Batch Processing

• Simulates a daily ETL job:
  I created a DataFrame of transactions, save it as a CSV, clean and transform the data, aggregate totals per customer, and save the results as Parquet files.
• SQL-like queries:
  I used DuckDB to run SQL queries on the aggregated Parquet data.
• Result:
  The final customer totals are saved in data/batch_customer_totals.parquet.
  Example (from my file):
  • David: 300
  • Bob: 200
  • Alice: 100
  • Charlie: 50

Stream Processing

• Simulates real-time transaction events:
  I used the Streamz library to emit random transaction events one by one.
• Real-time aggregation:
  Each event updates the running total for each customer instantly.
• Result:
  After all events, the final totals are saved in data/stream_totals.parquet.