Looks great! I would suggest to add the parametrize_with_checks testing right a the start, for "test driven development"

Adding the following method (BestEstimator) leads to an AttributeError: "property 'n_features_in_' of 'HyperactiveSearchCV' object has no setter". Adding a setter does not solve the problem, so I will remove this method for now.

    @property
    def n_features_in_(self):
        try:
            check_is_fitted(self)
        except NotFittedError as nfe:
            raise AttributeError(
                "{} object has no n_features_in_ attribute.".format(
                    self.__class__.__name__
                )
            ) from nfe

        return self.best_estimator_.n_features_in_

Looks great! I would suggest to add the parametrize_with_checks testing right a the start, for "test driven development"

Test added in 9c0d56e

Adding the following method (BestEstimator) leads to an AttributeError: "property 'n_features_in_' of 'HyperactiveSearchCV' object has no setter". Adding a setter does not solve the problem, so I will remove this method for now.

    @property
    def n_features_in_(self):
        try:
            check_is_fitted(self)
        except NotFittedError as nfe:
            raise AttributeError(
                "{} object has no n_features_in_ attribute.".format(
                    self.__class__.__name__
                )
            ) from nfe

        return self.best_estimator_.n_features_in_

Is the best_estimator_ supposed to be a blueprint (state as after __init__), or a fitted stimator (state as after fit)? Maybe it is the latter and that is the issue?

From a design perspective, the former might make more sense, but I would go with whatever is consistent with sklearn GridSearchCV.

I would suggest to separate the CI modifications to a separate PR - just for cleanness, and eas of localizing errors if occurring.

Btw, if you squash PRs, no rebase or cherrypick magic is needed if you want to avoid that

I will keep that in mind in the future.