Transform

A transform action sits between a load and a write, reshaping the rows that flow through. Pick the transform type by what you’re doing to the data.

You want to…	Reach for	Guide
Filter, join, or reshape with a query	A SQL transform	Reshape data with a SQL transform
Do something SQL can’t express cleanly (loop over columns, call a library)	A Python transform	Reshape data with a Python transform
Rename and cast columns to a target schema	A schema transform	Enforce a schema on a view
Stage an intermediate result the rest of the flow reads	A temp table	Stage an intermediate result in a temp table
Drop rows that fail expectations	A data-quality transform (mode: dqe)	Gate rows with data-quality expectations
Keep failed rows for inspection and recycling instead of dropping them	A data-quality transform in quarantine mode	Quarantine bad rows instead of dropping them

Note

Data quality has two homes. A data_quality transform gates rows inside a flow (drop or quarantine). To assert a standalone property — uniqueness, a row count, referential integrity — as a unit test for your data, use a test action instead.