fix: preserve UPPER and LOWER input nullability

[osipovartem]

Which issue does this PR close?

• Closes UPPER and LOWER always report nullable result fields #25380.

Rationale for this change

UPPER and LOWER currently use the default scalar UDF result field, which is always nullable. Both functions return a non-null value for a non-null input, so this loses a provable schema property and exposes inaccurate result metadata to downstream consumers.

What changes are included in this PR?

• Implement return_field_from_args for UpperFunc and LowerFunc.
• Preserve the input string type and nullability.
• Add nullable and non-nullable unit cases for each function.

The vectorized execution kernels are unchanged. This adds no per-row dispatch, conversion, allocation, or materialization.

Are these changes tested?

Yes:

• cargo +1.95.0 test -p datafusion-functions preserves_input_nullability
• cargo +1.95.0 clippy -p datafusion-functions --lib --tests --all-features --no-deps -- -D warnings
• cargo +1.95.0 fmt --all -- --check

Are there any user-facing changes?

Planned result schemas now retain non-nullability through UPPER and LOWER. Runtime values and public APIs are unchanged.

[jayzhan211]

Thanks @osipovartem , here is a suggestion

The unit tests only call return_field_from_args directly. Consider an end-to-end .slt case (e.g. in test_files/string/string_query.slt.part or functions.slt) so the planned schema and the outer-join case are covered:

statement ok
CREATE TABLE case_nullability (a VARCHAR NOT NULL, b VARCHAR) AS VALUES ('Ab', NULL);

statement ok
CREATE TABLE case_nullability_out AS
SELECT upper(a) ua, lower(a) la, upper(b) ub FROM case_nullability;

query TTT
DESCRIBE case_nullability_out;
----
ua Utf8View NO
la Utf8View NO
ub Utf8View YES

query T
SELECT upper(t.a) FROM (SELECT 1 k) x LEFT JOIN case_nullability t ON false;
----
NULL

[codecov-commenter]

Codecov Report

❌ Patch coverage is 94.73684% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.93%. Comparing base (ecb0fea) to head (be7d666).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
datafusion/functions/src/string/lower.rs	94.73%	0 Missing and 1 partial ⚠️
datafusion/functions/src/string/upper.rs	94.73%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #25381      +/-   ##
==========================================
- Coverage   81.93%   81.93%   -0.01%     
==========================================
  Files        1136     1136              
  Lines      428959   429190     +231     
  Branches   428959   429190     +231     
==========================================
+ Hits       351467   351656     +189     
- Misses      56469    56488      +19     
- Partials    21023    21046      +23     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[osipovartem]

Thanks for the suggestion. I added an end-to-end SQLLogicTest covering both sides of the nullability contract: UPPER/LOWER preserve non-nullable and nullable inputs in a CTAS schema, while a LEFT JOIN widens the input and UPPER correctly remains nullable. The focused SLT records and both UDF unit tests pass locally. I also rebased the branch onto the latest main.

[comphead]

you can use

> describe select upper('aaa');
+--------------------+-----------+-------------+
| column_name        | data_type | is_nullable |
+--------------------+-----------+-------------+
| upper(Utf8("aaa")) | Utf8      | YES         |
+--------------------+-----------+-------------+

[osipovartem]

Thanks. I replaced the CTAS-based schema check with direct DESCRIBE SELECT coverage. The test now checks both ordinary input nullability and nullability widened by a LEFT JOIN; it also retains the execution assertion that the widened value evaluates to NULL.

[comphead]

Thanks @osipovartem lgtm

[kumarUjjawal]

Thank you all!