Lazy definition of the experiment core tasks by wrapping instances, methods and arguments. Creates a graph with task nodes, while assigning task-parents to each node, depending on each Evaluation signature. The tasks can then be run sequentially as a list or asynchronous using the graph’s node dependencies. For instance:

  • A forecast can only be made if catalog was filtered to its window

  • A consistency test can be run if the forecast exists in a window

  • A comparison test requires the forecast and ref forecast

  • A sequential test requires the forecasts exist for all windows

  • A batch test requires all forecast exist for a given window.