A Python utility function for ingesting data into SQLAlchemy-defined PostgreSQL tables, automatically migrating them as needed, allowing concurrent reads as much as possible.
Allowing concurrent writes is not an aim of pg-bulk-ingest. It is designed for use in ETL pipelines where PostgreSQL is used as a data warehouse, and the only writes to the table are from pg-bulk-ingest. It is assumed that there is only one pg-bulk-ingest running against a given table at any one time.
pg-bulk-ingest exposes a single function as its API that:
- Creates the tables if necessary
- Migrates any existing tables if necessary, minimising locking
- Ingests data in batches, where each batch is ingested in its own transaction
- Handles "high-watermarking" to carry on from where a previous ingest finished or errored
- Optionally performs an "upsert", matching rows on primary key
- Optionally deletes all existing rows before ingestion
- Optionally calls a callback just before each batch is visible to other database clients
Visit the pg-bulk-ingest documentation for usage instructions.