Dynamic checking of requirement steps in pipelines

[pgierz]

Checks if the pipelines attached to a rule fulfill in any way some tagged requirements.

Requirements look like this in the steps:

def my_awesome_step(data, rule):
    _satisfies_cell_methods = "mean: time"
    data = data.mean(dim="time")
    return data

A step could satisfy multiple requirements. These should be internal variables, and might be used, but I think that they probably should just be used for checking.

A RuleRequirement helper class is available in rule.py, which is used to parse the requirements. Then, each step of each pipeline is checked for such tags:

my_requirements = {"requirement_name": "cell_methods", "requirement_value": "mean: time"}
rule_requirement = RuleRequirement(**my_requirements)
my_pipeline = Pipeline([my_awesome_step])
rule_requirements.pipeline_fulfills_requirements(my_pipeline)

I think that the checking method could use a better name, suggestions are welcome.

This idea is still untested, just a starting point for discussion....

[pgierz]

Maybe this should gather all requirements, instead of exiting at the first one.

[pgierz]

I'm not terribly convinced by this, it requires that the name and value are strictly defined to be on the same line. That might not be the best way of doing it.

[pgierz]

This technically does what it is supposed to, but we need to discuss the implications of the crosschecking function.

At the moment, everything fails because none of our steps implement the cell_methods described in the integration tests. I think we could do something "dirty" like this:

def maybe_mask_fesom(data, rule):
    _satisfies_cell_methods = "area: mean where sea"
    if "fesom" in data.attrs:
        return data
    else:
        raise ValueError("No idea how to mask the ocean")