Enhance StepFunctionPower component to allow multiple instances

[taldcroft]

Description

This allows including multiple instances of the StepFunctionPower component on the same node by adding an id attribute. This identifier is put into the component name (via the str function). The default id is an empty string for back-compatibility.

Reminder: this needs to be ported to the shiny branch (by cherry-pick or rebase).

To make the model more self-documenting, I would recommend setting the id to be the date of the step, for instance _2020213 for the July 31 swap.

Approval

Reviewed and approved by TWG on July 21, 2020.

Testing

• [x] Passes unit tests on MacOS (ska3-flight)
• [x] Functional testing

Functional

Made a new ACA model aca_spec_3_step which has a step function up at 2020:100 and same step function back down at 2020:120. This shows that the additional step functions are having the intended effect:

• No impact on model predictions prior to 2020:100
• Model prediction jumps up at 2020:100
• Model prediction jumps back down to original level at 2020:120

The new model spec as Python looks like:

model.add(xija.StepFunctionPower,
          node='aca0',
          time='2020:100:00:00:00',
          P=0.1,
          id='_2'
         )
model.add(xija.StepFunctionPower,
          node='aca0',
          time='2020:120:00:00:00',
          P=-0.1,
          id='_3'
         )

and as JSON:

        {
            "class_name": "StepFunctionPower",
            "init_args": [],
            "init_kwargs": {
                "P": 0.1,
                "id": "_2",
                "node": "aca0",
                "time": "2020:100:00:00:00"
            },
            "name": "step_power_2__aca0"
        },
        {
            "class_name": "StepFunctionPower",
            "init_args": [],
            "init_kwargs": {
                "P": -0.1,
                "id": "_3",
                "node": "aca0",
                "time": "2020:120:00:00:00"
            },
            "name": "step_power_3__aca0"
        }

[jeanconn]

This looks good to me. Could your functional test work as a unit test @taldcroft ?

[matthewdahmer]

Looks good to me as well!