No metrics property on the comparer/collection

[ecomodeller]

The only drawback is the need to explicitly set metrics on directional data (for now).

[jsmariegaard]

Mostly good, but we need to do something about the handling of default metrics for directional data.

How about defining a property on Comparer and Collection:

@property
def _default_metrics(self):
    default_metrics = [mtr.c_bias, mtr.c_rmse, mtr.c_urmse, mtr.c_max_error] if self.quantity.is_directional else None
    return default_metrics

and then do this in skill() and similar methods:

metrics = _parse_metric(metrics, default_metrics=self.default_metrics, return_list=True)

[jsmariegaard]

Or maybe passing is_directional to the _parse_metric function. I think we should have a better solution in the long run - maybe we should have a ms.options.metrics.list_directional along side with the current ms.options.metrics.list 🤔

[ecomodeller]

Or maybe passing is_directional to the _parse_metric function. I think we should have a better solution in the long run - maybe we should have a ms.options.metrics.list_directional along side with the current ms.options.metrics.list 🤔

See update...

[jsmariegaard]

Almost there, just need to allow user to define default metrics like this (see the failing notebook):

 ms.options.metrics.list = ["kge", "cc"]

[jsmariegaard]

Now fixed