Great Expectations 1.19.0 - Spark 4 And SQL Validation Fixes

Great Expectations 1.19.0 was published on July 13, 2026, with Spark 4 support as its main operator visible change. The release also corrects SQL distinct value handling for strings that resemble dates and rejects an empty regex list in a column expectation.

The full release notes and downloads are on the GitHub release page.

Spark 4 support is the largest runtime change in this release. Teams moving validation workloads to Spark 4 now have an explicitly supported Great Expectations release to include in their compatibility testing.

The release notes do not name new Spark configuration keys, migration steps, or changes to expectation semantics. This is a compatibility addition, not evidence that every existing validation suite will produce identical results on a new Spark runtime. Platform teams should test Spark and Great Expectations together with representative data, especially where SQL expressions, null handling, and type conversion affect expectation results.

That limited scope is useful when planning an upgrade. The Great Expectations change can be evaluated as part of the Spark 4 runtime move without assuming that this release changes how validation jobs are configured.

The SQL fix preserves strings that resemble dates when they appear in distinct value sets. The affected path is specific: SQL distinct values. Pipelines that compare these values with an expected text set should no longer lose the original string representation merely because a value looks like a date.

This matters for identifiers, partition labels, and imported source fields where date shaped text is still text. A validation result should reflect the stored value and type assumptions used by the expectation, rather than an inferred calendar value. Existing checks around this edge case are worth rerunning after the upgrade so that any previous workaround can be identified and reviewed.

The second validation fix changes ExpectColumnValuesToMatchRegexList. It now rejects an empty regex_list. An empty list does not express a useful matching rule, so rejecting it makes bad configuration visible before it can be treated as a meaningful expectation. Systems that generate expectation configuration should check list cardinality and decide whether an empty source list means an error, a skipped check, or a separately defined condition.

Two maintenance items improve the project test setup rather than the runtime API. Snowflake connection tests now provision their own data. Databricks CI now authenticates through an OAuth M2M service principal.

For maintainers, self provisioned Snowflake data reduces reliance on manually prepared test state. Service principal authentication gives Databricks CI a machine identity suited to automated runs. The release notes present both changes as CI maintenance. They do not announce a required Snowflake setup change or a Databricks authentication migration for user deployments.

The remaining release work is largely internal refactoring, CI cleanup, dependency maintenance, contributor workflow, and documentation repair. The notes do not claim a separate pipeline feature from that work.