Split out signatures from our current metrics (i.e., the "metrics" that only take one timeseries, mean, min, max, etc.)
Create a new class of metrics that first calculates the signature for primary and secondary (based on some initial grouping, event, annual, etc), and then calculates a metric on top of that. A two-step process.
The returned "metric" could be just the final metric or a dictionary that contains the signatures (A series of signatures) and the signature metric.
There is still something missing in this concept I think. We can currently do this with chained queries I think.
Under this issue, we want to:
There is still something missing in this concept I think. We can currently do this with chained queries I think.