September 2006 - Haskell-Cafe

Optimization problem
by Magnus Jonsson 29 Sep '06

29 Sep '06

Dear Haskell Cafe, When programming the other day I ran into this problem. What I want to do is a function that would work like this: splitStreams::Ord a=>[(a,b)]->[(a,[b])] > splitStreams [(3,x),(1,y),(3,z),(2,w)] [(3,[x,z]),(1,[y]),(2,[w])] I don't care about the order that the pairs are output, so the answer could just as well be [(2,[w],(3,[x,z]),(1,[y])]. However I do care about the order of the xyzw:s, so (3,[z,x]) could not be part of the solution in this example. Furthermore it should work on infinite lists. It can't eat the whole list before producing any output. Now, it's not too hard to come up with an inefficient solution that traverses the input list multiple times. For example a sieving solution: import Data.List splitStreams [] = [] splitStreams ((channel,msg):rest) = let (myMsgs,otherMsgs) = partition (\(c,_)->c==channel) rest in (channel, msg : map snd myMsgs) : splitStreams otherMsgs I'm afraid this algorithm is O(n*m) time in the worst case, where n is the length of the input list, and m is the number of unique channels. But is there any way to write it such that each element is touched only once? Or at least an O(n*log(m)) algorithm? Any takers? / Magnus Jonsson

14 43

Optimizing a title matcher
by Lyle Kopnicky 29 Sep '06

29 Sep '06

Hi folks, I'm competing in a contest at work, and we're allowed to use whatever language we want. I decided this was my chance to prove to people that Haskell was up to the challenge. Unfortunately, I ran into performance problems. Since the contest ends this Friday, I've decided to switch to C++ (gasp!). But if any of you have advice on how to speed up this code, it could help me advocate Haskell in the future. It's supposed to match movie titles from an imported database to a reference database. The version I've sent doesn't do anything very smart - it's just doing literal title matches. The first argument to the program is the filename of the table to be imported, and the second is the filename of the reference table. The first line of each table is a pipe-separated list of field names; the rest of the lines are records, each a pipe-separated list of values. The import files each have 3,000 records, and the reference table has 137,986 records. Building the hash tables out of the files is quick - it just takes a few seconds. But doing the matching of title_id in one table to title_id in the other, in a nested loop between both tables, takes way too long. It's matching two import titles (against each of the reference titles) per second. It needs to do at least 20 per second to qualify for the contest, and it's not doing anything fancy yet. I tried various "improvements" to speed it up. One was to specifically use ByteString, eliminating the AbsString class. Didn't make a difference. Another was to use arrays instead of lists to store each record, and precompute the indices of each of the fields within those records. I also iterated over a list of keys instead of the list of Maps, and only converted each record to a Map one at a time, hoping they would be disposed of sooner. Instead of speeding up the program, this slowed it down by a factor of 20! I've profiled it, and I can't make much out of that. It seemed to be spending 25% of its time doing scoring, and I though the problem must be due to laziness, but I'm not sure. So if anyone has any ideas how to speed this up by a factor of at least 10 times, it would be really appreciated! Even the Ruby solutions are doing that, which is embarrassing. Thanks, Lyle

8 11

instance of String illegal
by Adam Hupp 29 Sep '06

29 Sep '06

I am trying to create an instance of a class for String types. I have the following: class Foo a where mkFoo :: a -> String instance Foo String where mkFoo x = x and receive the following error: test.hs:9:0: Illegal instance declaration for `Foo String' (The instance type must be of form (T a b c) where T is not a synonym, and a,b,c are distinct type variables) In the instance declaration for `Foo String' Failed, modules loaded: none. I see that in general I am not able to create an instance of a class for any type [a], where a is a specific type. I can work around it by wrapping my Strings in some other data type but I'd like to understand why it's not possible. Any insight would be appreciated. -Adam

4 3

Creating DLLs with GHC
by Matthew Bromberg 29 Sep '06

29 Sep '06

I am having some difficulty with creating a dynamic link library using GHC on windows XP. I am attempting to follow the example in http://www.haskell.org/ghc/docs/6.4/html/users_guide/win32-dlls.html though I have a binary build of ghc 6.5 My problem (I think) is that some of my Haskell source files link to an external C library also contained in a DLL. I generate my Haskell object files when I compile my Haskell executable. I then attempt to use the object files in a ghc compiler statement like this, ghc --mk-dll -o netsim.dll ExternLib.o ExternLib_stub.o dllNet.o src1.o src1_stub.o src2.o -optl-lmatrixstack -optl-L"." My external C library is in matrixstack.dll and it has a corresponding static link stub library matrixstack.lib in the same directory as all the sources. It has references in one of the sources (say src1.hs). Unfortunately I get a host of undefined references to basically all the functions in matrixstack.dll and also some undefined references of the form Parsefile.o:ghc2996_0.hc:(.text+0x130): undefined reference to `TextziParserCombinatorsziParsecziError_show_closure' Parsefile.o:ghc2996_0.hc:(.text+0x220): undefined reference to `TextziParserCombinatorsziParsecziChar_spaces_closure' Parsefile.o:ghc2996_0.hc:(.text+0x24a): undefined reference to `TextziParserCombinatorsziParsecziChar_spaces_closure' Parsefile.o:ghc2996_0.hc:(.text+0x29c): undefined reference to `TextziParserCombinatorsziParsecziChar_spaces_closure' ... This appears to be coming from unsatisfied references to Text.ParserCombinators.Parsec, which I thought was a standard library. I note that my stand alone haskell executable links just fine and runs perfectly. What am I missing to make this work?

5 9

flip dot
by Greg Fitzgerald 28 Sep '06

28 Sep '06

Since there's talk of removal of the composition operator in Haskell-prime<http://hackage.haskell.org/trac/haskell-prime/wiki/CompositionAsDot>, how about this: Instead of: foo = f . g you write: foo = .g.f A leading dot would mean, "apply all unnamed parameters to the function on the right". A trailing dot would mean, "apply the result of the left to the function on the right". Prelude> notNull = .null.not => [a] -> Bool notNull [1,2,3] => True [1,2,3].notNull => True [1,2,3].null.not => True I like this because it has the same perks as the composition operator, yet it looks like OO code and the data flows nicely from left to right. It reads especially well when using the bind operator on the same line. Thoughts? Thanks, Greg

8 9

Re: Curious Functor Class
by Ashley Yakeley 28 Sep '06

28 Sep '06

On Sep 28, 2006, at 00:38, Jeremy Gibbons wrote: >> Perhaps the key is that there exist types P and Q s.t. there's an >> isomorphism >> >> F a <=> (P -> a,Q) > > F is Naperian iff there's a P with F a = P -> a; but what's the Q for? > >> This seems to be intuitively Napierian: >> >> ln (P -> a,Q) = (P,ln a) | ln Q >> >> I can believe that Hoistables are in fact Idioms, though I know >> there are Idioms that are not Hoistables (Maybe and Either, for >> instance). > > That's right. Every Monad is an Idiom. So are constant functors (F > a = Int) - which I guess a Naperian anyway. Hoistables are not always Idioms, it turns out. I think this can be made a Hoistable, but not an Idiom (because it has a "Q"): type WithInt = (,) Int -- i.e. WithInt a = (Int,a) I don't know if that counts as Napierian or not. >> (Also I think "Idiom" is a better class name than "Applicative".) > > Me too! Can you tell Ross and Conor? I've tried... Hey Ross, Conor, "Idiom" is a better name than "Applicative". Pretty much everyone thinks so. > * > > I don't subscribe to haskell-cafe, so apparently I can't post. Feel > free to post my reply there, if you think it is useful. -- Ashley Yakeley

2 1

source code for haskell web server?
by mvanier 28 Sep '06

28 Sep '06

Does anyone know where I could find the source code for the Haskell web server described in the papers "Tackling the Awkward Squad" by SPJ and "Writing High-Performance Server Applications in Haskell, Case Study: A Haskell Web Server" by Simon Marlow? Thanks in advance, Mike

8 10

Eager global IO actions (per module initialization)
by Einar Karttunen 28 Sep '06

28 Sep '06

Hello I am needing a way to run initializers defined in various modules in an eager fashion before main. I am doing this to load deserialization functions for a Typeable function. Basically I have code like: $(inferDecoderAndRegisterItOnStartup ''MyType) which defines a class instance, but additionally I want to call 'registerDecoderForType "MyType" decodeMyType' automatically on startup. Calling registerDecodeForType for all types in main gets very tedious and error-prone when doing things by hand. Thus an automated solution would be very nice. - Einar Karttunen

1 0

Re: [Haskell] BitSyntax for Haskell
by Einar Karttunen 28 Sep '06

28 Sep '06

On 26.09 10:01, Adam Langley wrote: > >For the decoding part: > >* Provide a monadic interface > > Are you suggesting a monad to pass in the input around, or that it > returns mzero on error? The latter makes more sense to me. Yes. Also make it possible for user supplied functions to fail in better ways than to produce Either or use error. > >* Add a test part to ReadType: > > Test :: ReadType a -> (a -> Bool) -> ReadType Test > > (or a -> m ()) in the monadic case. > > Again, I'm not clear what you are thinking of here? In some protocols I am using there are some fixed bytes which I want to ignore (no Haskell value produced), but check that they are valid in the data stream. - Einar Karttunen

1 0

Expressing seq
by Chad Scherrer 28 Sep '06

28 Sep '06

I was reading on p. 29 of "A History of Haskell" (a great read, by the way) about the controversy of adding seq to the language. But other than for efficiency reasons, is there really any new primitive that needs to be added to support this? As long as the compiler doesn't optimize it away, why not just do something like this (in ghci)? Prelude> let sq x y = if x == x then y else y Prelude> 1 `sq` 2 2 Prelude> (length [1..]) `sq` 2 Interrupted. There must be a subtlety I'm missing, right? -- Chad Scherrer "Time flies like an arrow; fruit flies like a banana" -- Groucho Marx

5 5