Fix PartialOrd for ScalarUDF #17182
Merged
Fix PartialOrd for ScalarUDF #17182
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
findepi
force-pushed
the
findepi/fix-partialord-for-scalarudf-dfb3ab
branch
from
7ab8252 to
2093f4c
Compare
Before the changes, `PartialOrd` could return `Some(Equal)` for two functions that are not equal in `PartialEq` sense. This is violation of `PartialOrd` contract. This was possible e.g. when - two functions have same name, but are of different types (e.g. when someone constructs DataFusion LogicalPlan and mixes DataFusion builtin functions with their own) - a function has a parameter (e.g. "safe" attribute for a CAST). The parameter is honored by `PartialEq` comparison but was transparent to `PartialOrd` ordering. The fix is to consult eq inside ord implementation. If ord thinks two instances are equal, but they are not equal in Eq sense, they are considered incomparable.
findepi
force-pushed
the
findepi/fix-partialord-for-scalarudf-dfb3ab
branch
from
2093f4c to
98c404f
Compare
Member
Author
alamb
reviewed
Aug 18, 2025
| let mut cmp = self.name().cmp(other.name()); | ||
| if cmp == Ordering::Equal { | ||
| cmp = self.signature().partial_cmp(other.signature())?; | ||
| } |
Contributor
There was a problem hiding this comment.
I think we could short circut the comparison if the ordering is not equal
Something like this, which should avoid computation in many cases
if cmp != Ordering::Equal {
return cmp
}
Member
Author
There was a problem hiding this comment.
That would make debug_assert below unreachable. Do you want me to delete it?
Contributor
There was a problem hiding this comment.
No, I think the debug assert is good. Sorry for not being clear
alamb
approved these changes
Aug 18, 2025
Contributor
alamb
left a comment
alamb
left a comment
There was a problem hiding this comment.
I thought about it some more and this PR already does short circuiting most likely (returning early would for sure avoid extra comparisions, but the compiler probably does the same thing with the way this PR is structured too)
Contributor
|
🤖 |
Contributor
|
🤖: Benchmark completed Details
|
Contributor
|
Look s good to me -- thanks @findepi |
This was referenced Sep 3, 2025
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.
Before the changes,
PartialOrdcould returnSome(Equal)for twofunctions that are not equal in
PartialEqsense. This is violation ofPartialOrdcontract. This was possible e.g. whensomeone constructs DataFusion LogicalPlan and mixes DataFusion builtin
functions with their own)
parameter is honored by
PartialEqcomparison but was transparent toPartialOrdordering.The fix is to consult eq inside ord implementation. If ord thinks two
instances are equal, but they are not equal in Eq sense, they are
considered incomparable.
PartialOrdforScalarUDF,WindowUDFandAggregateUDF#17064