A variant data type stores semistructured or evolving attributes when fields may appear, disappear, or change shape across records. It lets ingestion absorb source variability without forcing immediate rigid schema decisions, while known structure can remain governed and optimized through explicit schema evolution. section 6.1.3
Use variant columns when:
- Event streams, APIs, or logs produce records with changing fields or shapes. section 6.1.3
- A new source field should be routed into a flexible column instead of immediately extending the table schema. section 6.1.3
- The ingestion job still needs to inspect incoming schemas, validate mappings, and apply compatible transformations. section 6.1.3
Ingestion decision
When a new field appears, an ingestion pipeline can extend the table schema, route the field into a variant column, or cast it to a compatible type. Poor handling risks rejected records, silent corruption, or inconsistent queries. section 6.1.3
