"""
Generate metrics and enable regeneration of metrics if new metrics are generated
Key tasks:
- Specify metric functions that take in prediction vs actual to generate a metric value
- Given a model and test data, produce a suite of all metrics
"""

import pandas as pd
from core.Settings import OPTIMISE_METRIC
from MLModel.BaseMLModel import MLModel


def sort_by_metric(
    data: pd.DataFrame, optimse_metric: str, best_model_column_name: str
) -> pd.DataFrame:
    """
    Helper function to sort data frame by metric and append a best model flag
    """
    data = data.sort_values(optimse_metric, ascending=False).reset_index(drop=True)
    data[best_model_column_name] = [False] * len(data)
    data.loc[0, best_model_column_name] = True

    return data


class Metrics:
    """
    All metric functions used to generate a dictionary of metrics
    """

    @staticmethod
    def metric_1(predictions: pd.Series, actuals: pd.Series) -> float:
        """
        Can leverage ML packages like sklearn for individual metrics like MAPE etc
        """
        pass

    @staticmethod
    def metric_2(predictions: pd.Series, actuals: pd.Series) -> float:
        """
        Can leverage ML packages like sklearn for individual metrics like MAPE etc
        """
        pass

    def list_metric_functions(self) -> list:
        """
        Gather all metric functions to run
        """
        pass

    def generate_metric_suite(
        self, model: MLModel, data: pd.DataFrame, target_column: str
    ) -> pd.Series:
        """
        For the model, test data and target, generate predictions and then iterative over all metrics to generate a Series of metric values
        """
        predictions = model.generate_predictions(data=data)
        actuals = data[target_column]

        metric_dict = {}
        for key, metric_function in asd:  # TODO:
            metric_dict[key] = metric_function(predictions, actuals)

        metrics = pd.Series([metric_dict])

        return metrics