December 2001 - Haskell-Cafe

Re: Pointers in Haskell??
by Jeff Dalton 07 Dec '01

07 Dec '01

> All Haskell compilers use pointers internally. The > idea is that because Haskell is referentially > transparent and side effect free, you can overwrite > a function application with its result. For example, > > let > x = [1..1000] > in > foo (A x) (B x) > > Will internally have "x" pointing to the function > application [1..1000], when this function application > is evaluated it will overwrite the memory cell "x" > points to with the result. So eventually it will > look like this: > > +---+---+ +---+---+ > | A | x | | B | x | > +---+---+ +---+---+ > | | > +---+-------+ > | > V > +---+---+------+ > | : | 1 | tail |---> The list 2..1000 > +---+---+------+ > > Where each block is a memory cell and each arrow is a > pointer. What does it mean to have a letter ("A", "x", or a number or "tail", etc) inside a box? > A book that describes this a lot better is: > > Simon Peyton-Jones. The implementation of functional > programming languages. Prentice-Hall, 1987 Are they always implemented that way these days? -- J

1 0

Pointers in Haskell??
by Bryan Hayes (Hayes Technologies) 07 Dec '01

07 Dec '01

I am totally new to Haskell, so maybe this is a stupid question. Various languages have pointers (or references), for good reason. Haskell can at least partly do without them (they are only existing internally somehow). My question is: Does Haskell principally not need pointers (i.e. in case of 2 data structures needing to reference an other very large data structure) or is this a design flaw or have a overlooked something? -------------------------------------------------- HAYES TECHNOLOGIES Software Speed Optimization Bryan Hayes Managing Director Mannheimer Str. 8 D-67098 Bad Dürkheim Deutschland / Germany Tel. +49 / (0)6322 / 94 950 - 68 Fax +49 / (0)6322 / 94 950 - 69 Mobile Tel. +49 / (0)163 / 6111867 E-Mail: mailto:bryan.hayes@hayestechnologies.com Web-Site: http://www.hayestechnologies.com

4 3

How to write a lazy variant ?
by Dmitry Astapov 05 Dec '01

05 Dec '01

Consider the following functions: readRecFile fname = do ifh <- openFile fname ReadMode readRecords ifh readRecords ifh = do x <- decodeNextRecord ifh more <- isEof ifh if (more == False) then return [x] else do rest <- readRecords ifh return (x:rest) They allow me to read contest of file fname into a list of data structures. But "do rest <- readRecords ifh ; return (x:rest)" part makes them behave non-lazily. Maybe I'm just too tired right now, but it escapes me how can I write a lazy variant of those functions, so that they will behave like getContents. Does anyone have a clue? -- Dmitry Astapov //ADEpt E-mail: adept(a)umc.com.ua GPG KeyID/fprint: F5D7639D/CA36 E6C4 815D 434D 0498 2B08 7867 4860 F5D7 639D

2 3

Re: Functional Metapost revival
by Ralf Hinze 04 Dec '01

04 Dec '01

Dear Feri, > You can find a version of Functional Metapost, which works > with current (2001 February) Hugs, on the following page: > > http://afavant.elte.hu/~wferi/funcmp/ > > This version is able to produce all the illustrations in the > enclosed Tutorial on my machine. > > Good luck, waiting for your comments: Feri. sorry for not being responsive. Currently, I don't have time to look into this. However, I put my current version of FMP on the web, which works both with Hugs (2001 February) and GHC. http://www.informatik.uni-bonn.de/~ralf/FuncMP-0.2.tar.gz Cheers, Ralf

2 1

Integer to String Conversion?
by Chris 04 Dec '01

04 Dec '01

Hi everyone, is there a function that converts Integers to Strings and vice versa? Thanks in advance Chris

3 2

Graphic User Interface Question
by Gabriel Alix 02 Dec '01

02 Dec '01

Hello All; My name is Gabriel Alix; I am currently an undergrad Computers Science major at Colgate University working with Proffessor Phillip Mulary(http://cs.colgate.edu/new/faculty/mulry.html). I am currently working with a group of other people to build a UI but am finding a lack of good examples to study off of. Would anyone have any references, links, or code that shows the inner workings of UI in Haskell? We are look for anything that will help us designs a good GUI. Any help would be appreciated Thank You ~Gabriel Alix

2 1

Re: Strictness making it worst?
by Marcin 'Qrczak' Kowalczyk 01 Dec '01

01 Dec '01

Sat, 1 Dec 2001 15:58:47 +0000, Jorge Adriano <jadrian(a)mat.uc.pt> pisze: > So my question is, is there anyway to force an argument to be reduced to > *normal form*? The meaning of "normal form" (applied to an arbitrary Haskell type) is not precisely defined. It depends on the type and must be implemented for each type separately, you can at most group it in a class. It can't be implemented for functions at all if you want to evaluate their free variables. -- __("< Marcin Kowalczyk * qrczak(a)knm.org.pl http://qrczak.ids.net.pl/ \__/ ^^ QRCZAK

1 0

Re: Strictness making it worst?
by Jorge Adriano 01 Dec '01

01 Dec '01

> How high can you go with fibac before getting a stack overflow? I'm > guessing that it's no more than 200000, and that the suspensions have > the following structure. 100 000 works fine 110 000 stack overflow > > fibac (as you described): > > c = (+) a b > d = (+) b c > e = (+) c d > > sfibac: > > c = ($!) ((+) a) b > d = ($!) ((+) b) c > e = ($!) ((+) c) d > > i.e. the ($!) is in the wrong place, and is only making the suspension > twice as deep. Addition is already strict in both its parameters. correct, dumb mistake, thanks :) > > sfibac' n (a,b) | n == 0 = (a,b) > > | otherwise = sfibac' (n-1) ((,) b $! a+b) Yeap, got it. Well, in fact my first aproach was not using pairs: > sfibac2 :: IntPos -> IntPos -> IntPos -> IntPos > sfibac2 n a b | n == 0 = a > | otherwise = sfibac2 (n-1) b $! (a+b) With pairs I thought I'd write > dumbfibac :: IntPos -> IntPos -> IntPos -> IntPos > dumbfibac n a b | n == 0 = a > | otherwise = dumbfibac (n-1) $! (b,(a+b)) But this would obviously not work because ($!) reduces its 2nd argument to 'head normal form', not 'normal form'. Anyway IMO it looks way much better then yours sfibac' function :-) So my question is, is there anyway to force an argument to be reduced to *normal form*? a strict fold function defined like, > sfoldl f a [] = a > sfoldl f a (x:xs) = (sfoldl f $! (f a x)) xs would have the same problem as my dumbfibac function. How do you correctly define a strict fold function? J.A.

1 0