There is one TODO item in most sets that is we'd like to change the logic when ref data is not available when getting the reference climatology dataset for the variable and season.
The current logic:
If the reference climatatology does not exist or could not be found, it
will be considered a model-only run. For this case the test dataset
is returned as a default value and subsequent metrics calculations will
only be performed on the original test dataset.
When refactor, it has been troublesome, e.g. it would be useful to provide explicit error info for trouble shooting, also it can be easily to use test data as ref by mistake based on my recent experience refactoring.
Reference TODO item logged by @tomvothecoder
# TODO: This logic was carried over from legacy implementation. It
# can probably be improved on by setting `ds_ref = None` and not
# performing unnecessary operations on `ds_ref` for model-only runs,
# since it is the same as `ds_test`. In addition, returning ds_test
# makes it difficult for debugging.
I think we perhaps should prioritize this item before more sets are being converted.
There is one TODO item in most sets that is we'd like to change the logic when ref data is not available when getting the reference climatology dataset for the variable and season.
The current logic: If the reference climatatology does not exist or could not be found, it will be considered a model-only run. For this case the test dataset is returned as a default value and subsequent metrics calculations will only be performed on the original test dataset.
When refactor, it has been troublesome, e.g. it would be useful to provide explicit error info for trouble shooting, also it can be easily to use test data as ref by mistake based on my recent experience refactoring.
Reference TODO item logged by @tomvothecoder
I think we perhaps should prioritize this item before more sets are being converted.