[fix](be) Return NaN for avg_weighted when sum of weights is zero#64333
Open
jacktengg wants to merge 1 commit into
Open
[fix](be) Return NaN for avg_weighted when sum of weights is zero#64333jacktengg wants to merge 1 commit into
jacktengg wants to merge 1 commit into
Conversation
Issue Number: close #xxx Problem Summary: When the sum of weights passed to avg_weighted is zero but the weighted data sum is non-zero, the function computed data_sum / weight_sum and returned +/-Infinity (e.g. avg_weighted(f1, f2) over rows (1, 1), (2, -1) returned -Infinity). The expected mathematical result of a division by zero weight is undefined, so it should return NaN. This fixes AggregateFunctionAvgWeightedData::get() to return quiet_NaN() when weight_sum is exactly zero, instead of producing Infinity.
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
Author
|
/review |
Contributor
Author
|
run buildall |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Review result: no blocking code issues found.
Critical checkpoint conclusions:
- Goal and proof: the PR changes avg_weighted finalization so a non-empty aggregate state with accumulated weight_sum == 0 returns quiet NaN instead of +/-Infinity. The added regression case covers weights 1 and -1 and expects NaN.
- Scope: the implementation is small and focused: one include, one final-result branch, and one regression case.
- Concurrency and lifecycle: not applicable. This only reads aggregate state during result insertion; no shared state, locks, static initialization dependency, or lifecycle ownership is introduced.
- Config and compatibility: no config, storage format, function signature, serialized aggregate state, or FE/BE protocol change. Existing serialized state remains data_sum + weight_sum.
- Parallel paths: I checked the registration/combinator path. avg_weighted and foreach wrappers use this same nested implementation, so I did not find another result path that still returns infinity for the same zero-weight state.
- Conditions and error handling: the new weight_sum == 0 condition has a concrete semantic trigger and does not suppress errors or ignore Status values.
- Test coverage and result correctness: the added regression follows the existing suite style, drops the table before use, uses a hardcoded table name, and the expected NaN spelling matches existing floating-point output conventions. git diff --check is clean.
- Observability and performance: no new observability is needed for this finalization-only behavior. The extra branch is outside accumulation/merge loops and should be negligible.
- Transactions, persistence, data writes, and FE-BE variables: not applicable.
- User focus: no additional user-provided focus points were present.
I did not run a local Doris regression cluster. Current CI has pending BE/performance checks; the macOS BE UT failure I inspected stops before build with "ERROR: The JAVA version is 25, it must be JDK-17", which appears unrelated to this change.
Minor process note: the PR body still says Release note: None even though this is a user-visible bug fix. I am not blocking on that, but it may be worth aligning before merge.
Contributor
TPC-H: Total hot run time: 29181 ms |
Contributor
TPC-DS: Total hot run time: 169314 ms |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
BE Regression && UT Coverage ReportIncrement line coverage Increment coverage report
|
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
2 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.
Issue Number: close #xxx
Problem Summary: When the sum of weights passed to avg_weighted is zero but the weighted data sum is non-zero, the function computed data_sum / weight_sum and returned +/-Infinity (e.g. avg_weighted(f1, f2) over rows (1, 1), (2, -1) returned -Infinity). The expected mathematical result of a division by zero weight is undefined, so it should return NaN.
This fixes AggregateFunctionAvgWeightedData::get() to return quiet_NaN() when weight_sum is exactly zero, instead of producing Infinity.
What problem does this PR solve?
Issue Number: close #xxx
Related PR: #xxx
Problem Summary:
Release note
None
Check List (For Author)
Test
Behavior changed:
Does this need documentation?
Check List (For Reviewer who merge this PR)