, and some nice features (e.g., the "do" notation). I thought it would be easier to avoid errors when programming in a pure functional language. Indeed, the Haskell type system is great and catches many errors. But in practice, I found it very difficult to write correct programs in Haskell! This is because my definition of correctness includes asymptotic complexity (time and space). I have pondered why Haskell is so prone to space leaks and whether this can be fixed. I posted a related message (describing a space leak caused by inlining) to Glasgow-haskell-users, http://www.haskell.org/pipermail/glasgow-haskell-users/2009-November/018063… but apparently the GHC developers were busy preparing a new release. Perhaps on Haskell-cafe there are more people with some spare time; I would really appreciate your comments. So, there seem to be several reasons why controlling space usage is difficult: 1) The operational semantics of Haskell is not specified. That said, it seems that unoptimized programs behave in a predictable way if you follow the execution step by step. The Clean report explicitly says that the execution model is graph reduction; I believe that Haskell uses the same model. However, there are some subtleties, e.g., the tail call and selector optimizations. (I read about the selector optimization in Wadler's paper, http://homepages.inf.ed.ac.uk/wadler/topics/garbage-collection.html and saw it mentioned on the GHC development page. It's really nice and indispensable, but it seems to be missing from the user documentation. Is this the following description accurate? After the rhs of a lazy pattern like (a,b) = expression has been evaluated, the pattern does not take up any space, and the space occupied by a and b can be reclaimed independently.) There must be more subtleties. I imagine that a rigorous definition of operational semantics would be too complicated and impractical. But maybe an informal specification is better than nothing? Why people have not attempted to write it? 2) While each step is predictable, the overall behavior of a lazy program can be rather surprising. So one must be very careful. GHC provides two ways to control the evaluation order, seq and bang patterns, but I am not sure which of these (if any) is the right tool. Consider the following example (from the Real World Haskell book): mean :: [Double] -> Double mean xs = sum / fromIntegral num where (num,sum) = foldl' f (0,0) xs :: (Int, Double) f (n,s) x = (n+1, s+x) Although it uses foldl', there is a space leak because Haskell tuples are not strict. There are two possible fixes: f (!n,!s) x = (n+1, s+x) or f (n,s) x = let n1=n+1; s1=s+x in n1 `seq` s1 `seq` (n1,s1) The first one is short, but less natural than the second one, where the offending thunks are suppressed on the spot. Unfortunately, the syntax is awkward; it would be nice to write something like f (n,s) x = (!n+1, !n+1) Well, I am becoming too grumpy, perhaps the bang patterns are fine. More important question: are there established practices to *avoid* space leaks rather than fixing them afterwards? 3) The standard library was not designed with space efficiency in mind; for example, there is sum but no sum'. 4) GHC does a pretty good job optimizing programs for speed, but compiling with the -O option can easily introduce a space leak. I have not encountered such problems with Clean, though my experience is very limited. Apparently, the Clean developers have disallowed some unsafe optimization, see e.g., this message: http://mailman.science.ru.nl/pipermail/clean-list/2009/004140.html I doubt that the Clean solution is 100% reliable because the compiler still uses strictness analysis, which can change the asymptotic space usage (usually for better, but sometimes for worse). So I wonder whether it would be feasible to identify a set of conservative optimizations that do not change the space or time usage more than by a constant factor. For example, the strictness analysis could be limited to fixed-size data. Or instead of strictness analysis, one could inline the top-level patterns of every function. Of course, that would make the optimization less efficient on the average, but predictability is more important. Best regards, Alexei

4 4

Force -threaded from a library
by John Van Enk 15 Jan '10

15 Jan '10

Hi List, Is it possible to prevent a library from being used unless -threaded is enabled? I have a specific case where lots-of-nasty shows up if the library is linked against an executable built without -threaded. I suppose this is GHC specific. /jve

2 2

deleteBy type too restrictive
by Dan Rosén 15 Jan '10

15 Jan '10

Hello, I realized today that the type for deleteBy in Data.List is too restrictive. The code is: deleteBy :: (a -> a -> Bool) -> a -> [a] -> [a] deleteBy _ _ [] = [] deleteBy eq x (y:ys) = if x `eq` y then ys else y : deleteBy eq x ys though the type deleteBy :: (b -> a -> Bool) -> b -> [a] -> [a] will do good as well. Is there a particular reason that the type has this restriction? Otherwise, where can I post a suggestion to have it untightened? Best regards, Dan Rosén references: http://haskell.org/ghc/docs/latest/html/libraries/base-4.2.0.0/Data-List.ht… http://www.haskell.org/ghc/docs/latest/html/libraries/base-4.2.0.0/src/Data…

2 1

New blog: mainly Haskell and wxHaskell related
by Jeremy O'Donoghue 15 Jan '10

15 Jan '10

Hi all, I have just started a new blog at http://wewantarock.wordpress.com. At least initially, my intention is to provide a set of 'tutorial' style articles to help new wxHaskell users get beyond the 'Hello World' stage by offering a few worked examples on things you might want to do to put together a usable application. The idea is to offer something which fits between the first tutorial and the reference documents. The first few postings will explain the anatomy of a complete custom wxHaskell control written in Haskell. Comments gratefully accepted, including (gentle please - I'm not dons!) criticism of my Haskell style. Regards Jeremy O'Donoghue

2 1

Quick, somebody do something!
by Henk-Jan van Tuyl 15 Jan '10

15 Jan '10

Haskell has dropped out of the top 50 at Tiobe [1]; how could this hapen? Let's start selling mobile phones that can only be programmed in Haskell :-) [1] http://www.tiobe.com/content/paperinfo/tpci/index.html -- Met vriendelijke groet, Henk-Jan van Tuyl -- http://Van.Tuyl.eu/ http://members.chello.nl/hjgtuyl/tourdemonad.html --

5 4

Typed Configuration Files
by Sebastian Fischer 15 Jan '10

15 Jan '10

Dear Café, Neil Mitchell's cmdargs package [1] is pretty neat. It can be used to parse command-line arguments into a user-defined data structure. Is there something similar for parsing config files? There are a number of config file parsers on Hackage. But even the most sophisticated one I found, ConfigFile by John Goerzen [2], only yields primitive data like strings, booleans, and numbers. Did I overlook something? I'd like to write something like do fc <- configFile ... ac <- cmdArgs ... let conf = fc `mappend` ac where the type of `conf` is a user defined monoid. I found that the fez-conf package [3] provides a function `parseToArgs` which creates a list similar to the one returned by `System.getArgs` from a config file. Neil, if you would add a function to your cmdargs package that allows to specify the argument list, then one could reuse your machinery to create typed config data from config files too, right? Maybe along the lines of do args <- parseToArgs <$> readFile "/my/conf" conf <- cmdArgsWithDefault args "My Program" [myMode] The new function `cmdArgsWithDefault` could require the return type to be a monoid in order to allow users to specify themselves how to deal with multiple options of the same kind. Would that be reasonable or is there a better alternative? Cheers, Sebastian [1] http://hackage.haskell.org/package/cmdargs [2] http://hackage.haskell.org/package/ConfigFile [3] http://hackage.haskell.org/package/fez-conf -- Underestimating the novelty of the future is a time-honored tradition. (D.G.)

7 10

Design question, HTML for GUIs?
by Günther Schmidt 15 Jan '10

15 Jan '10

Hi everyone, as probably most people I find the GUI part of any application to be the hardest part. It just occurred to me that I *could* write my wxHaskell desktop application as a web app too. When the app starts, a haskell web server start listening on localhost port 8080 for example and I fire up a browser to page localhost:8080 without the user actually knowing too much about it. Is that a totally stupid idea? Which haskell web servers would make good candidates? Are there any *continuation* based web server in haskell, something similar to Smalltalk's Seaside? Is Hyena continuation-based? Günther

7 7

ANN: Lite Haskell IDE
by Mambo Banda 15 Jan '10

15 Jan '10

Hi Just started an open source project called Hoodoo. It is meant to be a Lite IDE for Haskell. I wrote it to learn Haskell, so if you are a beginner (like me) it might interest you. If you're interested follow this link (http://hoodoo.kenai.com/) and if you like what you see, grab the code and add something:) Mambo

1 0

Haskell implementation of ideas from StandardML as a Metaprogramming language
by CK Kashyap 15 Jan '10

15 Jan '10

Hi All, I was just going over the paper titled - "Standard ML as a meta programming language" by Samuel Kamin - It has a few ideas of generating C++ code from ML. The first one being generating C++ top down parser. I wanted to try out the sample in Haskell - I was wondering if anyone's already done that - I could just look at that implementation for reference. Regards, Kashyap

2 2

General Advice Needed ..
by Ian675 14 Jan '10

14 Jan '10

Hi, First of all, sorry if its in the wrong section.. But I'm just having trouble getting to grips with Haskell. I have my functional programming exam tommorow and I'm struggling to understand any of this. We worked through the book The Craft Of Functional Programming and Im trying to work my way through it but still no luck.. I find it really hard to put down code onto paper. I can understand the code in the book I can do the simple simple functions like cube.. cube :: Int -> Int cube n = n*n*n But after that im lost :( Is there any general advice? Just keep reading the book till it drills into my big head? You can make sarcastic n00b jokes if you like.. I'm just a very frustrated student right now! -- View this message in context: http://old.nabble.com/General-Advice-Needed-..-tp27161410p27161410.html Sent from the Haskell - Haskell-Cafe mailing list archive at Nabble.com.

8 10