Delta Lake Connector¶

Overview¶

This connector allows reading Delta Lake tables in Presto. The connector uses the Delta Standalone Library (DSR) provided by Delta Lake project to read the table metadata.

Configuration¶

To configure the Delta Lake connector, create a catalog properties file etc/catalog/delta.properties with the following contents, replacing the properties as appropriate:

connector.name=delta
hive.metastore.uri=hostname:port

Configuration Properties¶

The following configuration properties are available:

Property Name	Description	Default
`hive.metastore.uri`	The URI(s) of the Hive metastore where Delta Lake tables are registered. The connector relies on the Hive metastore to find the location of Delta Lake tables. From the Delta Log at given location, schema and data file list of the table is found.	`null`
`delta.parquet-dereference-pushdown-enabled`	Enable pushing nested column dereferences into table scan so that only the required fields selected in a `struct` data type column are selected. In order for this option to work, also set `experimental.pushdown-dereference-enabled` to `true`.	`true`
`delta.case-sensitive-partitions-enabled`	Allows matching the names of partitioned columns in a case-sensitive manner.	`true`

Delta Lake connector reuses many of the modules existing in Hive connector. Modules for connectivity and security such as S3, Azure Data Lake, Glue metastore etc. So the configurations for these modules is same those available in Hive connector documentation.

Querying Delta Lake Tables¶

Example query

SELECT * FROM sales.apac.sales_data LIMIT 200;

In the above query

sales refers to the Delta Lake catalog.
apac refers to the database in Hive metastore.
sales_data refers to the Delta Lake table registered in the apac database.

If the table is not registered in Hive metastore, it can be registered using the following DDL command.

Note

To register a table in Hive metastore, full schema of the table is not required in DDL as the Delta Lake connector gets the schema from the metadata located at the Delta Lake table location. To get around no columns error in Hive metastore, provide a dummy column as schema of the Delta table being registered.

Examples¶

Create a new Delta table named sales_data_new in the apac schema that has Delta Lake table location in an S3 bucket named db-sa-datasets using Delta Lake connector:

CREATE TABLE sales.apac.sales_data_new (dummyColumn INT)
WITH (external_location = 's3://db-sa-datasets/presto/sales_data_new');

To register a partition Delta table in Hive metastore, use the CREATE TABLE same as above. Only external_location is required in the properties, no need to specify partitioned_by in CREATE TABLE

Another option is querying the table directly using the table location as table name.

SELECT * FROM sales."$path$"."s3://db-sa-datasets/presto/sales_data" LIMIT 200;

In the above query the schema $path$ indicates the table name is a path. Table name given as s3://db-sa-datasets/presto/sales_date is a path where the Delta Lake table is located. The path based option allows users to query a Delta table without registering it in the Hive metastore.

To query a specific snapshot of the Delta Lake table use the snapshot identifier as suffix to the table name.

SELECT * FROM sales.apac."sales_data@v4" LIMIT 200;

Above query reads data from snapshot version 4 of the table sales.apac.sales_data.

To query the snapshot of the Delta Lake table as of particular time, specify the timestamp as suffix to the table name.

SELECT * FROM sales.apac."sales_data@t2021-11-18 09:45" LIMIT 200;

Above query reads data from the latest snapshot as of timestamp 2021-11-18 09:45:00 in the table sales.apac.sales_data.

DROP TABLE sales.apac.sales_data_new;

Above query drops the external table sales.apac.sales_data_new. This only drops the metadata for the table. The referenced data directory is not deleted.