Apache Hive Tables

Introduced or updated: v1.2.668

Databend can query data that is cataloged by Apache Hive without copying it. Register the Hive Metastore as a Databend catalog, point to the object storage that holds the table data, and then query the tables as if they were native Databend objects.

Quick Start

1. Register the Hive Metastore

   CREATE CATALOG hive_prod
   TYPE = HIVE
   CONNECTION = (
     METASTORE_ADDRESS = '127.0.0.1:9083'
     URL = 's3://lakehouse/'
     ACCESS_KEY_ID = '<your_key_id>'
     SECRET_ACCESS_KEY = '<your_secret_key>'
   );

2. Explore the catalog

   USE CATALOG hive_prod;
   SHOW DATABASES;
   SHOW TABLES FROM tpch;

3. Query Hive tables

   SELECT l_orderkey, SUM(l_extendedprice) AS revenue
   FROM tpch.lineitem
   GROUP BY l_orderkey
   ORDER BY revenue DESC
   LIMIT 10;

Keep Metadata Fresh

Hive schemas or partitions can change outside of Databend. Refresh Databend’s cached metadata when that happens:

ALTER TABLE tpch.lineitem REFRESH CACHE;

Data Type Mapping

Databend automatically converts Hive primitive types to their closest native equivalents when queries run:

Hive Type	Databend Type
BOOLEAN	BOOLEAN
TINYINT, SMALLINT, INT, BIGINT	Integer types
FLOAT, DOUBLE	Floating-point types
DECIMAL(p,s)	DECIMAL
STRING, VARCHAR, CHAR	STRING
DATE, TIMESTAMP	DATETIME
ARRAY<type>	ARRAY
MAP<key,value>	MAP

Nested structures such as STRUCT are surfaced through the VARIANT type.

Notes and Limitations

• Hive catalogs are read-only in Databend (writes must happen through Hive-compatible engines).
• Access to the underlying object storage is required; configure credentials by using connection parameters.
• Use ALTER TABLE ... REFRESH CACHE whenever table layout changes (for example, new partitions) to keep query results up to date.