TPCH connector#

The TPCH connector provides a set of schemas to support the TPC Benchmark™ H (TPC-H). TPC-H is a database benchmark used to measure the performance of highly-complex decision support databases.

This connector can be used to test the capabilities and query syntax of Trino without configuring access to an external data source. When you query a TPCH schema, the connector generates the data on the fly using a deterministic algorithm.

Configuration#

To configure the TPCH connector, create a catalog properties file etc/catalog/example.properties with the following contents:

connector.name=tpch

TPCH schemas#

The TPCH connector supplies several schemas:

SHOW SCHEMAS FROM example;
       Schema
--------------------
 information_schema
 sf1
 sf100
 sf1000
 sf10000
 sf100000
 sf300
 sf3000
 sf30000
 tiny
(11 rows)

Ignore the standard schema information_schema, which exists in every catalog, and is not directly provided by the TPCH connector.

Every TPCH schema provides the same set of tables. Some tables are identical in all schemas. Other tables vary based on the scale factor, which is determined based on the schema name. For example, the schema sf1 corresponds to scale factor 1 and the schema sf300 corresponds to scale factor 300. The TPCH connector provides an infinite number of schemas for any scale factor, not just the few common ones listed by SHOW SCHEMAS. The tiny schema is an alias for scale factor 0.01, which is a very small data set useful for testing.

Type mapping#

Trino supports all data types used within the TPCH schemas so no mapping is required.

SQL support#

The connector provides globally available and read operation statements to access data and metadata in the TPC-H dataset.