November 2008 - Reactive

Space leak if sub-Behaviors dynamically generated?
by Robin Green 24 Nov '08

24 Nov '08

I am trying to use Reactive to write a build system, initially for Haskell modules, which stays running and rebuilds stuff when files are modified, created etc. However, I have run into a problem with adding files to the project to be built. Suppose you generate separate Behaviors for the contents of each file (Behavior (Maybe ByteString)). Then, whenever you add a new file, a new Behavior must be dynamically created, which records Just the contents of this file, or Nothing when it did not exist, from the application startup time onwards. ("Dynamically created" because it's not possible to statically list all the filenames the developer might choose to use, obviously!) It seems to me that this entails, by the very design of Reactive, keeping around the complete contents of every version of every file, forever. *We* know that the old contents, if any, are "dead", but the garbage collector probably won't be able to detect that. If, on the other hand, you just have one single Behavior for this purpose (Behavior (Map FilePath ByteString)), this will change any time *anything* changes on the filesystem, so it doesn't directly support triggering a recomputation/recompilation based on *particular* changes. This problem seems to me to generalise to any case where you are trying to dynamically generate "sub-behaviors" from existing Behaviors, at arbitrary points in time. Is my analysis correct? Have I missed something? -- Robin

2 1

Whoever implements splitB gets a virtual beer from me.
by David Sankel 24 Nov '08

24 Nov '08

I'd like to design a function that splits behaviors based on the time of an event occurrence. Here is the type signature: splitB_ :: Behavior a -> Behavior a -> Event () -> Event (Behavior a) An example in semantics: splitB_ (const 1) (const 2) [(0,()), (1,())] => [ (0, (\t -> if t <= 0 then 1 else 2)) , (1, (\t -> if t <= 1 then 1 else 2)) ] Operationally, the implementation must avoid traversing all of the second behavior's reactive's steps at every occurrence of the event, so the following implementation isn't going to win anyone a beer: splitB_ :: Behavior a -> Behavior a -> Event () -> Event (Behavior a) splitB_ bLeft bRight e = fmap f (withTimeE_ e) where f ot = iffB (fmap (ot >=) time) bLeft bRight iffB :: Behavior Bool -> Behavior a -> Behavior a -> Behavior a iffB = liftA3 iff iff a b c = if a then b else c Thanks! David -- David Sankel Sankel Software

1 1

switcher == scanlB const?
by Greg Fitzgerald 24 Nov '08

24 Nov '08

I think my space leak in Bounce came from my assumption that "switcher == scanlB const." When I replace "scanlB const" with "switcher," the behavior is the same but that space leak goes away. -Greg

1 0

Bounce
by Greg Fitzgerald 22 Nov '08

22 Nov '08

How do I make a Behavior of a bouncing motion? I want to reverse the velocity when the object goes beyond a certain position, but since the position depends on the velocity, I have a loop. Is there a way to make recursive behaviors, or am I going about this all wrong? bouncingPosition :: Behavior Double bouncingPosition = ??? bouncingVelocity :: Double -> Behavior TimeT -> Behavior Double bouncingVelocity v0 t = velocityB v0 t `switcher` ??? collision :: Event a -> Behavior Double -> Event Double collision e = once . filterMP (< -2) . snapshot_ e positionB :: Double -> Behavior Double -> Behavior Double -> Behavior Double positionB x0 v t = (x0 +) <$> liftA2 (*) v t velocityB :: Double -> Behavior Double -> Behavior Double velocityB v0 t = velocity v0 <$> t velocity :: Double -> TimeT -> Double velocity v0 t = v0 - 9.8 * t Thanks, Greg

4 11

exporting animations
by Greg Fitzgerald 21 Nov '08

21 Nov '08

I'd like to pass animations around office. PNG or AVI would be most convenient. How do you guys recommend doing it? How did you make those FieldTrip YouTube videos? Thanks, Greg

3 2

"Time travel" with withNextE
by Robin Green 21 Nov '08

21 Nov '08

withNextE seems to allow occurrences' values to depend on occurrences which haven't yet occurred. Of course, it doesn't actually allow time travel. If the occurrence is not yet computed, presumably it will block until it does. Still, this is problematic theoretically. It allows "simulations" to be constructed which contain reverse causality (if the simulation makes all the input occurrences available immediately). Is this a case of making the framework a little too powerful? -- Robin

4 3

unsafePerformIO: is it correctly used?
by Peter Verswyvelen 21 Nov '08

21 Nov '08

When I was browsing the GHC documentation, I came across the following: If the I/O computation wrapped in unsafePerformIO<file:///C:/app/ghc-6.10.1/doc/libraries/base/System-IO-Unsafe.html#v%3AunsafePerformIO> performs side effects, then the relative order in which those side effects take place (relative to the main I/O trunk, or other calls to unsafePerformIO<file:///C:/app/ghc-6.10.1/doc/libraries/base/System-IO-Unsafe.html#v%3AunsafePerformIO>) is indeterminate. You have to be careful when writing and compiling modules that use unsafePerformIO<file:///C:/app/ghc-6.10.1/doc/libraries/base/System-IO-Unsafe.html#v%3AunsafePerformIO> : - Use {-# NOINLINE foo #-} as a pragma on any function foo that calls unsafePerformIO<file:///C:/app/ghc-6.10.1/doc/libraries/base/System-IO-Unsafe.html#v%3AunsafePerformIO>. If the call is inlined, the I/O may be performed more than once. - Use the compiler flag -fno-cse to prevent common sub-expression elimination being performed on the module, which might combine two side effects that were meant to be separate. A good example is using multiple global variables (like test in the example below). - Make sure that the either you switch off let-floating, or that the call to unsafePerformIO<file:///C:/app/ghc-6.10.1/doc/libraries/base/System-IO-Unsafe.html#v%3AunsafePerformIO> cannot float outside a lambda. For example, if you say: f x = unsafePerformIO (newIORef []) you may get only one reference cell shared between all calls to f. Better would be f x = unsafePerformIO (newIORef [x]) because now it can't float outside the lambda. Now in Reactive, in e.g. the internal IVar.hs module, I see readIVar :: IVar a -> a readIVar (IVar v) = unsafePerformIO $ readMVar v At first sight this function is not following all the rules described above. Could this cause problems? Cheers, Peter

3 3

To fmap fmap or not?
by Peter Verswyvelen 20 Nov '08

20 Nov '08

Many of the fmap fmaps I've seen come from (fmap f) . g = (fmap f) `fmap` g = fmap (fmap f) g = (fmap (fmap f)) g = ((fmap.fmap) f) g = (fmap.fmap) f g But... to me the former is much clearer than the latter... It is less abstract since it uses (.) instead of fmap. However, I also see (fmap.fmap.fmap), so Conal must have seen a pattern here? Of course one could also write (fmap `fmap` fmap `fmap` fmap) instead of (fmap.fmap.fmap). Or as was already mentioned with the generic replacement (.) = fmap, (fmap.fmap.fmap) can be written as ( (.) . (.) . (.) ) But besides this all being very Haskelly Hackery - and maybe ASCII art that needs to be censured - I don't see the light yet :) Could somebody enlighten me?

4 5

First draft of reactive-tetris
by Creighton Hogg 20 Nov '08

20 Nov '08

Hey Reactive, Since I have no hosting space I'm actually going to include the implementation as an attachment all in one file. It's not very long anyway, so it should be alright. A few caveats: This code is still _really_ ugly, but it does work as far as I've tested There is no game over, you need to quit manually There is no score or level adjustment yet. I need to add that when I have more time, i.e. after this week That being said, my first real reactive program was surprisingly simple once I got comfortable with the semantics. The actual reactive code isn't that complicated, but it did take a pretty big shift in thinking from writing explicit game loops. I've included a good bit of comments so it hopefully it isn't too hard to follow, but if there are any questions about why I did something a particular way (the answer may be 'I'm dumb') please e-mail me or find me on #haskell.

7 20

Representation Details of Events
by David Sankel 19 Nov '08

19 Nov '08

Consider these two events: a = [] b = [(+infty, _|_)] Semantically, a and b are different values. Consider their representations in reactive: a = Event (Future ((Max MaxBound), (_|_))) b = Event (Future ((Max MaxBound), (_|_ `Stepper` a)))) At the end of an event, the second element of the future tuple is bottom. Conversely, the second element of the future tuple is never bottom when not at the end of an event. There's no particular point to this post, just musing on my discovery of the above facts. -- David Sankel Sankel Software

1 1