- Haskell-Cafe - Haskell.org

Compiling an extremely large Haskell file (in GHC)
by Arjun Guha 29 Jun '05

29 Jun '05

I have an extremely large source file of about 11 MB. It's the all-pairs shortest paths data for a map of the Hyde Park area of Chicago (no real reason, really). I generated information in Scheme and printed the result to a Haskell source file as a list. I then edited the file to initialized an array with the data. GHC, with a 200MB stack, took up 1 hour and 1.3 GB of memory before getting killed by the system. How would I compile something of this size? I need to have the array of all-pairs shortest paths pre-computed. Any suggestions? Thanks. -Arjun

6 7

Announce: Yet Another Tool to Generate FFI Bindings: hsffig
by Dimitry Golubovsky 29 Jun '05

29 Jun '05

I am pleased to announce the very first alpha release of the (yet another) FFI binding autogeneration tool. Some time ago, I asked the list subscribers about available bindings to Berkeley DB. Several people kindly shared their code with me. One thing I noticed in all bindings: they required manually-written wrappers to accomodate for Berkeley DB indirect style of calling functions. Considering large number of such functions in BDB, I attempted to write a tool which would automatically parse the BDB include file, and extract function prototypes to generate FFI declarations. Soon it became clear that an almost full C syntax parser would be necessary. Finally it resulted in what I am presenting to the Haskell community with hope that this utility will be useful to many developers. Of course there are many omissions and shortcuts, but before I move forward, I would like to get some feedback from potential users of this tool. There is no detailed documentation. I have written a brief README file. More detailed documentation will follow. I haven't created a homepage for this project yet: the only way to get it so far is the darcs repo: darcs get http://www.golubovsky.org/repos/hsffig Files in the repository may be browsed at: http://www.golubovsky.org/repos/hsffig/ I will appreciate any feedback. Dimitry Golubovsky Middletown, CT

2 1

Re: ANNOUNCE: GHC survey results
by Bulat Ziganshin 29 Jun '05

29 Jun '05

Hello Simon, Tuesday, June 28, 2005, 1:58:13 PM, you wrote: SM> http://www.haskell.org/ghc/survey2005-summary.html SM> There's a lot to take in, but it's an interesting read. Enjoy! thank you all, processing all the 600 answers was not easy work :) i have several comments regarding results of this survey: 1) GHCi compiles to bytecode several times faster than GHC makes unoptimized compilation. can unoptimized GHC compilation just create bytecode (as Ocaml does)? 2) is your plans to support x86-64 platform includes Windows (and all other sorts of Unix), or it's for Linux only? 3) many users complaining about non-compatibility between GHC versions. if they mean library interfaces changes then how about using Pesco's library versioning scheme? (see http://www.haskell.org/tmrwiki/EternalCompatibilityInTheory) 4) i definitely will poll for nightly snapshot build for Windows (if it's possible) and frequent releases of stable versions i think there is a plans to move toward the "real world". but it encounter the chicken&egg problem - the commercial programmers don't want to use Haskell because it's too far-from-earth, and you cannot make movements toward "real world" because there is almost no voices from non-academic world. of course i can't monopolize the right to say over all these programmers. but from my point of view several things can be done to be closer to commercial programs and programmers: 1) adding OO-like features in O'Haskell style. for readers that don't know it's about adding new variants to ADT types: type Figure = Circle ... type Figure |= Box ... and adding new fields to existing records: type Point = {x,y::Int} type ColoredPoint = Point extended with {c::Color} imho these extensions are great because it's both in Haskell and C++ style 2) adding "real-world" libraries. i think about libraries for string processing, regular expression handling, command-line processing, serializing, ASCIIZ strings, working with filesystem and running external commands, networking, web-oriented libraries (http/ftp/smtp/pop3, constructing/deconstructing emails), DBMS access i enumerated here only working libraries which i seen in MissingH, or written by Pesco, John Meacham, or Peter Simons 3) may be it is time to include some FFI-generating tool in GHC? many newcomers have some C/C++ libraries they need to communicate 4) some newcomers wrote what they want to say example programs and even tutorials in the package. i think that at least links to such sources/texts can be provided so that a newcomer can easily get a TASTE of Haskell 5) of course, you must add words about Pugo and Darcs in readme :) anyway, ADT is a vehicle of progress ;) -- Best regards, Bulat mailto:bulatz@HotPOP.com

4 6

Noob error: Type b -> c b Does not match IO a
by kynn＠panix.com 28 Jun '05

28 Jun '05

--text follows this line-- I'm trying to learn Haskell from YAHT. My attempt at a solution of Exercise 3.10 is failing with a "Type does not match" error. The exercise is to write a function that will read numbers (one per line) from the command line, until the number 0 is entered. At this point the program is supposed to print out the sum of the numbers, their product, and for each number, the factorial. The interaction should resemble this: Give me a number (or 0 to stop): 5 Give me a number (or 0 to stop): 8 Give me a number (or 0 to stop): 2 Give me a number (or 0 to stop): 0 The sum is 15 The product is 80 5 factorial is 120 8 factorial is 40320 2 factorial is 2 The code I have for this is given next (the line that fails is indicated by a comment): module Main where import IO main = do hSetBuffering stdin LineBuffering numberList <- getList let s = sumList numberList let p = prodList numberList putStrLn ("The sum is " ++ s) putStrLn ("The product is " ++ p) printFact numberList fact 0 = 1 fact 1 = 1 fact x = x * fact x - 1 sumList [] = 0 sumList (x:xs) = x + sumList xs prodList [] = 1 prodList (0:xs) = 0 prodList (x:xs) = x * prodList xs printFact [] = return printFact (x:xs) = do -- triggers error message putStrLn (x ++ " factorial is " ++ fact x) printFact xs return getList = do putStrLn "Give me a number (or 0 to stop):" numString <- getLine let num = read numString if num == 0 then return [] else return (num:getList) Here's the full error I get (from HUGS): IO> :l Ex3_10 ERROR "./Ex3_10.hs":28 - Type error in generator *** Term : printFact xs *** Type : b -> c b *** Does not match : IO a If anyone can explain to me how to fix this error I'd appreciate it. Also, what is the difference between <- and let? Lastly, any comments on any other detail of the code, particularly coding, style are most welcome. Thanks! kj

5 5

Why I Love Haskell In One Simple Example
by John Goerzen 28 Jun '05

28 Jun '05

I posted this on my blog at http://changelog.complete.org/node/339 but I thought there may be some here that would find it of use. I recently implemented some new Haskell numeric types that, instead of performing calculations, can generate a rendering of the requested calculation or store units with it. Here you see a transcript of my session with a Haskell interpreter. The mathematical statements I am entering after the ">" are standard Haskell expressions, and, as I demonstrate, normally evaluate to a single result. Once I get a more powerful simplifier, I will probably write a LaTeX exporting function as well. The entire implementation of this, BTW, is less than 200 lines. NumTest> 5 + 1 * 3 8 NumTest> prettyShow $ 5 + 1 * 3 "5+(1*3)" NumTest> rpnShow $ 5 + 1 * 3 "5 1 3 * +" NumTest> prettyShow $ 5 + 1 * 3 "5+(1*3)" NumTest> prettyShow $ simplify $ 5 + 1 * 3 "5+3" NumTest> prettyShow $ 5 * (Symbol "x") + 3 "(5*x)+3" NumTest> 5 / 2 2.5 NumTest> (units 5 "m") / (units 2 "s") 2.5_m/s NumTest> (units 5 "m") / 2 2.5_m NumTest> 10 * (units 5 "m") / (units 2 "s") 25.0_m/s NumTest> sin (pi/2) 1.0 NumTest> sin (units (pi/2) "rad") 1.0_1.0 NumTest> sin (units 90 "deg") 1.0_1.0 NumTest> (units 50 "m") * sin (units 90 "deg") 50.0_m NumTest> ((units 50 "m") * sin (units 90 "deg")) :: Units (SymbolicManip Double) 50.0*sin(((2.0*pi)*90.0)/360.0)_m NumTest> rpnShow $ dropUnits $ ((units 50 "m") * sin (units 90 "deg")) "50.0 2.0 pi * 90.0 * 360.0 / sin *" NumTest> (units (Symbol "x") "m") * sin (units 90 "deg") x*sin(((2.0*pi)*90.0)/360.0)_m Also, I defined this in my source file: test :: forall a. (Num a) => a test = 2 * 5 + 3 Now, it can be used: NumTest> test 13 NumTest> rpnShow test "2 5 * 3 +" NumTest> prettyShow test "(2*5)+3" NumTest> test + 5 18 NumTest> prettyShow (test + 5) "((2*5)+3)+5" NumTest> rpnShow $ test + 5 "2 5 * 3 + 5 +" You can grab the very early experimental code with darcs get http://darcs.complete.org/num. Haskell has no built-in support for numeric types with units, arbitrary symbols carried through computations, etc. But it was trivial to add it. This kind of extensibility is a key part of why Haskell is so amazing.

5 6

RE: [Haskell-cafe] Compiling an extremely large Haskell file (in GHC)
by Simon Marlow 28 Jun '05

28 Jun '05

On 27 June 2005 19:24, Arjun Guha wrote: > I have an extremely large source file of about 11 MB. It's the > all-pairs shortest paths data for a map of the Hyde Park area of > Chicago (no real reason, really). I generated information in Scheme > and printed the result to a Haskell source file as a list. I then > edited the file to initialized an array with the data. > > GHC, with a 200MB stack, took up 1 hour and 1.3 GB of memory before > getting killed by the system. How would I compile something of this > size? I need to have the array of all-pairs shortest paths > pre-computed. Any suggestions? GHC doesn't have a good way to compile large amounts of static data. Either: (a) put the data into a static array in a C file, compile it with a C compiler, and access it via the FFI, or (b) encode it as a string, and access it using GHC's primitives, or (c) read it from a file at runtime. (Happy and Alex use (b) for encoding parse tables, BTW). Cheers, Simon

1 0

Why distinct tyvars in instance declarations?
by Josh Hoyt 28 Jun '05

28 Jun '05

Hello, I'm a new Haskeller, and I'm running into a problem attempting to declare certain types as instances. I was attempting something that's effectively equivalent to: > class Foo a > > instance Foo (Either b b) but GHC complains: > Illegal instance declaration for `Foo (Either b b)' > (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 (Either b b)' and so I looked in the report (http://www.haskell.org/onlinereport/decls.html#sect4.3.2) and found: > The general form of the corresponding instance declaration is: > > instance cx' => C (T u1 ... uk) where { d } > > where k>=0. The type (T u1 ... uk) must take the form of a type constructor T applied to simple type variables u1, ... uk; furthermore, T must not be a type synonym, and *the ui must all be distinct*. (emphasis mine) My question is, why this restriction that the types must be distinct? In particular, I'd like to declare a very specific type (Either String String) as an instance. What techniques can I use to accomplish this? Josh Hoyt

3 3

Re: [Haskell-cafe] Why distinct tyvars in instance declarations?
by Frank-Andre Riess 28 Jun '05

28 Jun '05

Hi there, > but GHC complains: > > Illegal instance declaration for `Foo (Either b b)' > > (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 (Either b b)' unless I'm totally mistaken, your problem isn't the distinction thingy, but rather an error like supplying an Int for where you need (Int -> Int -> Int). That is, you're trying make (Either String String) an instance of Foo, (Either String String) already being a fully constructed type; Foo, on the other hand, seems to require a type constructor that is yet to parameterize over three more types (e.g. StateT). Greets, Frank-Andre Riess

3 2

Re: [Haskell-cafe] Noob error: Type b -> c b Does not match IO a
by kynn＠panix.com 28 Jun '05

28 Jun '05

Some simplifications might help you here... > prodList [] = 1 > prodList (0:xs) = 0 > prodList (x:xs) = x * prodList xs Simplified: prodList xs = foldl (*) 1 xs But my original at least made some provision for short circuiting the whole operation if the list contained a 0. As far as I can figure, fold, map, etc., are not suitable for any situation in which short-circuiting would be desirable (e.g. and, or, etc.). Am I wrong? (I realize that, given that the function product is already available, there is no point in defining my function prodList above, but my question is a more general one, having to do with how best to implement short-circuiting.) Thanks! kj

6 8

type inference and named fields
by Malcolm Wallace 28 Jun '05

28 Jun '05

I have discovered something I believe to be a problem in Haskell'98, although it is not a simple bug as such - it has more of the flavour of an unintended mismatch in the interaction of two separate features. Since Haskell is deeply principled language, a feature conflict is extremely rare, and so worthy of note. So here it is: Some attempts to use named field updates are prohibited by type inference, even though the code is perfectly type safe. Here is an example. > module Fieldbug where > data Fields a = > VariantWithTwo { field1 :: a > , field2 :: a } > | VariantWithOne { field1 :: a } The key point here is that the data structure with named fields has more than one constructor, and some fields are omitted in one of the variants. Now let's try to write a simple conversion function over values of this type, using only the named-field update style: > data Void = Void > > voidcast :: Fields a -> Fields Void > voidcast v@(VariantWithTwo{}) = v { field1 = Void , field2 = Void } > voidcast v@(VariantWithOne{}) = v { field1 = Void } It looks simple enough doesn't it? But none of the widely available Haskell implementations will accept this code. Hugs: ERROR "Fieldbug.hs":11 - Inferred type is not general enough *** Expression : voidcast *** Expected type : Fields a -> Fields Void *** Inferred type : Fields Void -> Fields Void ghc: Fieldbug.hs:11:32: Couldn't match the rigid variable `a' against `Void' `a' is bound by the type signature for `voidcast' Expected type: Fields Void Inferred type: Fields a In the record update: v {field1 = Void} In the definition of `voidcast': voidcast (v@(VariantWithOne {})) = v {field1 = Void} nhc98: ====== Errors after type inference/checking: Derived type for Fieldbug.voidcast at 10:1-11:42 does not match due to: given free variable a is bound to Void Derived:((Fields Void) -> (Fields Void)) Given :((Fields a) -> (Fields Void)) As far as I can tell, the problem is that in the final line, the expression v { field1 = Void } is interpreted by the type inference algorithm as if the variable v could contain either constructor VariantWithOne or VariantWithTwo. Obviously in the latter case the expression would indeed be incomplete because it casts only one of the fields, not both. But we know that can never be the case! By pattern-matching, this particular v /must/ be the former constructor. Unfortunately we cannot pass that knowledge into the type inference algorithm. It turns out that it is in fact /impossible/ to write this conversion function using only named field updates. It can only be written by using an explicit constructor on the rhs, which thus forces you to initialise all its fields explicitly. This negates any value from using named fields in the first place. Regards, Malcolm

10 32