[mg87053] Re: Dynamic and J/Link

[David Bailey <dave at Remove_Thisdbailey.co.uk>]

J. McKenzie Alexander wrote:
> I'm using Java to do some agent-based modeling and am wondering what  
> the best way is to create a dynamic graphic that will update  
> automatically whenever the state of the simulation changes.
> 
> Suppose - for sake of argument -  that the simulation is a single Java  
> object saved in a Mathematica variable called "model".  Suppose that  
> calling the Java method step() advances the simulation one iteration  
> and that calling the Java method getState() returns a rectangular 2D  
> array of 0s and 1s.  Now, one easy way of representing the state of  
> the model is by using Raster, so the following gives a basic graphic  
> representation:
> 
> 	Graphics[ Raster[model@getState[]] ]
> 
> Suppose, now, that we wrap that statement with Dynamic as follows:
> 
> 	Dynamic[
> 		Graphics[ Raster[ model@getState[] ] ]
> 	]
> 
> If I then evaluate the following, the displayed graphic doesn't change  
> at all, even though (intuitively, at least) it should:
> 
> 	model@step[];
> 	Update[ model ]
> 
> For some reason, calling Update on the variable model doesn't cause  
> the Dynamic object in the notebook to recognize that it needs to  
> refresh the displayed graphic.  Why is that?
> 
> Now, the only way I've found to solve this problem is the draw the  
> display using a second variable which contains a local copy of the  
> state of the Java object. First, set up the initial display in the  
> notebook as follows:
> 
> 	array = model@getState[];
> 	Dynamic[
> 		Graphics[ Raster[ array ] ]
> 	]
> 
> then manually update the variable array each time the simulation  
> changes:
> 
> 	model@step[];
> 	array = model@getState[]; (* This triggers a redisplay *)
> 
> The two irritations about this are that it (a) it requires polluting  
> the current Mathematica's session namespace with another variable that  
> doesn't do anything except hold a value which could be easily obtained  
> by calling model@getState[], and (b) it requires some (minor)  
> additional code to configure each display.  Those aren't real worries,  
> but it does suggest that I'm not doing this the most efficient way.
> 
> Is the above solution the best (or, indeed, only) way to get Dynamic  
> to recognize state changes of J/Link objects?  Does anyone else have  
> any other suggestions?
> 
> Many thanks,
> 
> Jason
> 
> --
> Dr. J. McKenzie Alexander
> Department of Philosophy, Logic and Scientific Method
> London School of Economics and Political Science
> Houghton Street, London WC2A 2AE
> 
> 
> 
> 
> 
> 
> Please access the attached hyperlink for an important electronic communications disclaimer: http://www.lse.ac.uk/collections/secretariat/legal/disclaimer.htm
> 
Your variable 'model' contains a reference to an instance of a Java 
class (somewhat analogous to a pointer in C), it does not contain the 
instance itself, which lies in a completely separate Java process. Thus, 
as far as Mathematica is concerned, the variable does not change when 
you call step.

I would just wrap your two lines in a function and call that:

UpdateModel[]:=
(
model@step[];
array=model@getState[];
);

David Bailey
http://www.dbaileyconsultancy.co.uk