- Haskell-Cafe - Haskell.org

Fail: <<loop>> ???
by Ingo Wechsung 21 Dec '02

21 Dec '02

Dear Haskellers, when I execute a program made with GHC (standard Haskell only, no optimization) I get Fail: << loop>> I have -W on, but it only tells me about a single overlapping pattern, which is ok. The same program crashes Hugs alltogether. Any hints for what I should look in my sources? MfG Ingo

3 2

Tree insert, lookup with keys
by Ingo Wechsung 20 Dec '02

20 Dec '02

Hi, perhaps I don't understand the type system fully yet: I want a tree that carrys some information of type s; data Tree s = TNil | Branch s Tree Tree; This is fine for String, Integer and other atomic types. ins :: Tree s -> s -> Tree s; ins Tnil s = Branch s Tnil Tnil; ins (Branch s l r) x = case compare x s of { LT -> Branch s (ins l x) r; EQ -> Branch s l r; GT -> Branch s l (ins r x); } So far so good, however, I do not always want to compare everything. What I really want is to have (compare (key x) (key s)) instead in the definition of ins. My attempt: class Keyed a where { -- Type a is keyed if it has a key function. key :: Ord b => a -> b; -- key is a function, that, when applied to a yields some b that is comparable } and then data Sym = Sym String Int Ty; -- Ty is another algebraic type instance Keyed Sym where { key (Sym a _ _) = a; -- use string as key } But I always get the following message: Cannot unify the type-signature variable `b` with the type `String` Expected type: b Inferred Type: String In the definition of `key': a Why is this? Has "key" to return the very same type for every instance? After all, I need type `b' in compare only so it should not matter what it really is. The compiler could check if for every instance key is defined so that the return type is some instance of Ord, couldn't he. Merry Christmas Ingo

3 2

AW: Tree insert, lookup with keys
by Ingo Wechsung 20 Dec '02

20 Dec '02

Ketil wrote: >But it isn't obvious what b is supposed to be. Sure, it must be an instance of Ord. In a way, I promise the compiler not to use anything but compare on b's. Of course, when I define n instances, there will be n different b-Types that must not be confused. >(Requires -fglasgow-exts) Thanks for the tip. Since I am obviously still learning, I prefer not to use the extensions until I know exactly what they are about. (As it seems, I am just experiencing one case where the extensions are useful. I'm sure there are many other cases.) Meanwhile, I see 2 possibilities to solve the problem in standard Haskell: (a) Stick with String keys key :: a -> String; or, (b) forget class Keyed and for every type s in Tree s derive an instance from Ord that defines compare accordingly. Solution (b) seems not so good to me since we might need the standard Ord behaviour elsewhere. Solution (a) is practical and should perform fairly well, except when we must write something like key (Baz x s y) = show x; Are there any other solutions in standard Haskell? Ingo

1 0

Parsing date and time specifications
by Peter Simons 20 Dec '02

20 Dec '02

Hi, I have written a parser that turns an RFC2822 date and time specification into a datatype usable in Haskell. While the parser is working just fine so far, I have a problem with the CalendarTime datatype. It appears that in order to construct one of those, I need _all_ the information it contains, including the weekday (Day) and the number of the day in the year. The problem now is that I do not have this information! Of course I could calculate these values by hand, but this is immensely complicated. I thought about using a TimeDiff instead, but apparently the functions provided in the Prelude do not handle TimeDiff for anything except for "maths" -- what doesn't really help me for what I am trying to do. I briefly looked at the Posix module that comes with GHC as well, and which seems to provide CTime. But then, the Posix module appears to be non-standard. *sigh* Any suggestions what I could do? -peter

3 2

Re: Library for concurrent threads running
by Kevin Hammond 18 Dec '02

18 Dec '02

I am also not quite sure of your motivation! The current version of GpH is (or can be with very little difficulty) implemented using library calls. There is really no language extension per se, rather there are calls to tightly-coupled C code, plus a set of high level libraries to provide structure (evaluation strategies). There are, of course, modifications to the RTS to support proper scheduling, communications etc. , but these are necessary for a properly integrated high-level approach. If you want a low-level approach, one of the MPI libraries for Haskell may suit you better? Kevin At 7:16 pm +0100 16/12/02, Christian Sievers wrote: Dusan Kolar asked: > My question/wish is maybe naive, but anyway: > Is there a library (not a language extension, like > Concurrent Haskell, Glasgow Parallel Haskell, ...) enabling > to run two functions in parallel, possibly in cascade? I think the best you can get without language extensions is what is described in Koen Claessen's Functional Pearl "A Poor Man's Concurrency Monad". It's really nice, but not exactly what you seem to want. You can get that article from http://www.cs.chalmers.se/~koen/publications.html All the best Christian Sievers _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

1 0

Random
by Filipe Santos 18 Dec '02

18 Dec '02

I need some help to do a function so that I cant get 4 numbers between 1 and 6, i tried to use random but i can't make it work well.

3 6

(no subject)
by Olga Lucia Quintero Fonseca 17 Dec '02

17 Dec '02

Por favor, estoy aprendiendo el lenguaje y no se como usar Complex aqui, el prompt siempre esta como Prelude y no se como incluir ese package module. Podrian ayudarme ? De antemano gracias !!!

1 0

FW: Random
by Simon Peyton-Jones 17 Dec '02

17 Dec '02

I'm behind the times... Thanks Volker Simon -----Original Message----- From: Simon Marlow Sent: 17 December 2002 15:51 To: Simon Peyton-Jones Subject: RE: Random Volker Stolz already did this (and for a bunch of other H98 libraries), but I didn't merge it into the branch, sorry. Cheers, Simon > -----Original Message----- > From: Simon Peyton-Jones [mailto:simonpj@microsoft.com] > Sent: 17 December 2002 15:08 > To: Mark Carroll; haskell-cafe(a)haskell.org > Subject: RE: Random > > > | Ah - I was never sure what to make of that - I normally just use the > GHC > | online Haddockised stuff which tells me no more than the type > signatures, > | but I suppose "split" must be more than (\x->(x,x))! (-: (I'll be > happy to > | help with adding documentation once I'm sure of the > semantics myself.) > > The Random library is carefully documented in the Haskell Report. It > would be an excellent and wonderful thing if someone could Haddock-ise > the Haskell Report documentation and put it in the library > documentation. > > Volunteers get CVS write access! > > Simon > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://www.haskell.org/mailman/listinfo/haskell-cafe >

1 0

RE: Random
by Simon Peyton-Jones 17 Dec '02

17 Dec '02

| Ah - I was never sure what to make of that - I normally just use the GHC | online Haddockised stuff which tells me no more than the type signatures, | but I suppose "split" must be more than (\x->(x,x))! (-: (I'll be happy to | help with adding documentation once I'm sure of the semantics myself.) The Random library is carefully documented in the Haskell Report. It would be an excellent and wonderful thing if someone could Haddock-ise the Haskell Report documentation and put it in the library documentation. Volunteers get CVS write access! Simon

1 0

Library for concurrent threads running
by Dusan Kolar 16 Dec '02

16 Dec '02

Hello, My question/wish is maybe naive, but anyway: Is there a library (not a language extension, like Concurrent Haskell, Glasgow Parallel Haskell, ...) enabling to run two functions in parallel, possibly in cascade? Something like: > testAll (threadStart f1 (threadStart f2 f3)) > > where threadStart :: a -> b -> (WillBe a, WillBe b) > where type WillBe is something like Maybe, but > changing within a time (mutable values) ~ Processing, Just a, Nothing > > testAll (Nothing,p) = putStr "No result for f1" >> testPair p > testAll (l,(Nothing,r)) = putStr "Noresult for f2" >> testPair (l,r) > testAll (l,(ll,Nothing)) = putStr "No result for f3" >> testPair (l,ll) > testAll (Just a,p) = putStr ("res of f1: " ++ show a) >> testPair p > ... > > testPair ~ about the same like testAll Probably the explanation is not too clear and what I wish is out of the language scope (not really functional), but I would appreciate something like that. Is there something like that for Haskell/Hugs? Thx for any ref. Dusan

3 2

Re: Haskell run-time tutorial
by Jonathan Holt 16 Dec '02

16 Dec '02

Matt Hellige wrote: > I found the book: > > Simon Peyton-Jones and David Lester > "Implementing Functional Languages: A Tutorial" > Prentice Hall, 1992 Thanks for the detailed and helpful reply. I've already downloaded the book, and from what I've seen thus far, it seems to be exactly what I need! Thanks! -Jony __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com

3 2

Haskell run-time tutorial
by Jonathan Holt 14 Dec '02

14 Dec '02

Hi guys, Having completed a simple parser project using Haskell, I'm getting more and more "hooked" on functional-programming in general and Haskell in particular. And learning the language has been a real treat as well, using the tutorial with occasional peeks at the report. However, when I'm considering undertaking more complex projects, I realize that there is one important erea of the Haskell language that is still pretty much a "black-hole" for me: the run-time environment. I honestly can't say that I completely understand how the actual control and data flows work when a Haskel program is executed, and I seem to be constantly surprised by the garbage-collection, which sometimes seems as if it doesn't work at all, and at other times seems to do magic. I've tried reading some of the articles related to the design of GHC's run-time. But they don't seem to be much help. They either assume that the reader has prior knowledge of functional-programming run-time environments, or get into very low-level details. And at the end of the day, this is very confusing, and I find terms like thunks, entries, partial-function-applications, etc. floating in my head hoplessly seeking anchor. So, where can one find a good introductory book / article on the subject? Thanks, -Jony __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com

2 1

Error Handling
by Thomas L. Bevan 13 Dec '02

13 Dec '02

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, I'm writing a little command line app, but I have a question about error handling. The code will go something like, main = do (a:b:cs) <- getArgs i <- return (read a :: Int) j <- return (read b :: Int) putStr $ i + j How can I catch any possible cast exception? Ideally, I think I would like to use the ErrorT monad as a means of handling a potentially fairly complex set of inputs, but I can see no way of handling these errors neatly within that monad. Tom -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (GNU/Linux) iD8DBQE98+VqGse0w80KF2gRAqnMAJ9CMiOWFS+3y2ELu7851MyvTb1aCgCgru+z EtW/mO+WZ0lqtYUECcHyTDE= =r1DY -----END PGP SIGNATURE-----

6 9

Monads [beginner]
by maas＠cin.ufpe.br 13 Dec '02

13 Dec '02

What are Monads? Where can I read about then? I read that there´s some mathematical foundation under this concept, where can read about these theories? +---------------------------------+ | Marcos Aurélio Almeida da Silva | | E-mail: maas(a)cin.ufpe.br | +---------------------------------+

3 2

RE: Haskell related Debian packages
by Simon Marlow 13 Dec '02

13 Dec '02

> Please excuse me if anyone feels that this email isn't > appropriate for this mailing list. Though these are somewhat > Debian-specific questions, I thought that some Haskell folks > could lent their insight. You would be the users after all. Also see the recent discussion on glasgow-haskell-users about a library building/packaging system (which didn't really reach a conclusion). > There aren't many examples of libraries (without any > binaries) that are packaged thus far. I noticed that > ctklight installs stuff in /usr/share/hugs98/extentions, but > it can be used with GHC also, so I am not sure why its done > this way. I don't use hugs too much though, also I don't know > anything about nhc. > > So I'm considering sticking the *.lhs files for HUnit, for > example, in /usr/share/hunit. Then people can alter their > Makefiles to include that directory in either hugs or ghc, > but I see one problem with that: GHC would want to compile > the source files, and stick them in /usr/share/hunit, but > would be unable to (because of permissions problems when > users compile). > > The package could install binary files, but then it would > build-depend on ghc which doesn't exist for several Debian > platforms. So I guess it could check to see if ghc is > installed on the building computer and stick binaries in /usr/lib. I think I would have two versions of each package: one that builds & installs a package for GHC, and one that installs source files for use with Hugs (GHCi can use the GHC package). These really are entirely different beasts, and in most cases will have zero files in common, so I think it makes sense to separate them altogether. That way you don't have to require GHC in order to install the Hugs version of the package, and vice-versa. Cheers, Simon

1 0

Re: ffi
by Martin Huschenbett 13 Dec '02

13 Dec '02

> > $ ghc -ffi -o myprog Main.hs cfile.o > When I try this I also get an error: martin:~/work/prograemmelchen> ghc -ffi -o myprog Main.hs cfile.o ghc-5.02.2: unrecognised flag: -ffi Usage: For basic information, try the `--help' option. Do I use a wrong GHC version? THX, Martin.

3 4

Re: Monads [beginner]
by Benderjgdefault＠aol.com 12 Dec '02

12 Dec '02

In a message dated 12/12/2002 2:45:31 PM Eastern Standard Time, maas(a)cin.ufpe.br writes: > What are Monads? > Where can I read about then? > > I read that there´s some mathematical foundation under this > concept, where > can read about these theories? A large set of articles on monads in the context of Haskell are collected on my bibliography web site: http://haskell.readscheme.org/monads.html There are also a few, more tutorial, articles available at http://haskell.org/bookshelf/ that are not included on my monads page. For more theoretical works-- while still being programming language related--I suggest Eugenio Moggi's site: http://www.disi.unige.it/person/MoggiE/publications.html. Jim Bender

1 0

Haskell related Debian packages
by None 12 Dec '02

12 Dec '02

I'm considering packaging several Haskell libraries for Debian, and wonder what people think about where things should go. Please excuse me if anyone feels that this email isn't appropriate for this mailing list. Though these are somewhat Debian-specific questions, I thought that some Haskell folks could lent their insight. You would be the users after all. There aren't many examples of libraries (without any binaries) that are packaged thus far. I noticed that ctklight installs stuff in /usr/share/hugs98/extentions, but it can be used with GHC also, so I am not sure why its done this way. I don't use hugs too much though, also I don't know anything about nhc. So I'm considering sticking the *.lhs files for HUnit, for example, in /usr/share/hunit. Then people can alter their Makefiles to include that directory in either hugs or ghc, but I see one problem with that: GHC would want to compile the source files, and stick them in /usr/share/hunit, but would be unable to (because of permissions problems when users compile). The package could install binary files, but then it would build-depend on ghc which doesn't exist for several Debian platforms. So I guess it could check to see if ghc is installed on the building computer and stick binaries in /usr/lib. Another option would be to compile packages for GHC in the post-install script if ghc is installed, and stick those installed files in /usr/lib, and the .hs files in /usr/share/hugs98/extentions if hugs is installed. So my question is: if you were / are using Debian, what do you think that a well-behaved Haskell library package should do? Also Debian users might email me privately with suggestions on what are the most important programs to package: I plan to start with the ones I use the most. peace, Isaac

3 2

AW: ffi
by Markus.Schnell＠infineon.com 12 Dec '02

12 Dec '02

> $ cc -c cfile.c > $ ghc -o myprog Main.hs cfile.o won't you need -fffi also? Markus

1 0

ffi
by Martin Huschenbett 11 Dec '02

11 Dec '02

Hi all, I've got a simple question concerning FFI but as I am a newbie I don't know the answet. I've got a C-source-file, maybe something like: //cfile.c void cfun() { /* ... do something ... */ } //EOF and a Haskell-source-File, maybe something like: -- Main.hs module Main where main :: IO () main = do -- do something cfun -- I wan't to call the function written in C here -- do again something -- EOF And now my question is: How do i realize this? Can you please tell me how write a binding and how tu run ghc. Thank you very much, Martin.

2 1

video4linux
by Martin Huschenbett 11 Dec '02

11 Dec '02

Hi all, does anybody know how to read a single image from a video4linux device under Haskell? Thanks for all help, Martin

2 1

Re: AW: Editor Tab Expansion
by Marcin 'Qrczak' Kowalczyk 10 Dec '02

10 Dec '02

Fri, 6 Dec 2002 12:09:46 +0100, Ingo Wechsung <iw(a)contexo.de> pisze: > No. My editor produces the ASCII code for horizontal tab, when I > hit the tab key. Just as it produces the ASCII code for a when I > hit the a key. > > That's how it should be. It shouln't, becase tabs are 8 spaces, which is too much for indentation of programs in most languages. > The real problem is, that nobody could tell from just looking at > the screen, what a program means even if you know the layout rules > very well. Of course I can. If you can't, perhaps you are using a non-standard tab width? Please fix settings of your editor and it will work. > Instead, your text editor displays some space in place of the tab. > Only if the expansion rules of your editor are compatible with > Haskell's may you grasp the source codes meaning. But they are compatible because there is one most universally accepted interpretation of a tab (move to the next multiple of 8 columns). Any other interpretation hampers portability and should be avoided. > Think of a software development team. One member (in India) > uses "convert n leading spaces to tabs", the other (in Japan) > "convert leading tabs to m spaces". Everytime one of them makes a > small change in a file that has been changed by the other before, > the source code control system will have a huge diff. All these problems are caused by people who use a different tab size than 8. With consistent tab size there is no reason to convert between tabs and spaces. -- __("< Marcin Kowalczyk \__/ qrczak(a)knm.org.pl ^^ http://qrnik.knm.org.pl/~qrczak/

5 5

AW: AW: Editor Tab Expansion
by Ingo Wechsung 10 Dec '02

10 Dec '02

Beg your pardon, Marcin >But they are compatible because there is one most universally accepted >interpretation of a tab (move to the next multiple of 8 columns). Any >other interpretation hampers portability and should be avoided. No. It didn't hamper portability with C, Java, Perl or any other *nix stuff since more than 30 years except with COBOL, Python (?) and Haskell, where COBOL is excused due to its origin in the punch card era. The others are not excused, since they are yet to be established newcomers. The wise new language designer should obey the unix rule "whitespace does not matter" instead to expect that the world change their .exrc (by the way, according to ls -l my .exrc is several years older than Haskell) to accomodate. >All these problems are caused by people who use a different tab size than 8. They used to do so, since Jan 1 1970 00:00 GMT, without real problems apart from pure formatting problems. The formatting problems will not go away with the advent of Haskell. But 2 new problems suddenly arise: syntax error due to formatting and formatting dependend semantic. But even if it were so and you could expand a tab to 8 columns only, there is still the tools problem. No meaningful diff -b on Haskell source. Almost every language tool has to be a haskell parser, etc. etc If you can't or won't see that, fine. But don't expect Haskell to be too successfull in the professional software engeneering business very soon. Though, it was my impression, that the FP community would like to promote their ideas of how to make more reliable, better software to more people. This "obey the 8 column rule, use emacs, change your .exrc or get lost" attitude is less than helpful, at least in my opinion. Cause it's FP that matters, not the layout thing which has absolutely nothing to do with FP. (Anyway, I still love Haskell{;}) Cheers, Ingo

5 5

Editor Tab Expansion
by Ingo Wechsung 10 Dec '02

10 Dec '02

Hello, I'm new to Haskell and FP in general and I find it great. Having more than 10 years expirience with "whitespace does not matter" languages, the only thing that drives me crazy is the layout rule. As far as I understand it, I have 2 options: 1. Use braces and semicolons and ignore the layout rules. 2. Change the settings in all my editors so that the code looks like the Haskell compiler sees it. Currently, I expand tabs to 4 spaces only, so \tx=bar looks like ____foo = bar to me when the compiler sees ________foo = bar I would not want to change dozens of .exrc files, shell startup files with and/or ultraedit settings on many different machines. I also do not want to care whether there are spaces or tabs in front of my source code lines. So I'm stuck with option 1, right? Just to be sure, can I really, really forget about layout if I write fully braced and semicolonoized code? Besides, is there any reason why the syntax is LET { decl1; decl2; ... } IN expr when LET and IN are sufficient enough to enclose the declarations? Greetings, Ingo

6 5

AW: Editor Tab Expansion
by Ingo Wechsung 08 Dec '02

08 Dec '02

Simon wrote: >There's no reason not to use 8 column tab stops, so please don't do it. Ok, if "it just looks better to me" is no reason, then the following also holds: There's no reason not to use braces and semicolons, so please use them. > but it will be interpreted differently by a Haskell compiler. That's exactly my point. I consider it harmful, to put it mildly. >Tab characters are a primitive compression mechanism, nothing more. Well, layout rule may be regarded as an even more primitive compression mechanism, since it saves fewer bytes than tabs. (Remember that "it looks better" was outlawed before.) I have the impression that the view is somewhat biased toward the layout. Nobody seems to care about the trouble that is luring in the long run. You can't go professional with "use another editor", "change your long established habits", "don't use diff -b". Cheers, Ingo

3 2

AW: Editor Tab Expansion
by Ingo Wechsung 07 Dec '02

07 Dec '02

Thanks for the anserws. So I will write the braces and semicolons. It's better anyway in my opinion. I am not going to change my editing habits just to make hugs or ghc happy. Wether spaces or tabs are better in source files is a matter of taste and a language should not force me to use one or another. I can't help it, but to make the meaning of a program depend on the questions whether there are spaces or tabs in front of the line, and how many, seems silly to me. Sorry. Hal Daume III wrote: >so it needs to know where the boundaries between declarations are, hence >the need for semicolons. therefore, if you have embedded lets, you need >braces to delimit them: > let x = let y = z > q = r > in l > in q >would be ambiguous without layout/braces&semis. I don't see that. Sure, you need semis. But then let x = let y=z; q=r; in l in q or even let x = let y=z; q=r in l in q is just as unambiguous as (hypothetical) ( x = ( y=z; q=r ) l ) q I mean, you can see "let" as opening brace and "in" as closing brace. So no real need to write "let {" and "} in". Greetings, Ingo

6 5

AW: Editor Tab Expansion
by Ingo Wechsung 06 Dec '02

06 Dec '02

Andrew wrote: >My remark was merely in response to the claim that Haskell cares >whether you put tabs or spaces in your files. It does not, so long >as the tabs are of length 8. If your editor produces tabs of a >different size, that's a problem with your editor or the way you >dislike Haskell's layout rules. No. My editor produces the ASCII code for horizontal tab, when I hit the tab key. Just as it produces the ASCII code for a when I hit the a key. That's how it should be. The real problem is, that nobody could tell from just looking at the screen, what a program means even if you know the layout rules very well. The reason is, that the Haskell lexer sees tabs while you cannot see tabs unless you use a hex editor. Instead, your text editor displays some space in place of the tab. Only if the expansion rules of your editor are compatible with Haskell's may you grasp the source codes meaning. But, again, you can't know that by just looking at the screen. You have to use the :l command or the :se command in vi, for example, to make sure a) there are either only tabs or only spaces in front of the lines or b) the editor lets you view tabs expanded to 8. Or, to turn it another way: "What you see is not necessarily what you get". This may be fine for ad hoc scripts that one examines in hugs. But how can one hope to convince people that this is a good thing for reusable code, to be compiled code, serious software production! I mean business. Teams. Maintenance of code that has been written by someone else, etc. Some people suggested that the editor be configured so that it converts all tabs to spaces or vice versa on file open. Beware! Think of a software development team. One member (in India) uses "convert n leading spaces to tabs", the other (in Japan) "convert leading tabs to m spaces". Everytime one of them makes a small change in a file that has been changed by the other before, the source code control system will have a huge diff. Because, you can't use diff -b (ignore leading white space in comparision) on Haskell source. That implies that you cannot see later what this small change was. Instead, every indended line will be displayed as changed. This is a NIGHTMARE for project leaders, quality assistance staff, code maintainers. Greetings, Ingo

2 1

AW: AW: AW: Editor Tab Expansion
by Ingo Wechsung 06 Dec '02

06 Dec '02

Kevin, thanks for your helpfull comments. Ingo> Or, to turn it another way: "What you see is not necessarily Ingo> what you get". This may be fine for ad hoc scripts that one Ingo> examines in hugs. >So is that the language's fault (because of what you get) or the >editor's fault (because of what you see)? It depends on the abstraction level. Sure, I don't see Tabs and Spaces on the character level. I was not supposed to, up to now. On a higher level I see whitespace. Whitespace didn't matter, except for making things look nicer, up to now. I don't say the language is wrong. Beware. I only have my doubts about the question if its a good idea to cast a whitespace aware thing into an environment, namely *nix, where whitespace didn't matter. Up to now. Now that I know I can use semicolons and braces safely with any spacing, I can live with it. It's fine for me. I could write a haskell-insert-braces-and-semicolons-where-needed program if I feel like it. I think I will not feel like so, except for amusement. But then, I've got to get checks from my customers by the end of the month. So, perhaps I'll rather try to get my work done. But the problem will hurt others, again and again. Good night Ingo

1 0

Completeness of pattern matching
by Ingo Wechsung 06 Dec '02

06 Dec '02

Hi there, I wonder if the compiler could check, if all possible combinations have been checked in a pattern match. For example: data Colour = Red | Yellow | Green | Blue; trafficlight Red = True; trafficlight Yellow = True; trafficlight Green = True; trafficlight Blue = False; This is complete and okay. But, if somebody adds Magenta to Colour, then the function trafficlight becomes incomplete and the program might terminate at runtime whereas before it could not. (Of course, there could be dozens of more complicated case scattered over multiple files ...) If this could be checked, shouldn't it be an compile time error to have incomplete functions? Cheers, Ingo

3 2

RE: Editor Tab Expansion
by Simon Marlow 06 Dec '02

06 Dec '02

> No. My editor produces the ASCII code for horizontal tab, > when I hit the tab key. > Just as it produces the ASCII code for a when I hit the a key. > > That's how it should be. Not true: most editors don't insert a ^H when you hit the backspace key, so why should they insert ^I when you hit the tab key? This whole problem boils down to one thing: people configure their editors to display tab stops at something other than every 8 columns. There's no reason not to use 8 column tab stops, so please don't do it. Not only will your code be unreadable by someone using standard tab-stop settings, but it will be interpreted differently by a Haskell compiler. Tab characters are a primitive compression mechanism, nothing more. No-one is saying that you can't use units of 4 columns for indentation, just that you can't use tab characters for that. Cheers, Simon

1 0

AW: Editor Tab Expansion
by Ingo Wechsung 06 Dec '02

06 Dec '02

Markus wrote: >Perhaps somebody could come up with a "Haskell indent" (part of a pretty >printer?)? Hmmm... while for example "cb" is a comparably simple program, this "Haskell indent" had to be a full blown parser. I'd rather like to have a "Haskell brace" that makes fun (x,y) = case x of 7 -> case y of -- one tab 4 -> "seven four" -- two tabs _ -> "no seven in front" -- eight spaces into fun (x,y) = case x of { 7 -> case y of { 4 -> "seven four"; } _ -> "no seven in front"; } Regards, Ingo

1 0

AW: Editor Tab Expansion
by Markus.Schnell＠infineon.com 06 Dec '02

06 Dec '02

When I got somebody else's C-Code, I used gnu indent to bring it into a layout I liked (and could easily grasp). One of the points I like about Haskell is just its use of space: it makes things clearer and doesn't clutter up your screen as it does in C. I definitely had less problems with layout in Haskell than with forgetting braces in C (and line them up nicely). > unambiguous one using spaces) . I think it's really the fault of ASCII for > allowing the tab character code which has user-defined semantics, just > like the other abomination the form-feed character. That's because of its history: ASCII was used first for telegraphs, so TAB was just a shorter way of saying "make some space", same with form-feed, the bell, and all those other characters from 0-31 you never use. If somebody wants to write his/her Haskell code with braces, so be it - unless you develop in a group, then you should have layout rules anyway. Perhaps somebody could come up with a "Haskell indent" (part of a pretty printer?)? Markus

1 0

Re: Editor Tab Expansion
by Ashley Yakeley 06 Dec '02

06 Dec '02

At 2002-12-05 15:34, Andrew J Bromage wrote: >> Wether spaces or tabs are better in source files is a matter of taste and >> a language should not force me to use one or another. > >The language does not force you to do anything of the sort. It's your >editor's fault if it can't decouple the concept of hitting the tab >key from the concept of putting a ^I character in your file. Haven't we all been through this argument several months ago? I believe the conclusion was "people have different preferences, and Haskell allows for that". Certainly when I write my own Haskell I like to use tabs, real ones, to indent. And I can easily outdent by using the backspace key, once, which will cleanly delete exactly one tab. I'll have no truck with Haskell's "layout" and its picky interpretations of whitespace. Luckily for people like me there's braces and semicolons, which let you ignore this layout nonsense entirely. YMMV, of course. -- Ashley Yakeley, Seattle WA

2 1

Re: AW: Editor Tab Expansion
by Ashley Yakeley 06 Dec '02

06 Dec '02

At 2002-12-05 12:49, Ingo Wechsung wrote: >So I will write the braces and semicolons. It's better anyway in my opinion. > >I am not going to change my editing habits just to make hugs or ghc happy. >Wether spaces or tabs are better in source files is a matter of taste and >a language should not force me to use one or another. I can't help it, but >to make the meaning of a program depend on the questions whether there are >spaces or tabs in front of the line, and how many, seems silly to me. Sorry. I also use braces and semicolons for the exact same reason. -- Ashley Yakeley, Seattle WA

1 0

Re: Editor Tab Expansion
by C.Reinke 05 Dec '02

05 Dec '02

> Having more than 10 years expirience with "whitespace does not > matter" languages, the only thing that drives me crazy is the > layout rule. so in 10 years of programming, you've never written a Makefile?-) > As far as I understand it, I have 2 options: > > 1. Use braces and semicolons and ignore the layout rules. > > 2. Change the settings in all my editors so that the code looks > like the Haskell compiler sees it. 3. Change the settings in all your editors to get rid of those crazy hard-tabs! Good editors support this by enabling you to replace hard-tabs by soft-tabs (you type <Tab>, your editor inserts the right number of spaces, for a configurable notion of "right" => you can keep your favourite tab2space number and you and the compiler see the same code). (e.g., in Vim, type ":help tabstop" for a discussion of alternatives, my personal favourite is 2. Set 'tabstop' and 'shiftwidth' to whatever you prefer and use 'expandtab'. This way you will always insert spaces. The formatting will never be messed up when 'tabstop' is changed. you can still enter hard tabs when needed, e.g., in Makefiles, using CTRL-V <Tab>; there's also a useful help to make hard-tabs visible for debugging - ":help 'list'"). I got into the habit of hating hard-tabs before learning Haskell - in multi-developer projects, different layout styles are one thing, but combined with hard-tabs and different tab2space settings.. Hth, Claus

1 0

Re: Haskell-Cafe -- confirmation of subscription -- request 771383
by Nikita Ayzikovsky 05 Dec '02

05 Dec '02

On Thu, 05 Dec 2002 18:59:48 -0500 (EST) haskell-cafe-request(a)haskell.org wrote: > Haskell-Cafe -- confirmation of subscription -- request 771383 > > We have received a request from 24.78.145.92 for subscription of your > email address, <ayzik(a)shaw.ca>, to the haskell-cafe(a)haskell.org > mailing list. To confirm the request, please send a message to > haskell-cafe-request(a)haskell.org, and either: > > - maintain the subject line as is (the reply's additional "Re:" is > ok), > > - or include the following line - and only the following line - in the > message body: > > confirm 771383 > > (Simply sending a 'reply' to this message should work from most email > interfaces, since that usually leaves the subject line in the right > form.) > > If you do not wish to subscribe to this list, please simply disregard > this message. Send questions to haskell-cafe-admin(a)haskell.org.

1 0

actions
by Nuno Silva 04 Dec '02

04 Dec '02

I'm using an action that generates a random number, for the user to discover it. each time he guesses, if the number is diferent that the one generated, it is suposed to call itself again to try again, but each time I call it, it generates the number again. how can I overcome this? sample of the code: descodificador = do c1 <- gera c2 <- gera c3 <- gera c4 <- gera "code in the middle to check if the number is equal" if not than go down... descodificador --------------------------------- Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now

2 1

De-biforestation
by oleg＠pobox.com 04 Dec '02

04 Dec '02

Deforestation is usually defined as the elimination of an intermediate data structure between a single producer and a single consumer. Jorge Adriano showed an interesting example of one producer feeding two independent consumers. The two streams of data are mutually dependent. Furthermore, the rate of production is non-uniform: Generating an element in one stream may require generating trillion items in the other stream. If the first consumer is being evaluated, that consumer causes the evaluation of the producer until the latter yields an item. The trillion of items incidentally produced in the other stream have to be stored somewhere. In more OS-centric terms, the evaluator can't generally reduce two expressions in parallel. If it chooses one stream consumer, the other consumer is set aside. While the first stream consumer "waits" for a value, the second stream consumer is "blocked", and hence the corresponding stream must be buffered, sometimes at great expense. We will consider the problem of one producer and two consumers in the general setting. We derive a deforested version, which no longer needs to buffer any produced items. The retaining profile shows no space leaks. We also consider "parallel" writing of streams into two distinct files. Our solution is safe but the i/o is effectively interleaved. The deforested solution is indeed _derived_, via a sequence of equivalent transformations. In fact, the derived code worked on the first try. In the original example, the two consumers were writing received data into the their own files. Let us first generalize the problem. We first assume that the two mutually-recursive streams are specified by a transition function transition:: (a,b) -> (a,b) one the state of a _pre-stream_. The state of the latter is a pair of values. We obtain two streams in question by splitting the pre-stream into the fst and the snd components, and applying two different filters. The filtering part makes the rate of production non-uniform. > import IO > infixr 0 $^$ > -- Can't do pattern-matching-deconstruction on state because > -- pattern-matching is eager! > (f,g) $^$ state = (f $ fst state, g $ snd state) > > stream transition (p,q) state@(xstate,ystate) = > ((filter p) . (xstate :), (filter q) . (ystate :)) $^$ > (stream transition (p,q) $ transition state) For illustration, we will try the following transition function and the filters: > trans1 (x,y) = (y+1,x) > st_p x = x `mod` 1997 == 0 > st_q = const True The streams are: without filtering Main> take 9 $ fst $ stream trans1 (\x->True,\x->True) (0,0) [0,1,1,2,2,3,3,4,4] Main> take 9 $ snd $ stream trans1 (\x->True,\x->True) (0,0) [0,0,1,1,2,2,3,3,4] and with filtering: Main> take 9 $ snd $ stream trans1 (st_p,st_q) (0,0) [0,0,1,1,2,2,3,3,4] (443 reductions, 782 cells) Main> take 9 $ fst $ stream trans1 (st_p,st_q) (0,0) [0,1997,1997,3994,3994,5991,5991,7988,7988] (6262603 reductions, 9198209 cells, 42 garbage collections) Generating each item in the first stream indeed generates quite a few (thousand) items in the second stream. In our first example, the consumers will sum the corresponding streams: > consumers_sum stream n = (f,f) $^$ stream trans1 (st_p,st_q) (0,0) > where f = sum . take n More generally, we can write a consumer using a function sfoldl (stream fold). The function is similar to the regular foldl but can be applied to infinite streams (and so it needs a predicate to tell the termination). > sfoldl sf st z (x:_) | sf x z = z > sfoldl sf st z (x:xs) = sfoldl sf st (st z x) xs Our consumers_sum function can be written as > consumers_sum' stream n = > (fst,fst) $^$ (f,f) $^$ stream trans1 (st_p,st_q) (0,0) > where > f = sfoldl sf st (0,0) > sf x (_,count) = count >= n > st (sum,count) val = (val+sum,count+1) In general, the essence of producing and consuming two streams is captured by an expression (sfoldl sf1 st1 z1, sfoldl sf2 st2 z2) $^$ stream trans (p,q) (x,y) === { definition of stream } === (sfoldl sf1 st1 z1, sfoldl sf2 st2 z2) $^$ ( ((filter p) . (x :), (filter q) . (y :)) $^$ stream trans (p,q) $ trans (x,y) ) === { We know a$(b$c) === (a.b) $ c Similarly, a $^$ (b $^$ c) === ((((.),(.)) $^$ a) $^$ b) $^$ c or, in a less bizarre notation a $^$ (b $^$ c) === ((fst a . fst b),(snd a . snd b)) $^$ c } === ((sfoldl sf1 st1 z1) . (filter p) . (x :), (sfoldl sf2 st2 z2) . (filter q) . (y :)) $^$ stream trans (p,q) $ trans (x,y) where we also took advantage of the associativity of composition. We note that two folds, sfoldl and the fold implicit in filter, can be fused: sfoldlf sf st p z =def= (sfoldl sf st z) . (filter p) Then sfoldlf sf st p z (x:xs) === (sfoldl sf st z) $ (filter p (x:xs)) === { definition of filter, lifting the condition } if p x then (sfoldl sf st z (x:filter p xs)) else (sfoldl sf st z (filter p xs)) === { definition of sfoldlf } if p x then (sfoldl sf st z (x:filter p xs)) else sfoldlf sf st p z xs === { definition of sfoldl } if p x then if sf x z then z else sfoldl sf st (st z x) (filter p xs) else sfoldlf sf st p z xs === if p x then if sf x z then z else sfoldlf sf st p (st z x) xs else sfoldlf sf st p z xs We have just obtained a recursive equation for sfoldlf. Note that sfoldlf (just as sfoldl) are undefined for empty lists. We can now write our producer-consumers problem as (sfoldl sf1 st1 z1, sfoldl sf2 st2 z2) $^$ stream trans (p,q) (x,y) === ((sfoldlf sf1 st1 p z1) . (x :), (sfoldlf sf2 st2 q z2) . (y :)) $^$ stream trans (p,q) $ trans (x,y) We already see a notable simplification. The streams are now producing at the same rate! So we can replace a pair of two streams with one stream of pairs. But we can do better. Let's look at the equation for sfoldlf again sfoldlf sf st p z (x:xs) === if p x then if sf x z then z else sfoldlf sf st p (st z x) xs else sfoldlf sf st p z xs The equation suggests a useful generalization of sfoldl: > sfoldl' sf st z True _ = z > sfoldl' sf st z done (x:xs) | sf x z = sfoldl' sf st z True xs > sfoldl' sf st z done (x:xs) = sfoldl' sf st done (st z x) xs We can now introduce sfoldlf' sf st p done z =def= (sfoldl' sf st done z) . (filter p) which obeys the equation sfoldlf' sf st p done z (x:xs) === if done then z else if p x then if sf x z then sfoldlf' sf st z True xs else sfoldlf' sf st p done (st z x) xs else sfoldlf' sf st p done z xs === { pushing conditions inside } sfoldlf' sf st p (done || (p x && sf x z)) (ss sf st p done z x) xs where ss sf st p done z x = if done then z else if p x then if sf x z then z else (st z x) else z Something remarkable has happened: we've got the left recursion! We should note that the introduction of a 'done' argument is not unlike a common trick in physics of converting a value into a functional with the help of the Dirac delta-function. We can now write sfoldlfs sf1 st1 done1 z1 sf2 st2 done2 z2 (p,q) trans state@(x,y) =def= (sfoldl' sf1 st1 done1 z1, sfoldl' sf2 st2 done2 z2) $^$ stream trans (p,q) (x,y) === ((sfoldl' sf1 st1 done1 z1) . (filter p) . (x :), (sfoldl' sf2 st2 done2 z2) . (filter q) . (y :)) $^$ stream trans (p,q) $ trans (x,y) === (\tail-> sfoldlf' sf1 st1 p done1 z1 (x:tail), \tail-> sfoldlf' sf2 st2 q done2 z2 (y:tail) ) $^$ stream trans (p,q) $ trans (x,y) === (\tail-> sfoldlf' sf1 st1 p (done1 || (p x && sf1 x z1)) (ss sf1 st1 p done1 z1 x) tail, \tail-> sfoldlf' sf2 st2 q (done2 || (q y && sf2 y z2)) (ss sf2 st2 q done2 z2 y) tail ) $^$ stream trans (p,q) $ trans (x,y) === { eta reduction } (sfoldlf' sf1 st1 p (done1 || (p x && sf1 x z1)) (ss sf1 st1 p done1 z1 x), sfoldlf' sf2 st2 q (done2 || (q y && sf2 y z2)) (ss sf2 st2 q done2 z2 y) ) $^$ stream trans (p,q) $ trans (x,y) === { definition of sfoldlfs } sfoldlfs sf1 st1 (done1 || (p x && sf1 x z1)) (ss sf1 st1 p done1 z1 x) sf2 st2 (done2 || (q y && sf2 y z2)) (ss sf2 st2 q done2 z2 y) (p,q) trans (trans state) And we have obtained the primitive recursive solution with no lists whatsoever. We have achieved the full deforestation! Now we just have to program it: > ss:: (x->z->Bool) -> (z->x->z) -> (x->Bool) -> Bool -> z -> x -> z > ss sf st p done z x = > if done then z > else if p x then (if sf x z then z else (st z x)) > else z *> sfoldlfs sf1 st1 True z1 sf2 st2 True z2 (p,q) trans state = (z1,z2) *> sfoldlfs sf1 st1 done1 z1 sf2 st2 done2 z2 flt@(p,q) trans state@(x,y) = *> sfoldlfs sf1 st1 (done1 || (p x && sf1 x z1)) (ss sf1 st1 p done1 z1 x) *> sf2 st2 (done2 || (q y && sf2 y z2)) (ss sf2 st2 q done2 z2 y) *> flt trans (trans state) Or, factoring out common arguments > sfoldlfs sf1 st1 done1 z1 sf2 st2 done2 z2 flt@(p,q) trans state = > aux done1 z1 done2 z2 state > where > aux True z1 True z2 _ = (z1,z2) > aux done1 z1 done2 z2 state@(x,y) = > aux (done1 || (p x && sf1 x z1)) (ss sf1 st1 p done1 z1 x) > (done2 || (q y && sf2 y z2)) (ss sf2 st2 q done2 z2 y) > (trans state) > consumers_sum'' trans n = > (fst,fst) $^$ > sfoldlfs sf st False (0,0) sf st False (0,0) (st_p,st_q) trans (0,0) > where > sf x (_,count) = count >= n > st (sum,count) val = (val+sum,count+1) Let us now compare the forested version consumers_sum' with the deforested version consumers_sum'' in performance, using GHCi: *Main> consumers_sum' stream 8 (31952,12) (0.26 secs, 11934436 bytes) *Main> consumers_sum'' trans1 8 (31952,12) (0.18 secs, 7695084 bytes) *Main> consumers_sum' stream 16 (127808,56) (0.86 secs, 32660328 bytes) *Main> consumers_sum'' trans1 16 (127808,56) (0.32 secs, 14354768 bytes) *Main> consumers_sum'' trans1 16 (127808,56) (0.32 secs, 14579336 bytes) *Main> consumers_sum' stream 32 (511232,240) (3.32 secs, 97786900 bytes) *Main> consumers_sum'' trans1 32 (511232,240) (0.69 secs, 28490268 bytes) It appears that the deforestation was worth it. The resulting code needs less memory and executes notably faster. In another test, we compile the code with optimization and profiling enabled: > main = do > print $ consumers_sum' stream 32; > print $ consumers_sum'' trans1 32 The result indicate that consumers_sum' is responsible for only 15.4% of the time and 13.7% of the allocations (most of which are due to the transition function). What's more exhilarating, the retainer profile of "print $ consumers_sum'' trans1 32" shows nothing. Nothing is retained. In contrast, the retainer profile of "print $ consumers_sum' stream 32;" shows large a large space leak of 12 MB of retained memory. But what about the original question by Jorge Adriano? Can we write two stream into two different files, and _safely_ interleave the i/o? Yes, we can. We need a version of sfoldlfs that shares the singe seed between two consumers: > sfoldlfs1 sf1 st1 done1 sf2 st2 done2 flt@(p,q) trans z state = > aux done1 done2 z state > where > aux True True z _ = z > aux done1 done2 z state@(x,y) = > aux (done1 || (p x && sf1 x z)) (done2 || (q y && sf2 y z)) > (ss sf2 st2 q done2 (ss sf1 st1 p done1 z x) y) > (trans state) > write_in_two_files file1 file2 trans max = > (\(h1,h2) -> hClose h1 >> hClose h2) =<< > sfoldlfs1 sf (st fst) False sf (st snd) False (st_p,st_q) trans > (do h1 <- openFile file1 WriteMode; > h2 <- openFile file2 WriteMode; > return (h1,h2)) (0,0) > where > sf x _ = x >= max > st selector two_handles val = > do > handles <- two_handles > hPrint (selector handles) val; > return handles *> main = write_in_two_files "/tmp/a1" "/tmp/a2" trans1 80000 The profile shows that the i/o is interleaved indeed.

1 0

RE: Implementing forward refs in monadic assembler and interprete r
by Ashley Yakeley 02 Dec '02

02 Dec '02

At 2002-12-02 00:37, Simon Peyton-Jones wrote: >Main things in the next major release are > mdo > Template Haskell Can mdo be done in TH? It would be nice if there were some kind of macro system for this sort of thing; I know there's also a syntax for Arrows that requires a preprocessor... -- Ashley Yakeley, Seattle WA

1 0

RE: Implementing forward refs in monadic assembler and interprete r
by Simon Peyton-Jones 02 Dec '02

02 Dec '02

| For the moment I've chosen to do the mdo desugaring manually instead | of moving to the CVS ghc (or exclusively using Hugs.) Will mdo be in | the next release? (That's 5.04.2, I think -- though I've recently | been confused about GHC releases.) The relevant revision to Lex.lhs | is pretty recent and doesn't seem to be on ghc-5-04-branch. mdo won't be in 5.04.2; these patch releases are meant to be bug fixes only. As I mentioned in an earlier message, we are undecided about when to invest the effort to make another major release. The more people that yell, the sooner we will do it! Main things in the next major release are mdo Template Haskell a fairly major fix to the newtype-deriving mechanism (it's a bit broken in 5.04) Simon

1 0

someone help me
by Nuno Silva 30 Nov '02

30 Nov '02

I'm relatively new to haskell and have some questions: 1- I want to convert an integer to a char( if I say prelude> convert 4 ... it should return prelude> '4') 2- and vice versa? 3- I want to define a function that stores in a variable the following... pretended: (Int,Int,Int,Int) give :: IO Int give = randomRIO (0,9) main = do k <- (give,give,give,give) actualy this isn't possible. but can someone please tell how to do this??? thank you very much --------------------------------- Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now

3 2

RE: Implementing forward refs in monadic assembler and interprete r
by Erkok, Levent 30 Nov '02

30 Nov '02

> For the moment I've chosen to do the mdo desugaring manually instead > of moving to the CVS ghc (or exclusively using Hugs.) Will mdo be in > the next release? (That's 5.04.2, I think -- though I've recently > been confused about GHC releases.) The relevant revision to Lex.lhs > is pretty recent and doesn't seem to be on ghc-5-04-branch. The current CVS-head for GHC contains the mdo implementation (thanks to Simon PJ). I'm not sure what they're going to call it (i.e. 5.04.2, or 5.05, etc.), but the next release will have mdo for sure. I don't know when they'll make a new release though, maybe one of the Simon's can comment on that? -Levent.

1 0

Re: Implementing forward refs in monadic assembler and interprete r
by Mike Gunter 29 Nov '02

29 Nov '02

Thank to all who replied. After giving up on continuations (which make everything but forward branches seductively easy), I settled on something with the following interface: > class MonadGoto lbl gm | gm -> lbl where > liftG :: Monad m => m () -> gm m () > label :: Monad m => gm m lbl > gotoIf :: Monad m => m Bool -> lbl -> gm m () Note that, unlike a monad transformer, only ()-producing computations can be embedded. This limitation reflects the problem of def-use across basic-block boundaries. Of course, computations in the embedded monad can be done there, then lifted, e.g.: > ;top <- label > ; ... > ; liftG (do v <- m1 > ; ... > ; m2 v > ; ...) > ; brPositive r3 top > ; ... I've implemented MonadGoto as > newtype GotoT m a = GotoT { unGotoT :: WriterT [GotoLbldT m ()] LblSupply a } > > data GotoLbldT m a = Label !Lbl > | GotoIf (m Bool) !Lbl > | Inner (m a) > > wrL = GotoT . tell . (:[]) > instance MonadGoto Lbl GotoT where > liftG = wrL . Inner > label = GotoT $ lift fromSupply >>= \l -> tell [Label l] >> return l > gotoIf p l = wrL $ GotoIf p l where GotoLbldT gets converted to > data GotoLbldRunT m a = ThenLabel { before :: m (), next :: GotoLbldRunT m a, labelLbl :: !Lbl } > | ThenGotoIf { before :: m (), next :: GotoLbldRunT m a > , gotoPred :: m Bool, gotoLbl :: !Lbl } > | Tail (m a) which can then be run > type JumpArray m v = [(Lbl, m v)] > runGM :: Monad m => JumpArray m v -> GotoLbldRunT m v -> (JumpArray m v, m v) . My Verilog has delayed branches (because that's the easy thing to do.) So, I implemented a DelayedGotoT which provides for delayed branches by shuffling the "GotoLbldT"s and using a StateT to save the result of the gotoPred evaluation. "Erkok, Levent" <levent.erkok(a)intel.com> writes: > You need a fairly recent version of Hugs to run this > example, November 2002 release would do (try with -98). For ghc, you need > the CVS version, or wait till the next release. As far as I know, no other > implementation supports mdo, nor there are any plans for it. For the moment I've chosen to do the mdo desugaring manually instead of moving to the CVS ghc (or exclusively using Hugs.) Will mdo be in the next release? (That's 5.04.2, I think -- though I've recently been confused about GHC releases.) The relevant revision to Lex.lhs is pretty recent and doesn't seem to be on ghc-5-04-branch. thanks, mike

1 0

Merhaba;
by www.erotikmagaza.com 26 Nov '02

26 Nov '02

Sevgili internet kullanicilari; Bu mail sizlere rahatsizlik verdiyse �ok �ok �z�r diliyoruz. Biz Erotik �r�n satan bir firmayiz. Avrupada insanlarin erotik marketlerden satin alarak gerek ihtiya�larinda gerekse hediyelik esya olarak g�venle kullandiklari erotik �r�nleri ithal ederek �lkemizde m�sterilerimizin begenisine sunuyoruz. Eger ilginizi �ekerse sizleride Web sitemizi gezmeye davet ediyoruz. Saygilarimizla http://www.erotikmagaza.com

1 0

Random Color
by Mike T. Machenry 22 Nov '02

22 Nov '02

I am trying to construct an infinate list of pairs of random colors. I am hung up on getting a random color. I have: data Color = Blue | Red | Green deriving (Eq, Ord, Show) am I supposed to instantiate a Random class instance from color? I am not sure exactly how the random number generator works. Is this the best way to go about it? Should I instead get a random Integer index of an array of Colors? thanks, -mike

3 5

Re: Sets
by Christopher Milton 19 Nov '02

19 Nov '02

This might be of some help, circa Haskell 1.2, SetMap.hs - sets and maps in Haskell Author: Nick North ftp://ftp.cs.chalmers.se/pub/haskell/library/SetMap.hs Chris ===== Christopher Milton cmiltonperl(a)yahoo.com __________________________________________________ Do you Yahoo!? Yahoo! Web Hosting - Let the expert host your site http://webhosting.yahoo.com

1 0

newbie question re linear algebra in Haskell
by Jeff Austin 19 Nov '02

19 Nov '02

I'm a complete newcomer to Haskell, having learned about only recently. I'm intrigued by the possibility of in using it for numerical applications, specifically linear algebra. I understand that (at least in its present state) Haskell 98 isn't competitive with imperative languages when it comes to primitive matrix-vector operations, which often rely on destructive updating. It strikes me that one approach that takes advantage of the strengths of both paradigms would be create an imperative subsystem to handle primitive operations, then create a functional matrix algebra layer on top of it. Based on my very limited understanding of Haskell, I envision a linear algebra library which would work something like this: (1) Create a wrapper for the standard BLAS library. Since an optimized BLAS is available for various architectures with the same interface, this would provide a way of optimizing the library for a particular architecture. (2) Encapsulate this (using a monad?) in a way that provides safe access to the imperative routines from Haskell. (3) Using this as a foundation, define a matrix algebra in a "smart" way which avoids redundant evaluations. For instance, if A is an arbitrary invertible matrix, we know that A - A = 0 A + 0 = A A * Indentity = A A * inverse A = Identity so if we write an expression like "X = A * inverse A + B - B" we should expect our library to compute the correct result, "Identity" without actually performing any inversion, multiplication or elementwise matrix addition. (4) Create a Haskell module which exposes a pure functional interface, hiding the imperative subsystem from the user. My questions are: (a) Is this idea sound? (b) Has someone already done this? (c) Is there a better way of doing efficient linear algebra in Haskell? I would greatly appreciate any advice in this regard. --Jeff Austin

4 3

Network objects (was RE: deriving (...))
by Garner, Robin 19 Nov '02

19 Nov '02

> BTW, those typed returns on sockets that Shawn mentioned sounded > interesting. IIRC Modula-3 also had some approach to worrying about data > exchange between older and newer versions of the same program. Modula-3 (actually the Network Objects library) used subtyping to handle this. Successive versions of an interface are declared as subtypes of the previous version, and the type system will resolve a network object binding to the appropriate version. Hard to see how this could be done in a purely functional setting. robin

2 1

RE: Bug? [was: Implicit params]
by Simon Peyton-Jones 18 Nov '02

18 Nov '02

Now fixed in the HEAD, and will be in 5.04.2 Thanks for pointing it out. Simon | -----Original Message----- | From: Jorge Adriano [mailto:jadrian@mat.uc.pt] | Sent: 14 November 2002 21:10 | To: Iavor S. Diatchki | Cc: Haskell Cafe; glasgow-haskell-users(a)haskell.org | Subject: Bug? [was: Implicit params] | | On Thursday 14 November 2002 18:47, Iavor S. Diatchki wrote: | > hello, | > | > > Well, actually you must be right since the pure field defines a pure | > > (projection) function... Hmmm, ok, can someone explain this to me, | > > | > > data E s = E{ | > > refi :: STRef s Int, | > > refc :: STRef s Char, | > > m :: Int | > > } | > > | > > -- this is fine, obviously... | > > pure :: E s -> Int | > > pure e = m e | > > | > > -- but this is not... | > > pure2 :: (?e :: E s) => Int | > > pure2 = m (?e) | > > | > > Why exactly isn't this allowed? What is the workaround? | > > Error msg: | > > ------------------------------------------------------------------------ | > > Ambiguous constraint `?e :: E s' | > > At least one of the forall'd type variables mentioned by the | > > constraint | > > must be reachable from the type after the '=>' | > > In the type: forall s. (?e :: E s) => Int | > > While checking the type signature for `pure2' | > > Failed, modules loaded: none. | > > ------------------------------------------------------------------------ - | > >- | > | > this seems like a bug in GHC. Hugs is happy with it. The "s" in the | > pure2 signature is not ambiguous because it is determined when you give | > the value of the implicit parameter. in fact the way i think about | > implicit parameters is simply as a nice notation for computations in the | > environment monad, so in my mind the above two definitions are pretty | > much the same. | | Thanks Iavor! | GHC people, can you confirm this, Is it a bug? I'm using Ghc 5.0.4 SuSE rpms. | Is there a workaround? | | Thanks, | J.A. | | | | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

2 1

GTKHS
by Martin Huschenbett 18 Nov '02

18 Nov '02

Hi! Is there meanwhile any tutorial on GTK-HS? Huschi

2 1