# custom_preprocessor.py

from sklearn.base import BaseEstimator, TransformerMixin
import pandas as pd

class CustomPreprocessor(BaseEstimator, TransformerMixin):
    def __init__(self):
        # No hyperparameters right now; keep for sklearn compatibility
        pass

    def fit(self, X, y=None):
        # Stateless: nothing to fit, but keep method for Pipeline API
        return self

    def transform(self, X):
        # Work on a copy to avoid mutating caller data
        df = X.copy()

        # Step 1: Replace 'reg' with 'Regular'
        if "Product_Sugar_Content" in df.columns:
            df.loc[df["Product_Sugar_Content"] == "reg", "Product_Sugar_Content"] = "Regular"

        # Step 2: Process object columns
        obj_cols = df.select_dtypes(["object"]).columns
        for col in obj_cols:
            # Lowercase, replace spaces, then cast to category
            df[col] = (
                df[col]
                .astype(str)
                .str.lower()
                .str.replace(" ", "_")
                .astype("category")
            )

        # Step 3: Convert numeric columns to float32
        for col in df.columns:
            if df[col].dtype.kind in "biufc":
                df[col] = df[col].astype("float32")

        return df