[SPARK-56081][PS] Align idxmax and idxmin NA handling with pandas 3#54908
Closed
ueshin wants to merge 4 commits intoapache:masterfrom
Closed
[SPARK-56081][PS] Align idxmax and idxmin NA handling with pandas 3#54908ueshin wants to merge 4 commits intoapache:masterfrom
ueshin wants to merge 4 commits intoapache:masterfrom
Conversation
Member
Author
HyukjinKwon
approved these changes
Mar 19, 2026
Member
|
Merged to master. |
terana
pushed a commit
to terana/spark
that referenced
this pull request
Mar 23, 2026
### What changes were proposed in this pull request? This PR updates pandas-on-Spark `idxmax` and `idxmin` behavior to follow the pandas 3 semantics when NA values are involved. In `python/pyspark/pandas/series.py`, `Series.idxmax(skipna=False)` and `Series.idxmin(skipna=False)` now raise `ValueError` on pandas 3 instead of returning `np.nan` with a `FutureWarning`, while preserving the existing behavior on pandas versions below 3.0. In `python/pyspark/pandas/frame.py`, `DataFrame.idxmax(axis=1)` and `DataFrame.idxmin(axis=1)` now raise when pandas 3 would reject rows with all-NA values, instead of silently materializing `None`. The related pandas-on-Spark tests were updated to assert the version-specific behavior in: - `python/pyspark/pandas/tests/computation/test_idxmax_idxmin.py` - `python/pyspark/pandas/tests/series/test_index.py` ### Why are the changes needed? pandas 3 changed the NA handling for `idxmax` and `idxmin`. The existing pandas-on-Spark implementation still followed the older behavior in these paths, which caused mismatches against pandas 3 expectations. Aligning these code paths keeps pandas-on-Spark behavior consistent with the pandas version it is running against and makes the failure mode explicit instead of returning a deprecated result. ### Does this PR introduce _any_ user-facing change? Yes. When pandas 3 is used: - `Series.idxmax(skipna=False)` and `Series.idxmin(skipna=False)` now raise `ValueError` when an NA is encountered instead of returning `np.nan`. - `DataFrame.idxmax(axis=1)` and `DataFrame.idxmin(axis=1)` now raise on rows with all NA values instead of returning a null result. For pandas versions below 3.0, the existing behavior is preserved. ### How was this patch tested? Updated the related tests for the pandas-version-specific `idxmax` and `idxmin` behavior. ### Was this patch authored or co-authored using generative AI tooling? No. Closes apache#54908 from ueshin/issues/SPARK-56081/idxmax_idxmin. Authored-by: Takuya Ueshin <ueshin@databricks.com> Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
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.
What changes were proposed in this pull request?
This PR updates pandas-on-Spark
idxmaxandidxminbehavior to follow the pandas 3 semantics when NA values are involved.In
python/pyspark/pandas/series.py,Series.idxmax(skipna=False)andSeries.idxmin(skipna=False)now raiseValueErroron pandas 3 instead of returningnp.nanwith aFutureWarning, while preserving the existing behavior on pandas versions below 3.0.In
python/pyspark/pandas/frame.py,DataFrame.idxmax(axis=1)andDataFrame.idxmin(axis=1)now raise when pandas 3 would reject rows with all-NA values, instead of silently materializingNone.The related pandas-on-Spark tests were updated to assert the version-specific behavior in:
python/pyspark/pandas/tests/computation/test_idxmax_idxmin.pypython/pyspark/pandas/tests/series/test_index.pyWhy are the changes needed?
pandas 3 changed the NA handling for
idxmaxandidxmin. The existing pandas-on-Spark implementation still followed the older behavior in these paths, which caused mismatches against pandas 3 expectations.Aligning these code paths keeps pandas-on-Spark behavior consistent with the pandas version it is running against and makes the failure mode explicit instead of returning a deprecated result.
Does this PR introduce any user-facing change?
Yes.
When pandas 3 is used:
Series.idxmax(skipna=False)andSeries.idxmin(skipna=False)now raiseValueErrorwhen an NA is encountered instead of returningnp.nan.DataFrame.idxmax(axis=1)andDataFrame.idxmin(axis=1)now raise on rows with all NA values instead of returning a null result.For pandas versions below 3.0, the existing behavior is preserved.
How was this patch tested?
Updated the related tests for the pandas-version-specific
idxmaxandidxminbehavior.Was this patch authored or co-authored using generative AI tooling?
No.