[fluxus] 0.12b1 + cvs merge

[Claude Heiland-Allen]

Dave Griffiths wrote:
>> thinking about it some more, the front end app is scheme independant now
>> too, it would be doable to use it as an editor to a haskell
>> interpreter/repl with some equivelent to every-frame.

Excellent!

> and - thinking of haskell particually, there are a few things about the
> philosophy going on with fluxus. basically what it boils down to is that
> computer graphics works very very well imperatively.

Haskell can sequence actions too, most commonly with the IO monad.

> things like (push) (pop)

HOpenGL has something like:

withPushPop action = do { push ; action ; pop }

but I've forgotten its real name.

 > and (grab) use stacks for hierachically storing
> state, which would be (I imagine) considered very bad taste to the
> hardcore functional programmer.

I have impeccably bad taste ;)

 > scheme is quite well matched to this as
> it's a hybrid approach - and I like it for that. it's something I've been
> meaning to mention at some point, because it sort of happened by accident.

I understand.

> I guess you could implement (draw-cube) etc - but would you have to
> specify the entire state with it every time?

Yes and no.  You could have something like:

drawSphereBase :: FluxusStateHandle -> Float -> Float -> IO ()
drawSphereBase handle count1 count2 = do
    state <- readIORef handle
    ... call out to libfluxus ... [*]
    return ()

and define draw-sphere with the state handle inside it:

fluxus = do
   state <- ... call out to C to get a pointer to Fluxus' state ...
   handle <- newIORef state
   drawSphere 10 10
   where
     drawSphere :: Float -> Float -> IO ()
     drawSphere = drawSphereBase handle


But that's only necessary if the call in [*] needs to be passed a 
pointer to Fluxus' state.  Otherwise, if the state is stored entirely in 
libfluxus, you could just:

drawSphere :: Float -> Float -> IO ()
drawSphere = do { ..call libfluxus.. ; return () }

and

buildSphere :: Float -> Float -> IO (Ptr sometype)
sphere <- buildSphere 10 10


It's essentially the same issue as trying to do OOP in C rather than 
C++, you have to pass *this explicitly rather than it being passed for you.


 > I think the answer would
> involve the word monad somewhere :)

:)


Claude