July 2014 - Beginners

Data.Stream interleave implementation question
by Bryan Brady 18 Aug '14

18 Aug '14

I'm in the process of learning Haskell and am implementing my own Stream type. ( I'm working my way through homework problems I found at http://www.seas.upenn.edu/~cis194/hw/06-laziness.pdf ) One of the questions is: "Deﬁne the stream ruler :: Stream Integer which corresponds to the ruler function 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, 4, . . . where the nth element in the stream (assuming the ﬁrst element corresponds to n = 1) is the largest power of 2 which evenly divides n. " I got pretty close, but ended up looking at Data.Stream to nail down my mistake. My original solution was: ----- data Stream a = Cons a (Stream a) deriving (Eq, Ord) streamRepeat :: a -> Stream a streamRepeat a = Cons a (streamRepeat a) interleaveStreams :: Stream a -> Stream a -> Stream a interleaveStreams (Cons a as) (Cons b bs) = Cons a (Cons b (interleaveStreams as bs)) ruler :: Stream Integer ruler = foldr1 interleaveStreams (map streamRepeat [0..]) ----- Attempting to evaluate 'ruler' in ghci does not finish. I knew something was wrong with my implementation of interleaveStreams because it worked when I expanded it out manually for a few steps. The way Data.Stream implements interleave is: interleaveStreams (Cons a as) bs = Cons a (interleaveStreams bs as) I don't understand why this works and why my original implementation doesn't. I figure if the latter works, the former should as well. Here is my reasoning: In the latter definition, Cons a (interleaveStreams bs as), (interleaveStreams bs as) is a thunk. The thunk should only be evaluated when it is needed. In my original definition, (interleaveStreams as bs) is a thunk. The difference is an extra Cons (e.g., Cons b). It seems like the additional Cons is causing problems, I just don't understand why. Can anyone point out what I'm missing? thanks, bb

7 10

combining a trifecta parser with FreshMT from unbound
by Andreas Reuleaux 30 Jul '14

30 Jul '14

I can relatively easily combine a trifecta parser with some state: -- imports, language pragmas omitted here newtype InnerParser a = InnerParser { runInnerParser :: Parser a } deriving (Functor , Monad , Applicative , Alternative , Parsing , CharParsing , MonadPlus ) data PiState = PiState { foobar :: Integer -- ...some more stuff... } type PiParser = StateT PiState InnerParser instance TokenParsing PiParser where someSpace = buildSomeSpaceParser (skipSome (satisfy isSpace)) $ commentStart .~ "{-" $ commentEnd .~ "-}" $ commentLine .~ "--" $ commentNesting .~ True $ emptyCommentStyle idStyle = styleStart .~ letter $ styleLetter .~ (alphaNum <|> oneOf "_'") $ styleReserved .~ HS.fromList ["refl" ,"ind" ,"Type" -- ... ] $ emptyIdents identifier :: PiParser String identifier = token $ ident $ idStyle -- etc. This is the approach taken in Idris e.g. and it works, as I understand, because the trifecta/parsers type classes involved (CharParsing etc) are "already prepared" for attaching a StateT monad transformer: there are instances defined for all the usual transformers: StateT, ReaderT, RWST etc. I can easily add more convenient type classes, having them derived for my InnerParser: LookAheadParsing, DeltaParsing etc. Now when I want to use FreshMT from the unbound package instead of StateT, things get much more complicated. I don't really care at this point if with a type synonym type PiParser = FreshMT InnerParser or as a newtype wrapper (deriving as much as I can automatically) newtype PiParser a = P { runP :: FreshMT InnerParser a } deriving (Monad, Functor, Applicative) Anyway, I need to provide all these instances by hand (CharParsing, Alternative etc), am still struggling with the details (and might have more questions in this regard), even though FreshMT really is just defined in terms of StateT http://hackage.haskell.org/package/unbound-0.4.3.1/docs/src/Unbound-Locally… So I wonder, if there is an easier way to approach this, have some more instances derived automatically ? (Basically I just want to get Stephanie Weirichs pi-forall language working with trifecta instead of parsec, cf. this years' OPLSS) I can provide complete running code (with all the imports etc) if necessary. Thanks. -Andreas

1 1

Question on Lazy IO
by 양철웅 24 Jul '14

24 Jul '14

Hi, all. Haskell's lazy IO causes below program to write back "ECHO result" to the user even if he doesn't give any input to the program. test1 :: IO () test1 = do l <- fmap lines getContents putStrLn "ECHO result" putStr $ unlines l If the client logic, that is the part for feeding data to this program and wait for the response, expect the server logic to respond only after all the client's request data are feed to the server logic, the behaviour could be problematic. I'm now writing simple web server, as you might guess, and my web server responds "HTTP/1.1" before any client's request are given. I'm using hGetContents and web server's logic is basically same with above code skeleton. How can I tell the program to respond data only after all the requests are feed? Any directions are welcomed. Thanks. Chul-Woong

5 6

Type error in sub-function
by 양철웅 24 Jul '14

24 Jul '14

Hi, all! When I compile my program below, ghc emits Type error. Since the return type of foo Func is IO String and first case statement has "return timedVal", I think that ghc expects the type of timedVal as String. However, the error message shows that ghc expects timedVal should have type IO b0. What am I missing? Any comments are appreciated. Thanks in advance! Chul-Woong -- import Data.Time.LocalTime (getZonedTime, zonedTimeToUTC) import System.Locale (defaultTimeLocale) import Data.Time.Format (formatTime) import Data.Time.Clock (UTCTime) currentUTCTime :: IO UTCTime currentUTCTime = do zTime <- getZonedTime return (zonedTimeToUTC zTime) fooFunc :: IO String fooFunc = do putStrLn "Hello, world" barFunc "xutc" where barFunc s = case s of "apple" -> return s ('x': val) -> return timedVal where timedVal = case val of "utc" -> do utcTime <- currentUTCTime formatTime defaultTimeLocale "%a" utcTime --- Prelude> :l test [1 of 1] Compiling Main ( test.hs, interpreted ) test.hs:21:57: Couldn't match expected type `IO b0' with actual type `String' In the return type of a call of `formatTime' In a stmt of a 'do' block: formatTime defaultTimeLocale "%a" utcTime In the expression: do { utcTime <- currentUTCTime; formatTime defaultTimeLocale "%a" utcTime }

6 8

Re: [Haskell-beginners] Haskell in the Digital Humanities
by "René＠gmail" 23 Jul '14

23 Jul '14

Dear Niels-Oliver, having some relation to DH and working in my student's job with (dirty-tagsoup) XML. Struck by Haskells beauty and clearness of coding but also aware of the rather deep and technical type abstractions and combinatorics which are still hard to grasp for me, I also felt unsure about following the Haskell road. Still clinging on to Haskell because: a) bespoken clearness of code of Haskell I miss too hard when looking on other languages code b) elegance of Parser-Combinators (well this seems not really an argument because they are ported to many other languages, but I also suspect something might be lost, because you do not get the full customizability of functions and operators like in Haskell in other langs) c) stemming from the same Parser-Combinator vein: PANDOC: is there a better text conversion tool, which is so well maintained and so easy (well yes, some knowledge needed -> see learning tipps) to extend d) I don't know much about robust code in other languages, but then again I had so much input by Haskell blogs and tutorials, that the idea that type safety and testing functions in isolation might be a good one has stuck. So although you probably will not interface with a nuclear power plant in your DH coding, I think robustness could be an argument too. So the elegance of code and the safety of the program are very general pro-arguments for Haskell. Text handling capabilities through Parser-Combinators are really handy. Still for text-munging to produce a desired output for D3.js all this might seem overkill and to be honest I wouldn't want to parse tagsoup with Haskell either (but only because I know the benefits of a specialized text-processing-toolbox like TUSTEP [http://www.tustep.uni-tuebingen.de/tustep_eng.html] On the other hand, if you have a well formed XML there a enough XML-tools for haskell which produce a neat Haskell data structure which in turn you can work at with nice pure functions. Also there are enough possibilities to get these structures out into Javascript (Fay,...). (javascript? I also find elm-lang.org very compelling.) So for now my solution is to have Haskell as an important side-project. I still have to go through the "Write Yourself A Scheme" [1] which is said to give the needed exposure to some abstract types so that one can digest the theory behind those much better, but I hope by then the scariness of the abunding type system will abate a bit more. There's also a new book about data anlysis, which advertises "Recipes for every stage of data analysis, from collection to visualization" [2]. Interfacing with C? At least from my superficial obervance of my cabal installs Foreign.PointerC (or so) is by no means an exception. Don't know about Python, though. Well, anyway good to know, someone has similar worries like me. Good luck! Cheers, René Tobner [1] http://en.wikibooks.org/wiki/Write_Yourself_a_Scheme_in_48_Hours [2] http://www.packtpub.com/haskell-data-analysis-cookbook/book On 22.07.2014 14:00, beginners-request(a)haskell.org wrote: > The question is simple and was raised by many people before, probably everyone who began with another language and at some point got to know Haskell: Should I learn Haskell?

1 0

Manage the type of a variable in the show implementation
by Christian Sperandio 23 Jul '14

23 Jul '14

Hi, I wrote this kind of code: data DataLog a = SingleData a | MultipleData [a] instance (Show a) => Show (DataLog a) where show (SingleData v) = show v show (MultipleData vs) = join "|" vs where join _ [] = [] join _ (x:[]) = show x join s (x:xs) = show x ++ s ++ join s xs It works but when the DataLog is String typed I got this result: "My Test"|"Another Value" I'd prefer have: My Test|Another Value I know the quote problem comes from the show function on String value. How could I do a show for no-string values and return directly the value for strings? Thanks. Chris

3 3

Haskell in the Digital Humanities
by Niels-Oliver Walkowski 22 Jul '14

22 Jul '14

Dear Haskell Community, I am coming to you with a question which I try to solve for 2 month now, but still struggling, so I tried the step to refer to you. The question is simple and was raised by many people before, probably everyone who began with another language and at some point got to know Haskell: Should I learn Haskell? So I read a lot of things in the www that had the same question or that could support decision making. The reason why I couldn't manage to make a decision rest on – as so often – in the particular case or at least what seems particular for me. I would like to briefly outline why I would love to learn Haskell, what are the reasons why I am hesitating and how all this relates to my background. I would be tremendously grateful for feedback that will make a decision easier for me and hope that it is ok to ask such a question on a beginner's mailinglist. Background: I am a humanist, working in the so called Digital Humanities realm. I don't have any information technology education though I started coding in PHP at the beginning of University. In DH like in many other areas of 'data driven science' Python is 'Lengua Franca'. That might relate to the fact that coding is means not goal. In fact, often coding is more glueing things together than really code something and the ecosystem of scripts and libraries is as broad as scientific discourse itself. It is easy to reach a goal a goal (which is important when you want to test hypothesis) but things are often dirty of cause. Interest in Haskell: I got to know ideas of functional programing half a year ago when I digged a little bit into machine theory. Functional Programing expresses exactly how I think about computation and furthermore I enjoy very much to think within a functional logic about things I want to do. Additionally I think that the functional approach has some important advantages for coding in science at least on the publication level because it expresses ideas in a clear defined way that is not the same with an imperative approach. I started to read a little bit about functional programing in Python since Python is multi-paradigm but very quickly my impression was that it is something you do not really want to do, it makes writing code more verbose, you do not really have the benefits of functional programing and often feels like workaround, so I was brought back to Haskell again Concerns: As mentioned before coding in Digital Humanities is definitely always means for an end. On the other hand I like to to my things straight forward and on a certain abstraction level (maybe because I was trained in Philosophy ;-) ) General concerns rest on the fear that the Haskell ecosystem is not big enough for the area in which I code. I work in the area of text analysis (not identical with computational linguistics), topic modeling, I need sophisticated visualisations, I work with XML (state of the art in DH) and a lot with so called 'dirty data' which relates to the whole NoSQL world – especially Graph and JSON based. Managing Web technologies is important because of the importance of publishing. Approaches like the IPython Notebook are significant (I got to know that there is IHaskell which is IPython with a Haskell kernel but I don't know how valid and stable this project is). I absolutely do not need to have a library for any tiny task and I like to approach things from a higher level of abstraction but of cause there is a point where implementing something on your own is not reasonably time efficient anymore, especially when the goal is something completely different. And being not natively educated in informatics there is also a limit in capacity at a certain point. If you could provide me with feedback or with your judgment to help me evaluate the Haskell environment around my area and social field in a more substantial way and to get a more sophisticated feeling for or against the idea that learning Haskell will work out in my case, it would be wonderful. (A blog article about using Haskell in DH that summarizes the discussion and promotes the awareness of Haskell in the DH community would also something I might consider afterwards) If you feel I misused the list, please apologize. best, Niels-Oliver Walkowski

3 2

Critique My First Project - CampBX API Client
by Pavan Rikhi 19 Jul '14

19 Jul '14

Hiya, so I've spent the last 2 months or so learning Haskell. I've finished the UPenn course, LYAH and a little bit of RWH. I wanted to get a project going so I tried making an API client for the CampBX Bitcoin Market[1]. I started out by following along with the FPComplete Mailchimp API Tutorial[2]. I'm asking you beautiful people for any feedback on it so I can be sure that I'm on the right track here, stuff like code style/idioms, project layout, public interface, implementation, docs, ideas for new features, anything you want to tell me, etc.: https://github.com/prikhi/campbx-haskell Typical usage would be something like this[3]: main :: IO () main = do cfg <- defaultCampBXConfig _ <- runCampBX cfg $ do totalAskVolume <- calculateAskVolume <$> getDepth liftIO . putStrLn $ "Total Ask Volume: " ++ show totalAskVolume return () calculateAskVolume :: Depth -> BTCAmount calculateAskVolume depthList = sum . map askPrice . asks $ depthList where askPrice (Ask (_, q)) = q There are still some things I want to work on: * Define Asks and Bids[4] using Record Syntax. The JSON[5] for a Bid/Ask comes back as a 2 item Array but the generically derived instance[6] expects an object. I haven't completely wrapped my head around Aeson's Array parsing... * Write tests. I'm imagining they would be more "given this JSON, make sure the data structure is created correctly" instead of property-based testing. * Use something other than Doubles to represent Amounts + Prices. Is there a standard library for accurate math with decimals(I need up to 8 decimal places)? I suppose I could always just use Integers to represent Satoshis(the smallest subunit of bitcoins). Also I tried adding "default-extensions: OverloadedStrings" to the .cabal file and removing the pragmas from the source files, but the package wouldn't build :( [1] https://campbx.com/api.php [2] https://www.fpcomplete.com/school/to-infinity-and-beyond/competition-winner… [3] https://github.com/prikhi/campbx-haskell/blob/master/bin/CampBXMain.hs [4] https://github.com/prikhi/campbx-haskell/blob/master/src/Web/CampBX/Types.h… [5] http://sleepanarchy.com/p/K1wuSm [6] https://github.com/prikhi/campbx-haskell/blob/master/src/Web/CampBX/Types.h…

1 0

More simple continuation questions
by martin 17 Jul '14

17 Jul '14

Hello all, (1) when reading about continuations, there is this thing which needs an "other function" to be passed to produce a final result. And there is this "other function". Which of the two is "the continuation"? http://stackoverflow.com/questions/9050725/call-cc-implementation sais " ... instead, they are passed a function that represents the 'next step' in the computation - the 'continuation' " Which sound like the "other function" is called "continuation". But other articles hint in the opposite direction. (2) When comparing continuations with callbacks, it struck me that the type is newtype Cont r a = Cont { runCont :: (a -> r) -> r } So the type of the final result is r. But why does the function (a->r) need to return an r? With regular callbacks this does not seem to be the case. After returning from the callback the surrounding function is free to do anything it wants with the return value and return a value of a different type. Why is that so?

2 3

cabal sandbox add-source + cabal repl (for tracking Idris git version)
by Andreas Reuleaux 14 Jul '14

14 Jul '14

Being somehow new to cabal sandboxes, I would expect that sources that I add with cabal sandbox add-source /some/path are available to me in my local (sandbox aware) cabal repl, this is however not the case, or at least I haven't figured out how to make this this work. I am trying to use Idris as a library - roughly as described in http://brianmckenna.org/blog/idris_library, (and in particular the "cabal sandbox add-source..." step, described there) but my question is really more Haskell and cabal sandbox than Idris related, therefore I am asking here. I have successfully installed Idris from git in a sandbox (For completeness's sake here are the individual steps, including some test steps with highlight-versions, that are not strictly necessary). I install only the idris deps via cabal (i.e. from hackage), idris itself via make from its git sources: -------------------- (within my dir ~/idr, i.e. creating ~/idr/dev here) git clone git://github.com/idris-lang/Idris-dev dev cd dev cabal sandbox init cabal update cabal install highlight-versions (activate my sandbox with some bash magic, i.e. include /home/reuleaux/idr/dev/.cabal-sandbox/bin in my PATH, so that the command highlight-versions is found:) cabal install --dry-run --only-dependencies idris | highlight-versions (now the wet run:) cabal install --only-dependencies idris make (which installs idris from its git sources) I have two more wrappers in my sandbox for that purpose: (really some bash magic...) * ghc: exec "/usr/bin/ghc" -no-user-package-db -package-db=/home/reuleaux/idr/dev/.cabal-sandbox/x86_64-linux-ghc-7.6.3-packages.conf.d "$@" * ghc-pkg: exec "/usr/bin/ghc-pkg" --no-user-package-db --package-db=/home/reuleaux/idr/dev/.cabal-sandbox/x86_64-linux-ghc-7.6.3-packages.conf.d "$@" I know that "cabal sandbox hc-pkg" is the regular wrapper for ghc-pkg, but idris make expects a ghc-pkg command (not cabal sandbox hc-pkg) Now within ~/dev I can e.g. call * "idris" or * cabal repl and then > :m +Idris.Parser i.e. load some idris modules on the repl -------------------- So far, so good. Now I would like to track that dev version of idris in a separate directory / sandbox, I am therefore using "cabal sandbox add-source...": mkdir ~/foo cd foo cabal sandbox init cabal sandbox add-source ~/idr/dev cabal update (I don't mind reinstalling the deps, not sure if this is really necessary:) cabal install --only-dependencies idris I would think that within foo, I should be able to just do e.g. cabal repl > :m +Idris.Parser (as cabal repl is sandbox aware, as I learned) but nope !? Prelude > :m +Idris.Parser <no location info>: Could not find module `Idris.Parser' It is not a module in the current program, or in any known package. Prelude > Well I can see the sources with [sb] reuleaux@mirabelle ~/foo $ cabal sandbox list-sources Source dependencies registered in the current sandbox ('/home/reuleaux/foo/.cabal-sandbox'): /home/reuleaux/idr/dev To unregister source dependencies, use the 'sandbox delete-source' command. [sb] reuleaux@mirabelle ~/foo $ But that's about it. Similarly: How would I use these added sources with ghc? Do I have to use both: * /home/reuleaux/idr/dev/.cabal-sandbox/x86_64-linux-ghc-7.6.3-packages.conf.d * and /home/reuleaux/idr/foo/.cabal-sandbox/x86_64-linux-ghc-7.6.3-packages.conf.d in my -package-db ? I am happy to give more details of my installation, if necessary (I am on debian jessie, my cabal is v 1.20) Thanks in advance. -Andreas

2 4