9.21 Release notes: Jul 29, 2025-Aug 01, 2025

Attention

This release has been completed. For differences between the in-advance and final versions of these release notes, see the Release notes change log.

Security updates

GENERATE_SYNTHETIC_DATA: Consistency secret now optional in most cases

Previously, when you called GENERATE_SYNTHETIC_DATA with a replace column property, you needed to provide a SECRET for consistency_secret. With this change, consistency_secret is now optional. However, if you run GENERATE_SYNTHETIC_DATA in an owner’s rights stored procedure, you still must provide a value to consistency_secret.

SQL updates

Account Usage: TABLE_QUERY_PRUNING_HISTORY and COLUMN_QUERY_PRUNING_HISTORY views (General availability)

You can monitor data access patterns at the table and column level by querying two new Account Usage views:

The SEARCH_IP function supports searching for IPv6 addresses

You can use the SEARCH_IP function to search for IPv6 addresses in data. Previously, the function only supported searching for IPv4 addresses.

For more information, see SEARCH_IP.

Generating YAML for a semantic view and creating a semantic view from YAML

To generate the YAML specification for a semantic view, you can call the SYSTEM$READ_YAML_FROM_SEMANTIC_VIEW function. For example:

SELECT SYSTEM$READ_YAML_FROM_SEMANTIC_VIEW(
  'my_db.my_schema.tpch_rev_analysis'
);
Copy

You can also create a semantic view from a YAML specification by calling the SYSTEM$CREATE_SEMANTIC_VIEW_FROM_YAML stored procedure. For example:

CALL SYSTEM$CREATE_SEMANTIC_VIEW_FROM_YAML(
  'my_db.my_schema',
  $$
  name: TPCH_REV_ANALYSIS
  description: Semantic view for revenue analysis
  ...
  $$
);
Copy

For information, see:

Data pipeline updates

Snowpark Connect for Spark and Snowpark Submit (Preview)

With Snowpark Connect for Spark, you can run Spark DataFrame, SQL, and UDF APIs directly on the Snowflake platform using the same Spark code you use today. You can develop using client tools such as Snowflake Notebooks, Jupyter Notebooks, and others. With Snowpark Submit, you can run Spark workloads in a non-interactive, asynchronous way directly on Snowflake’s infrastructure while you use familiar Spark semantics.

Snowpark Connect for Spark and Snowpark Submit are in Preview.

For more information, see Run Spark workloads on Snowflake with Snowpark Connect for Spark.

Release notes change log

Announcement

Update

Date

Release notes

Initial publication (preview)

Jul 25, 2025

Creating semantic views from YAML and reading YAML for semantic views

Added to SQL updates

Jul 29, 2025

Tracing SQL statements run from handler code (General availability)

Added to Extensibility updates

Aug 01, 2025

Tracing SQL statements run from handler code (General availability)

Moved to 9.22 release notes

Aug 06, 2025