DuckLake is DuckDB’s new lakehouse format: data remains in Parquet files, but metadata is efficiently tracked in a
separate DuckDB database.

• DuckLake supports function-based partitioning, which allows data to be partitioned by year and month. This enables
  efficient file pruning on tp_timestamp without needing a separate tp_date filter. A tp_date column will still
  be present for compatibility, but it is no longer required for efficient query filtering.
• Existing data should be automatically migrated when Tailpipe runs. Migration does not
  occur if progress output is disabled (--progress=false) or when using machine-readable output (json, line,
  csv).