feat: Implement state conversion for remaining group accumulators - #23275
Merged
Conversation
kosiew
approved these changes
Jul 9, 2026
| } | ||
|
|
||
| #[test] | ||
| fn convert_to_state_roundtrips_through_merge() -> Result<()> { |
Contributor
There was a problem hiding this comment.
Nice, these new unit tests directly cover convert_to_state -> merge_batch, which is great to see.
One small follow-up idea: it may be worth adding a SQLLogicTest case in aggregate_skip_partial.slt for the newly enabled public skip-partial path for variance, stddev, and corr. Including a FILTER or NULL row would help catch wiring regressions outside the accumulator unit tests. COUNT(DISTINCT) could be included too if it is practical.
This is non-blocking since the accumulator-level roundtrip tests already exercise the core state conversion logic.
Contributor
Author
There was a problem hiding this comment.
Thanks @kosiew, I added SLT coverage in aggregate_skip_partial.slt for the skip-partial sql path. The new cases cover variance/stddev, corr, and count distinct with nullable inputs and FILTER.
I also compared the same queries against the non-skip path locally and confirmed the results match.
Contributor
lyne7-sc
added a commit
to lyne7-sc/datafusion
that referenced
this pull request
Jul 24, 2026
…nvert_to_state` (apache#23489) ## Which issue does this PR close? <!-- We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes apache#123` indicates that this PR will close issue apache#123. --> - Closes apache#23081. ## Rationale for this change <!-- Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed. Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes. --> Following apache#23275, all `GroupsAccumulator` implementations now provide `convert_to_state`. The `supports_convert_to_state` capability flag is therefore no longer needed. ## What changes are included in this PR? <!-- There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR. --> - Make `GroupsAccumulator::convert_to_state` a required trait method. - Remove `GroupsAccumulator::supports_convert_to_state` and its implementations. - Remove the corresponding capability checks from hash aggregation. - Simplify skip-partial aggregation to use the required `convert_to_state` implementation directly. - Add a regression test covering the partial hash aggregation skip path. - Document the breaking trait change in the 55.0.0 upgrading guide. - Remove `FFI_GroupsAccumulator::supports_convert_to_state`. This changes the FFI ABI layout, so providers and consumers must be rebuilt against DataFusion 55. ## Are these changes tested? <!-- We typically require tests for all PRs in order to: 1. Prevent the code from being accidentally broken by subsequent changes 2. Serve as another way to document the expected behavior of the code If tests are not included in your PR, please explain why (for example, are they covered by existing tests)? --> Yes. Added a regression test verifying that skip-partial aggregation uses the required `convert_to_state` implementation without a capability flag. Existing physical-plan and FFI tests continue to pass. ## Are there any user-facing changes? <!-- If there are user-facing changes then we may require documentation to be updated before approving the PR. --> <!-- If there are any breaking changes to public APIs, please add the `api change` label. --> Yes. This is a breaking Rust API change for external `GroupsAccumulator` implementations: - `convert_to_state` must now be implemented. - `supports_convert_to_state` should be removed. The migration is documented in the 55.0.0 upgrading guide. The `FFI_GroupsAccumulator` layout has changed. FFI providers and consumers must be rebuilt against DataFusion 55 and must not exchange this struct with older major versions.
kosiew
pushed a commit
to kosiew/datafusion
that referenced
this pull request
Aug 12, 2026
…nvert_to_state` (apache#23489) ## Which issue does this PR close? <!-- We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes apache#123` indicates that this PR will close issue apache#123. --> - Closes apache#23081. ## Rationale for this change <!-- Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed. Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes. --> Following apache#23275, all `GroupsAccumulator` implementations now provide `convert_to_state`. The `supports_convert_to_state` capability flag is therefore no longer needed. ## What changes are included in this PR? <!-- There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR. --> - Make `GroupsAccumulator::convert_to_state` a required trait method. - Remove `GroupsAccumulator::supports_convert_to_state` and its implementations. - Remove the corresponding capability checks from hash aggregation. - Simplify skip-partial aggregation to use the required `convert_to_state` implementation directly. - Add a regression test covering the partial hash aggregation skip path. - Document the breaking trait change in the 55.0.0 upgrading guide. - Remove `FFI_GroupsAccumulator::supports_convert_to_state`. This changes the FFI ABI layout, so providers and consumers must be rebuilt against DataFusion 55. ## Are these changes tested? <!-- We typically require tests for all PRs in order to: 1. Prevent the code from being accidentally broken by subsequent changes 2. Serve as another way to document the expected behavior of the code If tests are not included in your PR, please explain why (for example, are they covered by existing tests)? --> Yes. Added a regression test verifying that skip-partial aggregation uses the required `convert_to_state` implementation without a capability flag. Existing physical-plan and FFI tests continue to pass. ## Are there any user-facing changes? <!-- If there are user-facing changes then we may require documentation to be updated before approving the PR. --> <!-- If there are any breaking changes to public APIs, please add the `api change` label. --> Yes. This is a breaking Rust API change for external `GroupsAccumulator` implementations: - `convert_to_state` must now be implemented. - `supports_convert_to_state` should be removed. The migration is documented in the 55.0.0 upgrading guide. The `FFI_GroupsAccumulator` layout has changed. FFI providers and consumers must be rebuilt against DataFusion 55 and must not exchange this struct with older major versions.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
GroupsAccumulator::supports_convert_to_stateand makeconvert_to_statemandatory #23081.Rationale for this change
#23081 proposes making
GroupsAccumulator::convert_to_statemandatory and removingsupports_convert_to_state.Before that API cleanup can happen, all existing
GroupsAccumulatorimplementations need a correctconvert_to_stateimplementation.This PR is the first step toward that cleanup: it fills in missing
convert_to_stateimplementations for the remaining group accumulators that currently participate in grouped aggregation.What changes are included in this PR?
This PR adds
convert_to_statesupport for:PrimitiveDistinctCountGroupsAccumulatorHllGroupsAccumulatorforapprox_distinctVarianceGroupsAccumulatorStddevGroupsAccumulatorCorrelationGroupsAccumulatorGeometricMeanGroupsAccumulatorin the advanced UDAF exampleAre these changes tested?
Yes. This PR adds unit tests for the new
convert_to_stateimplementations.Are there any user-facing changes?
No. This PR does not remove
supports_convert_to_stateyet and does not change the publicGroupsAccumulatorAPI.