GH-36283: [C++] Fix predicate pushdown for nullable columns with range statistics #48716
+195
−0
Conversation
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
…h range statistics Prior to this change, predicate pushdown was broken for nullable columns with range statistics (min != max). Parquet generates guarantees of the form: or_(and_(field >= min, field <= max), is_null(field)) However, Inequality::ExtractOne() only handled single comparisons inside or_(..., is_null), not and_(...) expressions. This meant no inequalities were extracted and SimplifyWithGuarantee() could not simplify predicates. This affected ALL predicates on nullable columns: - Comparisons: equal, less, greater, less_equal, greater_equal - Set membership: is_in The fix adds ExpandNullableRangeGuarantees() which transforms: or_(and_(A, B), is_null(x)) into: [or_(A, is_null(x)), or_(B, is_null(x))] This expansion is valid because (A AND B) OR C ≡ (A OR C) AND (B OR C). Each expanded guarantee can then be processed by existing simplification logic. Added tests for both comparison operators and is_in with nullable range guarantees.
|
|
|
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Rationale for this change
Predicate pushdown for Parquet was broken for nullable columns with range statistics (min ≠ max),
which is the vast majority of real-world data. This caused row groups to be read even when predicates
could definitively exclude them.
The root cause: Parquet statistics for nullable columns generate guarantees of the form:
or_(and_(field >= min, field <= max), is_null(field))
However, Inequality::ExtractOne() only handled single comparisons inside or_(..., is_null), not
and_(...) expressions. This meant no inequalities were extracted and SimplifyWithGuarantee() could not
simplify predicates.
This affected all predicates on nullable columns:
See: #36283
What changes are included in this PR?
Added ExpandNullableRangeGuarantees() which transforms:
or_(and_(A, B), is_null(x))
into:
[or_(A, is_null(x)), or_(B, is_null(x))]
This expansion is logically valid because (A ∧ B) ∨ C ≡ (A ∨ C) ∧ (B ∨ C). Each expanded guarantee can
then be processed by existing simplification logic.
Also handles the reversed form or_(is_null(x), and_(...)).
Are these changes tested?
Yes. Added two new test cases:
Both tests fail without the fix and pass with it.
Are there any user-facing changes?
No API changes. Users will see improved query performance when filtering nullable columns in Parquet
files, as row groups can now be correctly skipped based on min/max statistics.
Fixes: [Bug] Pyiceberg row filter expression "In" takes longer to query than using "EqualTo" iceberg-python#1295