Scales cannot work properly from native code

[LourensVeen]

In a macro-micro situation, the macromodel runs once, and a native implementation of the macromodel can use will_stop() to test whether to exit its submodel execution loop. If you set the 'macro:T' and 'macro:dt' properties appropriately, this works.

However, for a native micromodel (which has to run more than once), it seems like it doesn't. A Java micromodel that inherits from muscle.core.kernel.Submodel can override init(), receive its f_init message in there, then return the timestamp of that message to be set as its originTime. The muscle.core.kernel.Submodel.willStop() method will then add the model's T to that originTime, and check that a message has been sent for that timestamp (i.e. an O_f) to decide whether to stop or not.

A native micromodel cannot do this, and has no way of setting originTime. As a result, will_stop() will only work correctly on the first submodel run. After that, it will always return true, since we're already past the time of the first model run. There does not seem to be any way for the native micromodel to communicate the originTime of its current run to the muscle.core.standalone.NativeKernel in the protocol or the native APIs.

The obvious work-around is to have the native model do its own time-keeping and not use will_stop(), but it would be better to fix this in a new native API.

[LourensVeen]

Actually, doing your own time-keeping is also problematic, since you can't get the timestamp that goes with an incoming message in the native API. So there's no way of setting the originTime, and your micromodel is left starting over and over again at time 0. Since physics is time-independent that shouldn't change your simulation results :), but it's still annoying.