September 2007 - Haskell-Cafe

Re: [Haskell-cafe] Troubles understanding memoization in SOE
by Peter Verswyvelen 25 Sep '07

25 Sep '07

Thanks Paul & Paul for the answers. I'll certainly read the paper Paul Liu reported. I just deleted 100 lines of text which explained my problem in more detail, and while I was explaining it, I answered it myself. Typical. I thought the lambda function that memo1 returns would be called over and over again, and instead of reevaluating the stream from the beginning, it would just return the stream since it is in the cache, but actually it just gets called twice in recursive situations: the first time it evaluates y = f x, stores the thunk in the cache, and returns the thunk, the second time it finds the same thunk in the cache, and then computation of the rest of the stream continues without consulting the cache anymore right? From my clumsy explanation you can see that I'm still thinking too imperative, too eager. Haskell is more lazy than I am, which is an incredible achievement :-) It would really help if I could see the lazy computation; do you think this kind of memo code is traceable using HAT? I'll guess I'll have to check out arrows / yampa again. A year ago I did not understand a single thing in those papers, but I should try it again now I read the SOE book :-) Thanks a lot, Peter

2 2

Shouldnt this be lazy too?
by Vimal 25 Sep '07

25 Sep '07

Hi all, I was surprised to find out that the following piece of code: > length [1..] > 10 isnt lazily evaluated! I wouldnt expect this to be a bug, but in this case, shouldnt the computation end when the length function evaluation goes something like: > 10 + length [11..] ? -- -- Vimal

13 26

representing differencial equations in haskell
by Thomas Girod 25 Sep '07

25 Sep '07

Hi there. Let's say I have mathematical model composed of several differential equations, such as : di/dt = cos(i) dc/dt = alpha * (i(t) - c(t)) (sorry my maths are really bad, but I hope you get the point) I would like to approximate the evolution of such a system iteratively. How would you do that in haskell ? cheers, Thomas

4 3

Template Haskell newbie questions
by Hugo Pacheco 25 Sep '07

25 Sep '07

Hi all, I'm try to write some function in TH that I don't even know if it is possible. Consider the example from the tutorials sel 1 2 = [| \(x,_) -> x |] sel 2 2 = [| \(_,x) -> x |] If I want to write some function that will dynamically create a selection function according to its arguments and do something with it. fsel :: Int -> Int -> x fsel x y = $(sel x y) ... Imagine a stupid case for fsel :: Int -> Int -> Int fsel x y = $(sel x y) (x,y) The compiler complains that GHC stage restriction: `x' is used in a top-level splice, and must be imported, not defined locally In the first argument of `sel', namely `x' In the expression: $[splice](sel x y) (x, y) In the definition of `fsel': fsel x y = $[splice](sel x y) (x, y) what, indeed, makes sense. However, is there a possible way to generate Haskell functions from TH without having to instanciate its arguments? Thanks, hugo

3 4

Re: [Haskell] Re: [Haskell-cafe] PROPOSAL: Rename haskell@to haskell-announce@
by Andrzej Jaworski 25 Sep '07

25 Sep '07

> (...) I subscribe to haskell so as not to miss > anything important, and when something I'm interested in moves to > haskell-cafe, (...) Wrong assumption. Even if you know the author it makes more sense to wait and see how others respond to his newest input. This means that you should check up Cafe first to see if it is worth moving to Cafe;-) Another false believe is that best things appear first in Haskell@. On the contrary, long threads in Cafe indicate that participants know what to expect of each other and pick up a thread on the same principle some pick up new Stephen King novel. Before reviews on the amazon.com publishers had to result to tricks like labelling a collection of science fiction stories by various lesser known authors with Isaac Asimov name even when his only cotribution was a preface. As I said ealier - really big traffic will need intelligent screening and lebbeling. And to what pathology it leads when unattended see comp.soft-sys.matlab. Cheers, -Andrzej Jaworski

1 0

Re: [Haskell-cafe] Very crazy
by Miguel Mitrofanov 25 Sep '07

25 Sep '07

> How can one function have more than one type signature? It's called "polymorphism".

1 0

Re: [Haskell] simple function: stack overflow in hugs vs none in ghc
by Isaac Dupree 25 Sep '07

25 Sep '07

(redirecting to haskell-cafe) Tom Pledger wrote: > | > sqnc p ts = let ( r, ts' ) = p ts in case r of > | > Nothing -> ([],ts') > | > Just x -> let (r',ts'') = (sqnc p ts') in > (x:r', ts'' ) > : > > > I don't know how ghc is avoiding the stack overflow, but it looks like > it's caused by strict pattern matching. The first call to sqnc wants > reassurance that the second call is returning a pair (as opposed to _|_) > before deciding that it's OK to return a pair. > > Try putting a tilde in the recursive call, to make the pattern lazy. > > let ~(r',ts'') = ... "let" already is lazy -- putting a tilde there has no effect. In fact that tuple is produced from (x:r', ts'' ) even if (sqnc p ts') is _|_. (The "let" at the beginning of the function can be discovered to be strict by strictness analysis (which GHC does), because "r" is analyzed by the "case", and the pair needs to be deconstructed to find the value of "r".) Isaac

4 9

Re: [Haskell-cafe] Very crazy
by Andrew Coppin 25 Sep '07

25 Sep '07

Chaddaï Fouché wrote: > 2007/9/25, Andrew Coppin <andrewcoppin(a)btinternet.com>: > >> Forget PrintfType - I can't even understand the haddoc page yet! >> >> (printf performs I/O, yet it is outside the I/O monad. It seems to >> accept an arbitrary number of arguments, which is obviously impossible. >> It's *almost* as scary as the original. Although the original will crash >> your program is you use it wr... oh, so does this one. Cool.) >> > > printf don't always perform IO : if you ask it for a String it will > happily turn into sprintf for you, if you use it in the IO Monad, it > will indeed perform IO, but there's nothing fundamentally IO bound in > printf logic. > That's even *more* impossible... o_O > Well printf crash your program because it can't check if you're > arguments correspond to the chain at compile time (quite obviously > since the chain itself can be dynamic), but it won't corrupt memory or > anything from this kind, so it's arguably much better than the C one. > All I know is that in C, any attempt to use printf() has an 80% probability of causing the machine to dump the entire contents of RAM to stdout until it happens upon a zero byte. Probably the most annoying part was all the 0x07 bytes in there. Do you know what a room full of beeping PCs is like? I do... Sometimes when we got bored, we would "race" our PCs to see which one would stop beeping first. (But they stopped that when I worked out how to make mine "win" every race...)

1 0

bindings to the xmms_remote API, GList and something more
by Andrea Rossato 25 Sep '07

25 Sep '07

Hi, I think there's a huge flaw in the Haskell design: if you conceive such a powerful and expressive programming language, one that empowers even the most ignorant computer user to write complex applications, instead of sticking to basic shell scripting - the only domain graspable by such stupid pseudo-coders -, well you end up with messages like this one flooding the mailing list you are providing for your users. This is what I call inevitability... To make a short story long, I needed some client for the Audacious media player, something I could use to remote control it and, since I'm addicted to Haskell, instead of surfing the web to find a suitable client I surfed the web to find the API documentation. With the FFI, in a matter of half an hour I had the client with all the commands I needed. (The long introduction was written just to let you know that you'd better not waste too much time with what follows...;-) So far so good, but, since everything was so easy, I thought: "Why not packaging a library (and the client) for Hackage? Maybe others may be willing to use it, maybe..." And then the problems start, problems due to my basic ignorance of the C language ("How comes you dare to import in Haskell functions written in a language you don't even understand?" you should replay!). Basically there are two C types I'm having problem with: 1. GList: even though I read the gtk2hs code I do not exactly understand how to create a GList to feed to this function: void xmms_remote_playlist_add(gint session, GList * list); As far as I understand this function takes a session number and a list of files' names. Still I seem not to be able to create a wrapper function around the imported one. 2. what a "gchar ** list" is? As far as my C goes, it should be an array of strings, right? What should I use on the Haskell side, newArray? void xmms_remote_playlist(gint session, gchar ** list, gint num, gboolean enqueue); 3. c2hs v. hsc2hs? Which should I prefer? In c2hs I write {#pointer * GList#}. What is the equivalent in hsc2hs? 4. As I said, I solved my personal problem: I imported enough functions to create the client I needed. Releasing the library is just something I would like to do, to give something back to the Haskell community. So this is something useful if I can provide robust and well designed code. This is a name space question: what is the name of the exported module? Sound.XMMS, Sound.XmmsRemote? Is the some kind of convention I should be referring to? Sorry if I was too long. Moreover I'm asking help I need just to publish a piece of software that is probably useless. Do not waste your time. But if you do, well... thank you, I really appreciate. Andrea

2 5

PROPOSAL: Rename haskell@ to haskell-announce@
by Neil Mitchell 24 Sep '07

24 Sep '07

Hi I've just replied to another first poster with "wrong list". Its entirely not their fault, but its also probably a bit off-putting that your very first post gets a (very polite) "you got it wrong" message. To steal the reasons and explanations from Ian: --------------------- pretty much what would be haskell-announce@ anywhere else, and haskell-cafe@ what would be haskell@ elsewhere. A haskell-cafe@ list elsewhere would probably be where the haskell@ people discuss things which aren't actually related to Haskell (like e.g. the demon.local newsgroup). I think the number of posts in the wrong place would be lower if these were more conventionally named (although there aren't a lot of them anyway). I think it would make sense to: * Rename haskell@ to haskell-announce@, and redirect mails from haskell@ to haskell-announce@ for some period. (more, but can be done step by step in the future) --------------------- I suggest that following somewhat the http://www.haskell.org/haskellwiki/Library_submissions we allow 4 weeks for discussion, and depending on the result we then move the mailing list. Bug to track: http://hackage.haskell.org/trac/ghc/ticket/1732 Deadline: 23 October 2007 Thanks Neil

6 8