Great Expectations 1.18.2 - Spark Connect Metric Fixes

Great Expectations 1.18.2 was published on June 26, 2026, with a narrow fix for Spark Connect compatibility in distinct values metrics. The release removes .rdd usage from Spark metric code, which matters when validation suites run against Spark Connect instead of a classic Spark driver path. It is not marked as a prerelease.

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

The main bug fix is in the Spark distinct values metric path. PR #11922 fixes .rdd usage so that validation can work with Spark Connect.

For data teams, the practical issue is compatibility. Spark Connect changes how client code talks to Spark, and direct access patterns that assume the old driver side object model can fail before a metric result is computed.

Distinct values checks are common in data quality suites because they sit close to enum checks, partition sanity checks, and column profile jobs. A small API assumption there can block a larger batch validation run. This release is therefore more about execution path correctness than new validation surface area.

The scope is also useful to keep in view. The note only mentions Spark distinct values metrics. It does not claim wider Spark Connect coverage, and it does not describe new expectations or new Spark execution options.

The notes also call out PR #11924, which fixes a BigQuery Python 3.13 collection error from a NumPy generic unit deprecation warning.

That is maintenance, but it is relevant for platform teams that test validation stacks on newer Python runtimes before moving scheduled jobs. Data quality jobs often fail on dependency edges before they reach the actual data rule. Python version moves make those edges visible.

The release notes do not describe a new BigQuery feature. Read this as compatibility cleanup for environments that are already moving toward Python 3.13, especially where warehouse backed checkpoints are part of CI or scheduled validation.

PR #11921 fixes a pytest parametrize deprecation that was breaking scheduled CI. That does not change runtime behavior for validation jobs, but it matters for release health because broken scheduled tests make small fixes harder to trust.

Three dependency bumps are limited to /docs/docusaurus: http-proxy-middleware moves to 2.0.10, webpack-dev-server moves to 5.2.5, and @babel/core moves to 7.29.6. Those are documentation site dependencies, not data runtime dependencies.

For operators, this means the release is not a broad dependency refresh of the validation engine. Treat the docs changes as supply chain hygiene around the documentation site, while the Spark and BigQuery fixes are the runtime items worth checking.

If your Great Expectations deployment runs Spark validations through Spark Connect, prioritize a focused validation job that hits distinct values metrics. That is the clearest operational path covered by the GitHub release page.

Teams with local custom expectations should remember that this release note covers project code, not local code that might call .rdd directly. If a custom Spark expectation still depends on that access pattern, this patch will not remove that local assumption.

For BigQuery users, include BigQuery backed checkpoints in Python 3.13 runtime checks. The release note is small, but it points at the kind of version skew between Python, NumPy, and warehouse connectors that can stop validation before data logic runs.

• Release page: Great Expectations 1.18.2
• Project repo: great-expectations/great_expectations
• Tag: 1.18.2