Unclear operator evaluation from iteration and time step index

[Veljko Lipovac]

According to the docs of Operator.evaluate, it will pull values from the previous iterate, alternatively from the previous time step.

But when I instantiate my variable like this

and evaluate an algebraic combination of variables using simply .evaluate(ad_system), I get the following error.

This error happens during a call to .get_variable_values (in line 710 of operators.py)
The iterate_index argument is None and the time_step_index is 1, as per call from evaluate

Assuming I did everything right, there seems to be either an inconsistency in the documentation or a missing implementation?
According to the docs the implementation should be able to find the values stored at iterate index 0.

If I instantiate a variable with values at both iterate and time, the error does not happen.

[Eirik Keilegavlen]

Your final statement is the key: You need to initialize the relevant arrays before they can be fetched. So this is not a bug, but arguably a tacit assumption in the EquationManager.

[Ivar Stefansson]

Your final statement is the key: You need to initialize the relevant arrays before they can be fetched. So this is not a bug, but arguably a tacit assumption in the EquationManager.

Or EquationSystem, rather. But yes, I think it's fair that you need to set the value for the previous time step before being able to get that value.

[Veljko Lipovac]

Alright.

Then maybe the documentation should be updated, because to my understanding it says that it is optional ("alternatively").

And does that mean that time-independent problems also need a first time-index set for their variables?
It seemed intuitive to me that the iterative index is prioritized (like the previous approach took ITERATE first, than STATE)

[Ivar Stefansson]

From the point of view of the set_variable_values method, the parameter is indeed optional.