November 2008 - Reactive

Link to Tetris
by Creighton Hogg 16 Dec '08

16 Dec '08

http://code.haskell.org/~wchogg/Tetris.hs Conal told me how to host the files on code.haskell.org, so now the following link will work as well.

4 7

windowPoint function in Reactive/GLUT/SimpleGL.hs
by Peter Verswyvelen 01 Dec '08

01 Dec '08

The windowPoint function in Reactive/GLUT/SimpleGL.hs has the following comment: -- | Convert a window position to a logical X,Y. Logical zero is at the -- origin, and [size??]. -- -- TODO: this def is completely broken. What do I even want? Probably to -- go backward through the viewing transform to the XY plane. The function I made in the past for doing this is: windowPoint :: GL.Position -> IO (Double,Double)windowPoint pos(a)(GL.Position x y) = do viewport(a)(_,GL.Size _ height) <- get GL.viewport modelviewMatrix <- getMatrix $ GL.Modelview 0 projectionMatrix <- getMatrix $ GL.Projection let pos3 = GL.Vertex3 (fromIntegral x) (fromIntegral (height-y)) 0 GL.Vertex3 vx vy _ <- GLU.unProject pos3 modelviewMatrix projectionMatrix viewport return (vx,vy) where getMatrix :: GL.MatrixMode -> IO(GL.GLmatrix Scalar) getMatrix mode = get $ GL.matrix $ Just mode

3 2

Is the code in makeTVal free of race conditions?
by Peter Verswyvelen 29 Nov '08

29 Nov '08

I'm looking at the makeTVal code: data TVal t a = TVal { timeVal :: (t,a), definedAt :: t -> Bool } makeTVal :: Clock TimeT -> MkFed (TVal TimeT a) amakeTVal (Clock getT _) = f <$> newEmptyIVar where f v = (TVal (readIVar v) (unsafePerformIO . undefAt), sink) where undefAt t = -- Read v after time t. If it's undefined, then it wasn't defined -- at t. If it is defined, then see whether it was defined before t. do -- ser $ putStrLn $ "sleepPast " ++ show t sleepPast getT t-- maybe True ((> t) . fst) <$> tryReadIVar v value <- tryReadIVar v case value of -- We're past t, if it's not defined now, it wasn't at t. Nothing -> return False -- If it became defined before t, then it's defined now. Just (t',_) -> return (t' < t) sink a = do t <- getT writeIVar v (t,a) I assume the sink function is called from a different thread than the one on which undefAt is called. Suppose a thread switch happens just before the (t,a) value is written to the IVar in sink a = do t <- getT writeIVar v (t,a) Lets call the t above t1. Now sleepPast getT t is evaluated on the other thread. sleepPast getT t Assume the parameter t passed to sleepPast is really close to t1 but larger than t1. Call it t2. So t1 < t2. Call t3 the time as evaluated by getT in the sleepPast function. So t1 < t2 < t3 (since t3 is read a after t1, and since the clock is serialized, this is the case I guess) Since t3 > t2, sleepPast will not call threadDelay, so it is not unlikely that a thread switch happens. But then value <- tryReadIVar v evaluates to Nothing, since the IVar is not written to yet. So undefAt will return False, although clearly t1 < t2 and so it should have returned True. I'm not sure if my reasoning is correct. But if it is, it would be solvable my making sure that getting the time from clock and writing it to the IVar is an atomic operation. Does any of the above make sense?

1 0

reactive adapter for OpenAL
by Richard Smith 28 Nov '08

28 Nov '08

Hello list, I'm working on a reactive adapter for OpenAL, allowing sounds to be generated and played as Behaviour Double. It's a little slow, but working. You can get the source from http://metafoo.co.uk/darcs/resound Comments, constructive criticism and especially patches welcomed :) Thanks, Richard (IRC:lilac)

3 2

Snapshotting a Behavior built with accumB seems to go wrong
by Eyal Lotem 27 Nov '08

27 Nov '08

It seems that going from Event(1) -> Behavior(2) -> Event(3) -> Behavior(4) is broken: http://trac.haskell.org/reactive/ticket/13 The first event(1), is for example a keyPressed, which is accumB'd into a Behavior(2). Then that Behavior is snapshotted(3) against a regular interval ticker (atTimes [0, gap..]), and then "stepper" is used to convert it to a behavior(4) again. The whole behavior(4) is broken and does not change at all, while the intermediate behavior(2) is just fine. Any idea what could cause this? Eyal

1 0

Re: GTK + Reactive
by Levi Stephen 26 Nov '08

26 Nov '08

Hi, This is no longer a problem I managed to get this working by adding the following lines: updater <- mkUpdater (cGetTime clock) (return () `stepper` quitOnDestory windowDestroy) idleAdd (updater >> return True) Levi On Mon, Nov 24, 2008 at 10:30 PM, Levi Stephen <levi.stephen(a)gmail.com>wrote: > Hi, > > I'm trying to set up a simple program combining Reactive and GTK. I have > the program below so far, but mainQuit is never called. After looking at a > couple of adapters I tried adding the commented out line, but this didn't > work for me. > > I'm sure I've probably missed something, but I'm not sure what. > > Thanks, > Levi > > main = do > initGUI > > clock <- makeClock > > (windowDestroy, sink) <- makeEvent clock > > w <- windowNew > w `onDestroy` sink () > widgetShowAll w > > -- forkE (tSync clock) $ quitOnDestroy windowDestroy > > mainGUI > > quitOnDestroy :: Event () -> Event (IO ()) > quitOnDestroy e = fmap (const mainQuit) e > > tSync :: Clock TimeT-> ITime -> IO () > tSync clock t = sleepPast (cGetTime clock) (exact t) > >

3 4

'stateE' --> 'mealy'
by Conal Elliott 25 Nov '08

25 Nov '08

I renamed 'stateE' and 'stateE_' to' mealy' and 'mealy_' in reactive-0.9.4 (just released). I've been meaning to change these names for a while now. I prefer these new names as being more clear & specific and connecting clearly to automata theory rather than to imperative programming. Please update your code if you use either function. David -- would you tweak the Events tutoria<http://netsuperbrain.com/blog/posts/introducing-reactive-events/>l accordingly? - Conal

3 9

Transforming Behaviors
by Greg Fitzgerald 25 Nov '08

25 Nov '08

Should it be possible to write this: spinningTorusPair :: Double -> Behavior Geometry3 spinningTorusPair sz = uscale3 sz *% spinningG torusPair Instead of this: spinningTorusPair :: Double -> Behavior Geometry3 spinningTorusPair sz = (uscale3 sz *%) <$> spinningG torusPair That is, should a Behavior of something transformable be an instance of Transform as well? Has anyone tried to implement this instance? Thanks, Greg

2 6

Re: [Haskell-beginners] GTK + Reactive
by Bas van Dijk 24 Nov '08

24 Nov '08

CCing this to the Reactive mailinglist. See: http://www.haskell.org/mailman/listinfo/reactive Bas On Mon, Nov 24, 2008 at 1:00 PM, Levi Stephen <levi.stephen(a)gmail.com> wrote: > Hi, > > I'm trying to set up a simple program combining Reactive and GTK. I have the > program below so far, but mainQuit is never called. After looking at a > couple of adapters I tried adding the commented out line, but this didn't > work for me. > > I'm sure I've probably missed something, but I'm not sure what. > > Thanks, > Levi > > main = do > initGUI > > clock <- makeClock > > (windowDestroy, sink) <- makeEvent clock > > w <- windowNew > w `onDestroy` sink () > widgetShowAll w > > -- forkE (tSync clock) $ quitOnDestroy windowDestroy > > mainGUI > > quitOnDestroy :: Event () -> Event (IO ()) > quitOnDestroy e = fmap (const mainQuit) e > > tSync :: Clock TimeT-> ITime -> IO () > tSync clock t = sleepPast (cGetTime clock) (exact t) > > > _______________________________________________ > Beginners mailing list > Beginners(a)haskell.org > http://www.haskell.org/mailman/listinfo/beginners > >

1 0

instance Monad Behavior where ...
by Thomas Davie 24 Nov '08

24 Nov '08

While hacking a bit on Creighton Hogg's Tetris game, I came across something interesting re composability. I have made each tile appear as a taurus, and decided it might be fun if the tauruses span while you played (just for added vomit-worthyness). So I set about changing my tile from a Geometry3 to a Behavior Geometry3. This went well, until it came to getting the spinning tiles to display something based on the game state. The tetris game is a Behavior GameState, where the game state stores the positions of the tiles on the board etc. What I *want* is that I compute the current game state, by looking into the Behavior GameState, get a GameState up, and then apply a function createTetrisGeometry :: GameState -> Behavior Geometry3. This sounds a lot like (>>=) to me, unfortunately there's no Monad instance for Behavior though :(. I've heard from various sources that the Monad instance is problematic for a couple of reasons. I'd love it if we could have a bit of a discussion about why it's problematic, and if it can be fixed. Thanks Bob p.s. It is possible to hack round this problem by making a function createTetrisGeometry :: TimeT -> GameState -> Behavior Geometry3, and then using (uncurry createTetrisGeometry) <$> (time <^(,)^> tetris). The result currently looks something like this... http://www.cs.kent.ac.uk/people/rpg/tatd2/tetris.png

1 0