Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

coverage: 88.664% (+0.5%) from 88.199%
when pulling 7dbfe6c on sampling_simplification
into 1ef4854 on master.

1. It removes the Policy and Scenario class. Behaviourally, there was no difference between them, so from a coding pov there is no reason for them to exist.

So Policy and Scenario are both replaced by Sample?

Are these often used by users? I'm thinking a bit about how Sample is a more abstract name than Policy and Scenario, which have a clear meaning. Also a Policy is always a set of policy levers, while a scenario is a set of uncertainties, right? Having that clear separation for a user might beneficial, one way or the other.

Also I think if you have one Policy and one Scenario there are never duplicate variables (levers/uncertainties). That's something you could check for. With only Sample that might be different.

Maybe it doesn't need to be different classes, but it could have a type or label, which could be "policy", "scenario" or "mixed".

So Policy and Scenario are both replaced by Sample?

Yes

Are these often used by users? I'm thinking a bit about how Sample is a more abstract name than Policy and Scenario, which have a clear meaning.

I agree that this is a drawback of this change.

Also a Policy is always a set of policy levers, while a scenario is a set of uncertainties, right? Having that clear separation for a user might beneficial, one way or the other.

Currently, there is no difference between Point, Policy, and Scenario other than the class name. Also, whether a given parameter is a lever or uncertainty is defined at the model level, but Point, Policy, and Scenario are not tied to a model and so cannot do any check of this kind.

Also I think if you have one Policy and one Scenario there are never duplicate variables (levers/uncertainties). That's something you could check for.

See above, there is currently already no way to do this within any of these classes. It might be something to check in Experiment, but I am not sure about the value of this.

Would it be useful to add deprecation warnings, at least for the much used Scenario and Policy?

# ema_workbench/em_framework/points.py
class Scenario(Sample):
    def __init__(self, *args, **kwargs):
        warnings.warn(
            "The `Scenario` class has been removed. Use `Sample` instead. See https://github.com/quaquel/EMAworkbench/pull/420",
            DeprecationWarning,
            stacklevel=2,
        )
        super().__init__(*args, **kwargs)

class Policy(Sample):
    def __init__(self, *args, **kwargs):
        warnings.warn(
            "The `Policy` class has been removed. Use `Sample` instead. See https://github.com/quaquel/EMAworkbench/pull/420",
            DeprecationWarning,
            stacklevel=2,
        )
        super().__init__(*args, **kwargs)

__all__ = ["Sample", "Scenario", "Policy"]

# ema_workbench/__init__.py
from .em_framework.points import Sample, Scenario, Policy
__all__ = ["Sample", "Scenario", "Policy"]