Documentation
¶
Overview ¶
Command adbc shows how a database, exposed through ADBC, feeds into Cosma as just another scan source. ADBC result batches are already Arrow, so the internal/ingest seam returns an array.RecordReader exactly like CSV and Parquet do, and the rest of the pipeline (DataFrame, Filter) is unchanged.
The real DuckDB ADBC driver is CGo and was not available in this module version, so this example uses an in-memory mock adbc.Database. Swapping in a real driver is a one-line change: build the driver's adbc.Database and pass it to ingest.ADBC.
Click to show internal directories.
Click to hide internal directories.