Parameter validation seems broken?

[richrobe]

The parameters of a class should have the same name as the arguments passed to the __init__. tpcp should show a descriptive error when this happens. However, this is not the case.

Example:

class Preprocessing(Algorithm):

    _action_methods = ["clean"]

    preprocessing_method: Parameter[str]

    icg_clean_: pd.DataFrame

    def __init__(self, method: str = "butterworth"):
        if method not in ["butterworth", "elliptic", "savgol"]:
            raise ValueError("Filter type can only be 'butterworth', 'elliptic', or 'savgol'")
        self.preprocessing_method = method

[AKuederle]

Turns out this was never supported... implemented it now, but was more complicated than expected. I hope I cought all edge cases.

Just as a side-note for future readers. The class shown by @richrobe in the issue, is NOT a valid tpcp algorithm class. Not just because it "changes the name", which we now check with the update, but also because it performs parameter validation within the init, which is an antipattern in tpcp that can cause unintended side effects, as validation might be skipped if parameters are set via set_params instead of using the init