A new bronze table ingests JSON events from cloud storage. An engineer proposes casting every field to its final numeric or date type during bronze ingestion so downstream code is simpler. What does Databricks medallion guidance recommend instead?
Choose an answer to check it. Afterward, select any option to review its explanation.
Bronze keeps data close to its raw form, mostly as string, VARIANT or binary, so a source schema change does not drop data. Type casting is one of the operations performed when silver tables are built from bronze.
Silver exists to deliver validated, typed records; pushing casting to every analyst query skips the cleanup the silver layer is responsible for.
Gold holds aggregated, business-aligned data. Casting has to happen before aggregation, in silver, or the aggregates would be built on untyped values.
Schema enforcement and type casting are silver-layer operations; bronze is meant to do minimal validation, so moving casting there and dropping enforcement in silver inverts the pattern.
id q1d160f8f09