The DATASET data type is a new Complex Data Type (CDT) supported by Version 16.0 of the Teradata database that provides a mechanism to store self-describing data. Self-describing data contains both raw data and matching schema. A schema describes how the data is structured. The schemas do not need to be predefined, for example, during table creation time. Schemas can be defined just before inserting the data, or they can be packaged together with the data as data containers; this data architecture can loosely be called late binding.
In the initial 16.0 release, the adopted storage format of a DATASET data type is the Apache Avro format. Starting from 16.20, in addition to Avro, it supports CSV storage format as well. DATASET data can be inserted or imported into a DATASET column in the Avro binary format or in CSV format.
The name SQL_TD_DATASET_AVRO and SQL_TD_DATASET_CSV is defined in the tdsql.h file.
DATASET data can be retrieved as Avro binary or as CSV format and can be converted to JSON format.