When timeseries have mismatched indexes, SpineOpt execution sometimes neither errors nor continues

[Karlis-Baltputnis]

Describe your problem

Normally, if there is something wrong with the time indexes of a timeseries, SpineOpt produces an error during constraint creation. However, there are sometimes specific cases when the constraints are seemingly created successfully, an error is not produced, the Run SpineOpt hammer in the workflow keeps hammering indefinitely and the Julia console shows no new information.

I have attached an example input database, which is designed to have the above mentioned behavior happen. It is the rolling horizon tutorial, which I have modified by changing the vom_cost of plant a to a fixed resolution timeseries with same number of elements as electricity demand timeseries, but shifted in indexes five hours forward. So the expected outcome would be for some sort of error to occur, seeing how there is data missing for the first five hours of the model.

Instead, the last output to console is the beginning of optimization:

Perhaps it is not necessarily a problem, as it does not lead to false results or anything like that. But it is a bit annoying when I have made a small mistake in a large model and for a while sincerely think that it is being solved, before it dawns on me that a solution is not going to be produced.

Attach your model as a JSON file (encouraged)

Example_w_infinite_execution.json

What versions are you using?

SpineOpt: 0.9.2
SpineInterface: 0.15.1
SpineToolbox: 0.9.2
Julia: 1.11.2

[tarskul]

Strange, when I run your example, I get an error message and it does not keep running indefinitely. Though, the tool does give me a green checkmark instead of a red crossmark.

Bringing simple to the first window...the expression below resulted in a non-finite value - this is probably due to missing time-series data:

{vom_cost(analysis_time=Base.RefValue{Dates.DateTime}(Dates.DateTime("2000-01-01T00:00:00")), unit=power_plant_a, node=fuel_node, direction=from_node, stochastic_scenario=realization, t=2000-01-01T00:00~(1 hour)~>2000-01-01T01:00) = NaN} * {node_stochastic_scenario_weight(node=fuel_node, stochastic_scenario=realization) = 1.0}

The only difference between our setups is that I have Julia 1.10 (and probably that I run spine tools on linux) but I would be surprised if that is the actual issue here.