Snowflake Connector for Spark release notes for 2022¶

This article contains the release notes for the Snowflake Connector for Spark, including the following when applicable:

Snowflake uses semantic versioning for Snowflake Connector for Spark updates.

Version 2.11.1 (December 13, 2022)¶

Added support for AWS VPCE deployments by adding the S3_STAGE_VPCE_DNS_NAME configuration parameter to specifying the VPCE DNS name at the session level.
Added a new configuration option treat_decimal_as_long to enable the Spark Connector to return Long values instead of BigDecimal values, if the query returns Decimal(<any_precision>, 0). WARNING: If the value is greater than the maximum value of Long, an error will be raised.
Added a new option proxy_protocol for specifying the proxy protocol (http or https) with AWS deployments. (The option has no effect on Azure and GCP deployments.).
Added support for counting rows in a table where the row count is greater than the maximum value of Integer.
Updated the connector to use the Snowflake JDBC driver 3.13.24.

Compatible JDBC Driver version: 3.13.22

Added support for Spark 3.3 and fixed some bugs:
- Upgraded the version of the PostgreSQL JDBC Driver that tests depend on to avoid the security vulnerability CVE-2022-31197.
- Updated the connector to use the Snowflake JDBC driver 3.13.22 and the Snowflake Ingest SDK 0.10.8.

Note

Starting from version 2.11.0, the Snowflake Connector for Spark supports Spark 3.1, 3.2 and 3.3. Version 2.11.0 of the Snowflake Connector for Spark does not support Spark 3.0. Note that previous versions of the connector continue to support Spark 3.0.
For Snowflake GCP accounts, the Snowflake JDBC driver versions 3.13.16 through 3.13.21 do not work with the Spark connector.

Compatible JDBC Driver version: 3.13.14

Removed unnecessary dependencies on libraries to avoid the security vulnerabilities CVE-2020-8908 and CVE-2018-10237.
Added support for using the JDBC data type TIMESTAMP_WITH_TIMEZONE when reading data from Snowflake.
Changed the logic for checking for the existence of a table before saving a DataFrame to Snowflake:
- The connector now reuses the existing connection (rather than creating a new connection) to avoid potential problems with token expiration.
- If the table name is not fully qualified (i.e. does not include the schema name), the connector now checks for the table under the schema specified by sfSchema, rather than the schema that is currently in use in the session.
  
  Note
  
  If you need to save a DataFrame to a table in a schema other than sfSchema, specify the schema as part of the fully qualified name of the table, rather than executing USE SCHEMA to change the current schema.
Improved performance by avoiding unnecessary parse_json() calls in the COPY INTO TABLE command when writing a DataFrame with ArrayType, MapType, or StructType columns to Snowflake.
Added the getLastSelectQueryId and getLastCopyLoadQueryId methods to the Utils class. These methods return the query ID of the last query that read data from Snowflake and the last COPY INTO TABLE statement that was executed (respectively).

Compatible JDBC Driver version: 3.13.14

Added support for Spark, version 3.2. Beginning with this release, the Snowflake Connector for Spark only supports Spark 3.0, 3.1 and 3.2. Spark version 2.4 is no longer supported.