SEMI-STRUCTURED DATA IN APACHE ICEBERG: MEET THE VARIANT TYPE

Summary from the TLDR Data newsletter

Iceberg v3 added a Variant type for semi-structured data, so fast-changing JSON-like records can live in one column while preserving native types like timestamps, decimals, and binary values. It uses Parquet's binary encoding, works across supported engines including Spark and Flink, and can be queried from Spark SQL with variant_get.