Guides

Each guide covers one task, with examples you can run. For why genson behaves as it does, see How it works.

• Infer a schema from a JSON column: as a Polars schema, a JSON Schema or an Avro schema, per row or merged.
• Normalise a JSON column: into typed columns, one struct column, or JSON strings.
• Normalise JSON stored in Parquet: straight from a Parquet column to typed Parquet output, keeping other columns alongside.
• Factor out repeated embedded data: store data that repeats per key once, in a lookup table.
• Process large inputs: speed and memory for big files.