- Haskell-Cafe - Haskell.org

Haskell modules: Morton's fork?
by Nikita Churaev 22 Jul '17

22 Jul '17

Option 1: module App.Component (Foo, Bar, Qux) -- re-exports module App.Component.Foo (Foo, make, func1, func2) module App.Component.Bar (Bar, make, func1, func2) module App.Component.Qux (Qux, make, func1, func2) API clients will have to write long import lists: import App.Component import qualified App.Component.Foo as Foo import qualified App.Component.Bar as Bar import qualified App.Component.Qux as Qux Option 2: module App.Component (module Foo, module Bar) -- re-exports module App.Component.Foo (Foo, makeFoo, fooFunc1, fooFunc2) module App.Component.Bar (Bar, makeBar, barFunc1, barFunc2) module App.Component.Qux (Qux, makeQux, quxFunc1, quxFunc2) Now you can import easily: import App.Component But now you have to prefix your functions, just like in C... Is there another, cleaner option?

2 1

Re: [Haskell-cafe] Feeding a monad into itself
by Olaf Klinke 22 Jul '17

22 Jul '17

Joshua Grosso wrote: >Could MonadPlus (with mzero) or Alternative (with empty) provide the >termination condition, if this pattern turned out to be more generally >useful? MonadPlus and Alternative provide a way to put the zero into the monad, but no way to check for zeroness. In the Maybe type you have the isNothing function, but the MonadPlus and Alternative classes don't require such a thing, and probably shouldn't. In the following I present a class that allows to do stuff conditional on mzero. The starting observation is that in logic, False => anything == True. Hence the implication arrow is a (binary) function that turns nothing into something. Consider the following binary operation. (Logicians: think conjunction) (/\) :: Monad m => m a -> (a -> m ()) -> m a ma /\ f = (\a -> liftM (const a) (f a)) =<< ma It's a bit like Control.Monad.mfilter but deletes stuff by mapping it to mzero instead of False. For example, ghci> [0,1,2,3] /\ (\n -> replicate n ()) [1,2,2,3,3,3] Some MonadPlus monads are set-like. Think of Nothing as the empty set, Just x as the singleton set {x} and think of the list [x,y,z] as the finite set {x,y,z}, disregarding muliplicities. For such monads m, the types m a have a preorder ≤ which is set inclusion under the "set interpretation". For example, Nothing ≤ Just () [1,2,3] ≤ [1,2,3,4] but not [1,2] ≤ [1,4]. Now consider the following type class. class (Monad m) => Heyting m where (-->) :: (Eq a) => m a -> m a -> a -> m () We reqire that for all x, y and f x /\ f ≤ y if and only if f ≤ x --> y That is, (x /\) is left adjoint to (x -->), which reminds the logician of the Heyting implication arrow. For MonadPlus instances, we suggest that (mzero -->) == const (const (return ())) Here are some instances: instance Heyting [] where ys --> xs = \a -> if (a `notElem` ys) || (a `elem` xs) then [()] else [] instance Heyting Maybe where y --> x = listToMaybe.((maybeToList y) --> (maybeToList x)) Without the Eq constraint, I couldn't have used the elem function, hence the Heyting class carries that constraint. (There is also a Heyting instance for the monad of finite distributions [1], see the package probable [2] and its siblings.) Heyting monads have a sort of negation, turning mzero into something: neg :: (MonadPlus m, Heyting m) => m a -> m () neg x = ((liftM (const ()) x) --> mzero) () Using neg, we can do stuff on mzero: ghci> neg Nothing Just () ghci> neg (Just ()) Nothing ghci> neg [1,2,3] [] ghci> neg [] [()] Finally we implement Jake's function, but more general. untilMZero :: (Heyting m, MonadPlus m) => (a -> m a) -> a -> m a untilMZero f x = (liftM (const x) (neg (f x))) `mplus` ((f x) >>= (untilMZero f)) ghci> untilMZero (\n -> if n > 5 then Nothing else Just (n+1)) 0 Just 6 ghci> untilMZero (\n -> if n > 5 then [] else [n+1,2*n]) 1 [6,10,8,6,6,10,8,6,10,8,6,6,10,8] -- Olaf [1] For the Giry monad, --> computes the Radon-Nikodym derivative. It type-checks. [2] https://hackage.haskell.org/package/probable-0.1.2/docs/Math-Probable-Distr…

2 1

Feeding a monad into itself
by Jake 21 Jul '17

21 Jul '17

I have a function f :: a -> Maybe a that I want to feed into itself with some initial input until it returns Nothing. untilNothing :: (a -> Maybe a) -> a -> a untilNothing f x = case f x of Just x' -> untilNothing f x' Nothing -> x Is there a better way to do this? I feel like there is something more general going on with Monads being fed into themselves, but maybe I'm wrong. Thoughts?

8 12

How is this variable not in scope?
by Jeffrey Brown 21 Jul '17

21 Jul '17

The following compiles and runs perfectly: main = do (inputFile:_) <- getArgs input <- readFile inputFile let pairs = pairFilesToContents $ readSmsnLines input mapM_ f pairs where f :: (FilePath, String) -> IO () f (name, content) = writeFile name content But the following triggers the error "Not in scope: ‘root’". main' = do (root : inputFile :_) <- getArgs input <- readFile inputFile let pairs = pairFilesToContents $ readSmsnLines input mapM_ f pairs where f :: (FilePath, String) -> IO () f (name, content) = writeFile (root+name) content -- Jeff Brown | Jeffrey Benjamin Brown Website <https://msu.edu/~brown202/> | Facebook <https://www.facebook.com/mejeff.younotjeff> | LinkedIn <https://www.linkedin.com/in/jeffreybenjaminbrown>(spammy, so I often miss messages here) | Github <https://github.com/jeffreybenjaminbrown>

3 3

How to deal with last item with concatMapAccumC in Conduit.
by jun zhang 21 Jul '17

21 Jul '17

Dear cafes I use Conduit to parse a huge file. And I need merge lines by condition. I find the concatMapAccumC can do that and I write a demo as blow(with conduit-combinators-1.0.6,lts-6.18). The problem is if the last item didn’t make condition true, the data only keep in the accum but missing in stream. Any one can give me some advises? Thanks ---------------------------- import Conduit test'::Int->Int->(Int,[Int]) test' a s = case (a+s) > 5 of True -> (0,[a+s]) False -> (a+s,[]) testlog::IO [Int] testlog = runConduit $ (yieldMany [1,2,3,4,5,6,3]) $= (concatMapAccumC test' 0 ) $$ sinkList

2 4

getChar (NoBuffering) not working on Windows
by Francesco Ariis 20 Jul '17

20 Jul '17

Hello list, I am writing a small console application. Today I was bitten by #2189 [1] ("hSetBuffering stdin NoBuffering doesn't work on Windows"). To reproduce: import System.IO main :: IO () main = hSetBuffering stdin NoBuffering >> getChar >>= putChar -- on linux you just need to press, say, 'c', on Win you -- need to press <Enter> too. Is there any sensible (cross-platform) way of working around this? [1] https://ghc.haskell.org/trac/ghc/ticket/2189

1 1

runghc and linking in libraries
by Brian Sammon 20 Jul '17

20 Jul '17

If I want to have (and run without explicitly invoking ghc or runghc) a script that starts as follows: #!/usr/bin/runghc module Main where import MyLib I can do it just fine if MyLib is a pure-haskell MyLib.hs -- it doesn't have to be packaged or installed or anything complicated. If I want to create a MyLib that uses C code (that I wrote) through FFI, it doesn't seem to be possible to use it in a runghc script like this, unless I create a package and install it. Am I correct about this? In the course of researching this, I realized that I have a rather poor understanding of what "packages" are in ghc-land. I have a high-level-overview understanding ("It's like rubygems, but for haskell"), but I'm rather unclear about the nuts-and-bolts of it. The tight coupling between packages and the ghc core is rather different than what I've seen in other programming languages I've used (I haven't programmed in Java though) It would be useful if ghc had an argument to take a "MyLib.lib" file (or maybe a directory) and temporarily act like it's an installed package for the one invocation. Unsurprisingly, web-searching for "haskell" and "package" doesn't give me the focused result set that I want. I found the relevant section of the GHC Users Guide https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/packages.ht… but links to other reading materials would be appreciated.

2 1

[ANN] New release of SBV (v7.0), now with incremental solving
by Levent Erkok 20 Jul '17

20 Jul '17

I'm pleased to announce v7.0 release of SBV, a library for integrating SMT solvers into Haskell. This release adds incremental solving capabilities to SBV: Users can now interact with the underlying solver using a typed API capturing the SMTLib language, issuing commands, getting model values, and asserting further constraints on-the-fly. The possibility of issuing multiple check-sat commands allows incremental assertion of facts, a key feature of modern constraint solving systems. Hackage: https://hackage.haskell.org/package/sbv Homepage: http://leventerkok.github.io/sbv/ Release notes: https://github.com/LeventErkok/sbv/blob/master/CHANGES.md SMTLib initiative: http://smtlib.cs.uiowa.edu/ Feedback and bug reports are most welcome! -Levent.

1 0

How to statically link some external libraries?
by Kai Zhang 20 Jul '17

20 Jul '17

I want to produce a statically-linked executable. First I tried "-optl-static -optc-static -optl-pthrea", which statically links all external libraries. But when I ran this executable it gives me "abort (core dumped)" error. I suspect the statically-linked glibc library might be the cause. So I decide to statically link all other libraries and leave the glibc as a dynamic library. How to do this using Cabal? Should I write a custom "Setup.hs"?

3 2

DeepSeq and random generators
by Dennis Raddle 19 Jul '17

19 Jul '17

I'm having trouble using Control.DeepSeq.force (necessary for parallel code) on a StdGen from System.Random. I don't know how to derive NFData. A standalone deriving statement, like this deriving instance NFData StdGen gets me the error "the data constructors of StdGen are not all in scope..." Is it actually possible to derive an NFData instance of StdGen, and how would I do that? Or if not possible, how can I work about this? Note: I'm going to see if I can use tf-random in this application. Maybe it will be easier. D

3 2