biplobgon/product-recommendation-system

1"""2data_prep.py3------------4Load and inspect the raw Retailrocket events dataset.5 6Data source priority:7  1. Local file at data/raw/events.csv  (fast path – no network required)8  2. Google Cloud Storage bucket        (auto-downloaded when local file is absent)9 10Set GCS_BUCKET_NAME (and optionally GCS_CREDENTIALS / GCS_PROJECT_ID) in a11.env file or as environment variables to configure the GCS connection.12See .env.example for details.13"""14 15import os16import sys17 18import pandas as pd19from dotenv import load_dotenv20 21load_dotenv()22 23# ---------------------------------------------------------------------------24# Resolve project root so the script works when called from any directory25# ---------------------------------------------------------------------------26_HERE = os.path.dirname(os.path.abspath(__file__))27_PROJECT_ROOT = os.path.dirname(_HERE)28EVENTS_PATH = os.path.join(_PROJECT_ROOT, "data", "raw", "events.csv")29 30 31def load_events() -> pd.DataFrame:32    """Return the events DataFrame, fetching from GCS if the file is missing."""33    if not os.path.exists(EVENTS_PATH):34        print(f"Local events file not found at {EVENTS_PATH}.")35        print("Attempting to download from Google Cloud Storage …")36        try:37            from gcs_loader import load_events as _gcs_load  # noqa: PLC041538        except ImportError as exc:39            print(f"ERROR: Could not import gcs_loader: {exc}", file=sys.stderr)40            sys.exit(1)41 42        try:43            return _gcs_load(local_path=EVENTS_PATH)44        except FileNotFoundError as exc:45            print(f"ERROR: Dataset file not found in GCS: {exc}", file=sys.stderr)46            sys.exit(1)47        except Exception as exc:  # noqa: BLE00148            # Covers google.auth.exceptions.DefaultCredentialsError and other49            # GCS / network failures with a helpful remediation hint.50            print(f"ERROR: Could not download events data from GCS: {exc}", file=sys.stderr)51            print(52                "Please ensure:\n"53                "  1. GCS_BUCKET_NAME is set correctly (see .env.example).\n"54                "  2. You are authenticated (gcloud auth application-default login\n"55                "     or GCS_CREDENTIALS points to a valid service-account key).",56                file=sys.stderr,57            )58            sys.exit(1)59 60    return pd.read_csv(EVENTS_PATH)61 62 63if __name__ == "__main__":64    events = load_events()65 66    print(events.head())67    print(events.info())68    print(events["event"].value_counts())69