Refactor Feature Store as a Strategy Pattern

import pandas as pd
from sklearn.base import TransformerMixin
from collections import namedtuple
from sklearn.preprocessing import StandardScaler
from sklearn_pandas import DataFrameMapper
import numpy as np

AttributeValue = namedtuple('AttributeValue', ['a', 'v'])

class Identity(TransformerMixin):
    def fit(self, X, y=None):
        return self
    def transform(self, X):
        return X

class TimeFilter(TransformerMixin):
    def __init__(self, start_date=AttributeValue("start_date", 1), end_date=AttributeValue("end_date", 4)):
        """
        Presume the configuration is attribute-value pairing on the data set
        We can make them objects, but set them as a dict for ease of reading.

        Usage:
        X = pd.DataFrame({"x": [1,2,3,4], "start_value":[0,1,2,3], "end_value": [2,3,4,5]})

        TimeFilter().tranform(X)
        """
        self.start_date = start_date
        self.end_date = end_date

    def fit(self, X, y=None):
        return self

    def transform(self, df):
        return df[(df[self.start_date.a] >= self.start_date.v) & (df[self.end_date.a] <= self.end_date.v)]


if __name__ == "__main__":
    mapper = DataFrameMapper([
        (['x'], StandardScaler()),
        ('start_date', Identity()),
        ('end_date', Identity()),
    ], df_out=True)

    mapper.fit_transform(TimeFilter().transform(X))

    # now apply onto a bigger dataset, for multiple time periods...

    X_full = pd.DataFrame({"x": np.arange(100), "start_date":np.arange(100), "end_date": np.arange(100)+2})

    df_all = []
    for start_date, end_date in zip(np.arange(90), np.arange(90) + 3):
        df_all.append(mapper.transform(TimeFilter(start_date=AttributeValue("start_date", start_date), end_date=AttributeValue("end_date", end_date)).transform(X_full)).assign(timestamp=start_date))

    # and here we have applied the said feature over multiple time periods!
    df_all = pd.concat(df_all)
    print(df_all)

    for start_date, end_date in zip(np.arange(90), np.arange(90) + 3):
        df_all.append(mapper.transform(TimeFilter(start_date=AttributeValue("start_date", start_date), end_date=AttributeValue("end_date", end_date)).transform(X_full)).assign(timestamp=start_date))