release-22.2: sql: fix ambigious udf overload with null arguments #109193
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewable status:
complete! 0 of 0 LGTMs obtained (waiting on @mgartner and @michae2)
pkg/sql/logictest/testdata/logic_test/udf line 2849 at r1 (raw file):
statement ok CREATE FUNCTION f88374(i TEXT) RETURNS INT STRICT LANGUAGE SQL AS 'SELECT 2';
Note: I couldn't replicate the original example using INT2 and INT4 in the issue in 22.2 due to an ERROR: function "f88374" already exists with same argument types which is outside the scope of this PR. However, this patch fixes other types.
|
Gentle ping for a review on this backport! |
mgartner
changed the title
There was a problem hiding this comment.
Nice job tracking down the change that fixed this!
For backports (even when they are non-standard like this), remember to prefix PR titles with release-XX.X: to denote that it's not being merged to the master branch. I added the prefix to this PR.
Reviewed 3 of 3 files at r1, all commit messages.
Reviewable status:
-- commits line 4 at r1:
nit: Can you link to the PR?
pkg/sql/logictest/testdata/logic_test/udf line 2849 at r1 (raw file):
Previously, rharding6373 (Rachael Harding) wrote…
Note: I couldn't replicate the original example using INT2 and INT4 in the issue in 22.2 due to an
ERROR: function "f88374" already exists with same argument typeswhich is outside the scope of this PR. However, this patch fixes other types.
That's fine. We never backported the fix (#96481) that allows multiple UDFs with non-identical types in the same family - which is why you're getting that error in 22.2.
This change backports a small change from a larger PR (cockroachdb#98162) that prevents short-circuiting type checking for UDF function overloads. This allows us to do more precise type checking even if inputs are NULL. Epic: none Fixes: cockroachdb#88374 Release note (bug fix): Fixes ambiguous calls to UDFs with NULL arguments.
rharding6373
force-pushed
the
backport22.2-98162
branch
from
d511aef to
4d7382e
Compare
There was a problem hiding this comment.
Thanks for that.
Reviewable status:
Previously, mgartner (Marcus Gartner) wrote…
nit: Can you link to the PR?
Done.
There was a problem hiding this comment.
Add the "larger PR" number to the issue description too, if you'd like.
Reviewable status:
|
TFTR! |
This change backports a small change from a larger PR (#98162) that prevents short-circuiting type checking for UDF function overloads. This allows us to do more precise type checking even if inputs are NULL.
Epic: none
Fixes: #88374
Release note (bug fix): Fixes ambiguous calls to UDFs with NULL arguments.
Release justification: Fixes a bug referencing UDF overloads.