Vertica is the columnar warehouse found beside many enterprise estates. In February 2025 it became a supported engine: a driver with the full type surface the platform expects, ten days of running the whole JDBC test suite against a live server, and one behaviour of the vendor driver that the platform now handles on the caller's behalf.
The driver
- Types Array, JSON, multiset, geometry, blob, user-defined types and text search each have a Vertica implementation. The numeric types are mapped and tested for decimal, double, float, integer, money and smallint.
- Administration Database creation, metadata, users, properties, activity and execution plans go through the same driver interfaces as every other engine, so the workbench and the storage tools work unchanged.
- Reserved words Vertica's PostgreSQL ancestry shows in its reserved-word list, which the driver carries so generated SQL quotes what it must.
One quirk, handled
- Two result sets The Vertica driver cannot keep two result sets open on one connection: the second fails once the first's 8 KB buffer is full. The platform now drains a result set into memory before opening another on Vertica, logging progress every thousand rows, and the report cursor consumes its buffer the same way.
- Proven by the suite The driver was added on 4 February. The following ten days were the platform's JDBC, data-type, date and table suites run against Vertica, with each difference fixed in the driver rather than in the tests.
Ten months later the dialect refactor gave Vertica its own DDL, DML, query and procedure dialects, with MERGE compiled natively. This release is where it entered the list.