September 2006 - Haskell-Cafe

Haskore
by David Curran 22 Sep '06

22 Sep '06

Hi I have been trying to learn haskell (tip over the vending machine) for a while and eventually decided the Haskore music library might be a good way to start understating the language. I am using windows and hugs98. The IOExtensions.hs file will not work under windows. Any ideas on how to make it work or is this library *nix only? Regards David ----------------------------------------------------------------------------- -- -- Replacement for IOExtensions.hs of hugs for ghc -- -- Implements only the functions necessary for Haskore! -- -- Does not work under Windows (where there is a difference between -- binary and text files) -- -- Suitable for use with Hugs 98 ----------------------------------------------------------------------------- module IOExtensions( writeBinaryFile, openBinaryFile, readBinaryFile ) where import System( getArgs ) import IO( Handle, IOMode(WriteMode), openFile, hPutStr, hClose ) import IOExts openBinaryFile :: FilePath -> IOMode -> IO Handle openBinaryFile path mode = openFileEx path (BinaryMode mode) writeBinaryFile :: FilePath -> String -> IO () writeBinaryFile = writeFile -- this does not work on Windows!!! readBinaryFile :: FilePath -> IO String readBinaryFile = readFile -- this does not work on Windows!!! -- It chews up lies and spits out the gristle of truth http://liveatthewitchtrials.blogeasy.com/

5 4

Trouble with simple embedding of shell in haskell
by George Brewster 22 Sep '06

22 Sep '06

I'm just tried writing a function to allow convenient embedding of shell commands, but I'm running into behavior I don't really understand somewhere at the intersection of lazy evaluation, IO, and threading. The function in question is: sh :: String -> String -> IO String sh cmd = \input -> do (stdin, stdout, _, pid) <- runInteractiveCommand cmd forkIO $ hPutStr stdin input >> hClose stdin -- get exit status to prevent zombie accumulation forkIO $ waitForProcess pid >> return () hGetContents stdout Simple uses like this work (prints "hi" and returns): return "hi" >>= sh "cat" >>= putStrLen However, this blocks: return "hi" >>= sh "cat" >>= sh "cat" >>= putStrLen this prints "hi\nblar" and returns return "hi" >>= sh "cat" >>= (\s -> putStrLn "blar" >> sh "cat" s) >>= putStrLn and this always returns but sometimes prints nothing and sometimes prints "hi\n" sh "echo hi" "there" >>= sh "cat" >>= putStrLn All these were tried with ghc 6.4.1 compiled with -threaded Just wondering if anybody has a good explanation of this range of behavior and/or a working rewrite of this function. -George

3 4

Non-existant existential?
by Jim Apple 21 Sep '06

21 Sep '06

Consider the following: data SimpExist a = SimpExist (forall x . x -> a) f :: SimpExist Bool f = SimpExist (const True) g = SimpExist id What is the type of g? In a similar example, GHC tells me it is of type SimpExist c. Yet, I can't unify it with any other SimpExist c'. It seems to me that this is something like exists x . SimpExist x, and is similar to: data ExistWrap = forall a . ExistWrap (forall x . x -> a) Jim

2 3

Re: Haskell web forum
by Albert Lai 21 Sep '06

21 Sep '06

As requested, I continue here the thread on the proposal for a web forum. You will soon enough find out what I think of web forums. Let me first show you what this mailing list looks like to me using software of my choice: ! [ 192: Mark Carroll ] [Haskell-cafe] Card trick ! [ 37: Thomas Conway ] Re: [Haskell-cafe] How can we detect and fix E [ 19: Albert Lai ] Re: [Haskell-cafe] Either e Monad E [ 19: Deokhwan Kim ] [ 62: Bas van Dijk ] E [ 47: apfelmus@quantentunn] [Haskell-cafe] Re: Optimization problem [ 51: Ross Paterson ] E [ 42: Conor McBride ] [ 56: Robert Dockins ] [ 42: Conor McBride ] [ 61: apfelmus@quantentunn] [Haskell-cafe] Re: Optimization prob [ 14: Ross Paterson ] [ 154: Jan-Willem Maessen ] Re: [Haskell-cafe] Traversing a graph in STM [ 52: Josef Svenningsson ] [ 18: Bulat Ziganshin ] Re[2]: [Haskell-cafe] Traversing a graph [ 26: Sebastian Sylvan ] [ 22: Ashley Yakeley ] [Haskell-cafe] Re: Wiki contact An elaboration of the cool features I have always enjoyed is in order: On browsing threads, both inter-thread jumps and intra-thread navigation, the screenshot speaks for itself. (If for any perverted reason I want the messages listed chronologically rather than well-organized, the software can do it too.) The messages marked with "E" are messages I have just read. With this mark, they will not show up next time I browse this mailing list again. Take note that I can mark and unmark any subset of messages, unlike certain people's linear-time idea that one single timestamp distinguishes "already read" from "new" across the board. (I have always heard that non-linear thinking is superior to linear thinking. I think I believe it for at least this application.) (Is this an example of what is meant when someone said that a web forum requires "less technical knowledge", i.e., programmers on a web forum will not need to know about subsets?) (Speaking of which, is the thread-browsing part also an example, i.e., programmers on a web forum will not need to know about trees and forests?) Though the marked messages will not show up next time, they are not thrown away yet. They are kept on my disk for quite a while. (I get to set how many days they stay.) The software offers several ways to show them, but I think the most useful one is this. Suppose I look at apfelmus's message on "Optimization problem" (two are shown, I'm referring to the first one), and wonder, "gee, what is it replying to?" To find out, I position myself to that message, then press a button, then the software will show the desired parent message. In fact, the updated screen looks like this (I just include the thread in question): E [ 18: Ross Paterson ] Re: [Haskell-cafe] Re: Optimization problem E [ 47: apfelmus@quantentunn] [ 51: Ross Paterson ] E [ 42: Conor McBride ] [ 56: Robert Dockins ] [ 42: Conor McBride ] [ 61: apfelmus@quantentunn] [Haskell-cafe] Re: Optimizatio [ 14: Ross Paterson ] So even if apfelmus's message quotes nothing from Ross's, I can still find out, provided it is recently enough to be still on my disk. Take note that, on a linear-thinking web forum, there could be any number of intervening messages in the same thread between Ross's and apfelmus's, and even though it is true that Ross's is immortalized in the database of the forum, the real question is how to fish for it. (ObRant again about how web forums are "friendly" to "programmers" who have no clue about trees.) Here is a feature unrelated to threading, and in fact it is much more fundamental and pervasive (since almost all email software, not just mine, provides this), and in fact it has much to do with pervasive computing too. Here in Canada, in cafes one can get wireless Internet access, but it has to be paid by the minutes. So suppose I want to visit haskell-cafe when I visit a Starbucks cafe, and there are 30 messages I want to read, and I can read them at the rate of half a minute per message. If haskell-cafe were on a web forum, I would end up keeping my wireless connection for 15 minutes. (Could I disconnect and reconnect between messages? First of all that's really a pain. Secondly, let's say I do that, the billing is still such that I connect for 2 seconds and I'm still charged for 1 minute.) But since haskell-cafe is on a mailing list, I can just connect for a few seconds to download the messages, then disconnect and read offline; I'm charged for at most 1 minute. (Alternatively but equivalently, suppose I want to read haskell-cafe while riding subway - underground trains. Here in Toronto the subway doesn't provide wireless Internet access yet, and probably won't for another decade. If haskell-cafe were on a web forum, I'd be SOL. But since it is on a mailing list, I can pull the same trick as above.) If you can find a single implementation of web forums that provides the features I have always taken for granted... but wait, there is one more feature, actually it's a meta-feature, a principle, a belief of mine. I am an individualist and non-conformist. I believe in each individual choosing and customizing his/her software and human-computer interface. There must be a common protocol for the exchange of content and meta-data, of course, but that should be where the commonality ends. How this content is presented to me, how I navigate and manipulate this content, what human-computer interface I use, and whose software I use, should be entirely up to me. The very idea of a web forum is an antithesis to this. In practice, because a web forum has to include some human-computer interface, and because someone has to write code for that, and because that "someone" is at best a few people with only a few hours per week for this, the web forum is forced to include little or no choice in its human-computer interface and its expressive power. And since you always have to worry about members who "have less technical knowledge such as professional programmers who don't understand trees and subsets", its expressive power cannot be too great, and so the mandatory web interface must suck. In other words, in theory you can always say "let's code up a superior, even diverse, web forum", but in practice who's going to do that? In contrast, superior software for mailing lists and newsgroups already exist, and if you have Mozilla or Thunderbird or Evolution or Outlook Express, you probably already have decent or half-decent such software. So let me make a daring counter-proposal. We go back to newsgroups. We set up our own news server for this; use just one server so everyone is always synced to it, and we do not pose a heavy load on it by today's standard anyway. Here are the benefits. First and foremost, all my favourite features are provided. (I'm allowed to be a bit self-centred for just a little while, ain't I? :) ) Everyone can use his/her favourite software. For example Mozilla can do newsgroups and news servers, and everyone knows how to use Mozilla, and if someone doesn't yet, the GUI is easy to pick up. It's a no-brainer. Normally news servers are set up to throw away old messages, but we can tell ours to keep everything forever. You can always consult the server for old messages for context or archaeology, in case you don't keep local copies of your own. Your favourite software can search. There may even be a way to get Google and other search engines to do the search. There is a valid concern raised about the ritual overhead on the part of a new member in subscribing and unsubscribing to a multitude of mailing lists, and it is suggested that a web forum mitigates this. A web forum mitigates this by requiring just one registration and you can subsequently post in any sub-forum. The newsgroup way answers this beautifully: it is the same number of registration and you can subsequently post in any newsgroup. Last but not least, for those who are excited about the very idea of a one-size-fits-all web interface: we can put up a web interface as a front-end to the news server too! As programmers, we are very individualist about programming environments: we like to pick our own editors, colours, key bindings... Actually we don't mind all sorts of compromises on those fronts, but here is one thing all of us insist: we require our code repository to be highly concurrent and distributed. We expect to work offline. In light of this, I find it rather odd that a programmer would propose the exact opposite for our discussion repository - a web forum, which not only superficially limits customizations (which, like I said, we can live with), but also fundamentally expects us to work online. More ironically, highly concurrent distributed discussion repositories have always existed, and they are now dismissed under the false impression of "hard to use". (They are certainly easier to use than highly concurrent distributed code repositories, since you post new messages rather than mutate existing messages, and so there is no need to learn about conflicts, merging, forking, and all that.) It makes me wonder: what's next, are we going to move our IDEs to the web too, and henceforth must we debug our code through the web?

9 10

what is a difference between existential quantification and polymorhic field?
by Bulat Ziganshin 21 Sep '06

21 Sep '06

Hello haskell-cafe, now i'm reading Haskell' proposals and found that these two things considered as different: http://hackage.haskell.org/trac/haskell-prime/wiki/ExistentialQuantification http://hackage.haskell.org/trac/haskell-prime/wiki/PolymorphicComponents can you please explain me what is the difference between data Ex = forall a. Num a => Ex a and data Po = Po (forall a. Num a => a) ? -- Best regards, Bulat mailto:Bulat.Ziganshin@gmail.com

4 5

Re[2]: [Haskell-cafe] [newbie] How to test this function?
by Bulat Ziganshin 21 Sep '06

21 Sep '06

Hello Bruno, Thursday, September 21, 2006, 6:16:49 PM, you wrote: >>> Second, I want to test this function, without hitting the filesystem. >>> In >>> C++ I would use a istringstream. I couldn't find a function that >>> returns >>> a Handle from a String. >> >> http://haskell.org/haskellwiki/Library/Streams > I'll give this a try. btw, although this may be considered as bug, it's possible to use Streams to combine getContents and other I/O on the same stream: h <- openFD "test" ReadMode str <- vGetContents h print (take 5 str) c <- vGetChar h print с print (take 5 str) :) so it's possible to use Streams to simplify your program. of course, you should understand how it works in order to use it in safe manner and that means that you should understand how unsafeInterleaveIO works. look at my today message in haskell list and expalanation of unsafePerformIO in "IO inside" -- Best regards, Bulat mailto:Bulat.Ziganshin@gmail.com

1 0

Re: Haskell web forum
by Aaron Denney 21 Sep '06

21 Sep '06

Moved to haskell-cafe from haskell-general. On 2006-09-20, Niklas Broberg <niklas.broberg(a)gmail.com> wrote: > On 9/20/06, Aaron Denney <wnoise(a)ofb.net> wrote: >> And I disagree with you. Web forums are usenet reinvented poorly. >> It's impossible to keep track of what's new, threading is either poor or >> nonexistent. Mailing lists with searchable archives work well. gmane >> provides a nice usenet interface to mailing lists. > > I don't recognize the forums I frequent in your description at all. I > have absolutely no problem keeping up with what's new. In fact I find > it a lot easier since many of the topics take place in subforums that > I know don't interest me, so I don't even need to go there to check. And half of the time messages are miscategorized, because conversations are miscategorized, so I do need to go check. > With this mailing list, I have to manually "mark as read" about 2/3 of > all incoming mails because they don't (from the title) interest me, > and I have to read a few that don't interest me because I couldn't > tell from the title. Where as an entire page of messages under a particular "thread" (it's not a thread, because they almost never have the messages showing the exact post they're responding to, just the general list of messages in this topic.) Then I need to go locate the tiny link that gets me to the next page. They're nearly unuseable. > Threading also works as well as could be expected, better than > threading in gmail for instance, so I don't see the problem here > either. That has never been my experience. Threads branch annd mutate and drift. And this is a good thing. The rigid categorizations can't handle this, whereas I can easily ignore entire threads and subthreads with a client designed for it. >> > It's all there, all the time. >> >> Yes, that's part of the problem. > > How is that? You mean because all the old stuff gets in the way of the > new? Then you're just using a bad forum software that can't properly > point out the new stuff for you. I agree that not all forums are good, > but there are definitely those that are. Exactly. but the problem with forums is that everynone must use the same forum software because it's sitting on a website. With usenet or mailing lists, everyone gets their own software that provides the features they want, with the interface they want. >> > It is also easy to create sub-groups/forums for specific projects, >> >> This is the one semi-useful thing. Of course, what you end up with is >> another not-so-useful forum. > > This is probably the one most useful thing yes. I wouldn't call it > semi-useful though. Just remember all the responses that the HCAR gets > each year of the form "wow how many cool projects are out there that I > had no clue existed". What if all (or many of) those projects were > actually there, on the forums, where everyone knew where to look? Forums are another place to look, instead of on the wiki, on the mailing lists, etc. It fragments the community. > And your last sentence, I just don't understand. "not-so-useful" just > because it's a forum, or did you mean somethine else as well? Yes, because it's a forum. Almost any feature can be implemented in a web-forum. And many will, but almost none will be done well, because (a) it's hard to get something that will please everyone and it has to please everyone because everyone must use the same software (b) The experience isn't there. Mail & usenet software has been worked on for over 25 years. It works. Your mail client sucks? Get a new one. There are lots, and one should surely work for you. I don't have the option of using different web-forum software when something about it annoys the hell out of me. The biggest thing I see for them is web accessibility. But we already have that. We have archives. We have gmane. Tons of other gatewaying software exist, and could be easily set up, because mailing lists are designed to be used by multiple clients. Web forums are only "pull". I have to go check each individual one, on a semi-regular basis. Yes, there is RSS. It's another hacky workaround for something that mail inherently has. Mailiing lists can be either pull or push. I can dump the list mail in my inbox and get it regularly, or segregate it out to another mailbox, or several mailboxes, or whatever. -- Aaron Denney -><-

1 0

Either e Monad
by Deokhwan Kim 20 Sep '06

20 Sep '06

Dear all, Where is the Monad instance declaration of Either e? >From the description of Control.Monad.Error, I deduce that Either e is an instance of Monad. http://haskell.org/ghc/docs/latest/html/libraries/mtl/Control-Monad-Error.h… class Monad m => MonadError e m | m -> e where ... Error e => MonadError e (Either e) But, I cannot find the Monad instance declaration of Either anywhere. Thanks. -- Deokhwan Kim

3 3

MonadList?
by Michael Shulman 20 Sep '06

20 Sep '06

In another thread, I wrote: > main = (>> return ()) $ runListT $ do > arg <- ListT getArgs > n <- ListT $ return [1..3] > liftIO $ putStrLn ((show n) ++ ") " ++ arg) The frequent occurence of "ListT $ return" in my code when I use the ListT monad transformer has made me wonder why there isn't a standard typeclass `MonadList', like those for the other monad transformers, encapsulating the essence of being a "list-like" monad -- in this case, the ability to select from a list of things. I quickly wrote one for myself: class MonadList m where option :: [a] -> m a instance MonadList [] where option = id instance (Monad m) => MonadList (ListT m) where option = ListT . return Has anyone else thought about or done something like this? Mike

4 6

Re: Numeric type classes
by Henning Thielemann 20 Sep '06

20 Sep '06

On Mon, 11 Sep 2006, Ross Paterson wrote: > On Mon, Sep 11, 2006 at 04:26:30PM +0200, Henning Thielemann wrote: > > On Sat, 9 Sep 2006, Ross Paterson wrote: > > > I think that a finer grain numeric hierarchy, while retaining Num, etc, > > > is feasible without changing the language: unlike the case of monads, > > > the people who will be defining instances of numeric classes are the > > > very ones who are inconvenienced by the current hierarchy. The main > > > impact on clients of the classes is that some functions would have > > > more general types. > > > > There are many Num instances around in libraries where people wrap to > > external libraries: functionalMetapost, CSound wrapper in Haskore, > > SuperCollider (GSL too?). What about Num (algebraically Ring) instances of > > polynomials, residue classes and other such advanced mathematical objects? > > And what do abs and signum mean for Haskore's orchestra expressions, > polynomials, residue classes, vectors, matrices, functions, etc? For clarification: Haskore does not define any arithmetic for music, but CSound provides some arithmetic and Haskell wraps it with Num instances. > The people who define those wish they were defining Ring, but they must > define Num. It seems we are at a point, where we have to define what is a 'number'. More precisely: Can you tell me the difference between numbers and "more complex mathematical objects"? Is a complex number a number? Is a quaternion a number? Is a residue class a number? We can calculate with integers modulo some other integer like with integers - is that considered computation with numbers? Shall we distinguish between matrices of numbers and matrices of more complex mathematical objects? In signal theory matrices of polynomials are very common.

14 30