Fix 500 from coalesce, mvappend, array and percentile_approx types on short, byte and float values

[cnoramut]

Description

coalesce, mvappend, array and percentile_approx return HTTP 500 with a ClassCastException for some numeric types. coalesce(s) on a short or byte field fails, and so does any of the first three with a float literal such as 2.5F. percentile_approx and median fail in memory on short, byte, float and half_float fields.

Root cause. Each function hands its result to Calcite as an Object, and Calcite casts it to the Java class of the declared type. That class is Short for SMALLINT, Byte for TINYINT, Float for REAL and Double for FLOAT, since Calcite's FLOAT is double precision per its data types. Each function built the wrong class for some of these.

// EnhancedCoalesceFunction.coerceToType
case "SMALLINT", "TINYINT" ->
    tryConvert(() -> ExprValueUtils.integerValue(value.integerValue()), value);
case "FLOAT", "REAL" ->
    tryConvert(() -> ExprValueUtils.floatValue(value.floatValue()), value);

mvappend and array built Float elements for FLOAT in the same way. percentile_approx.result returned a double for SMALLINT, TINYINT and REAL and a float for FLOAT. A float field is typed REAL, so it reaches FLOAT only next to a literal such as 2.5F.

Fix. Each switch now builds the class Calcite expects. SMALLINT and TINYINT build a short and a byte, FLOAT joins DOUBLE, and REAL keeps the float.

Before and after. HTTP status on one index with s mapped short, b mapped byte and f mapped float, on main at 00262797c and on this branch.

Query	Before	After
eval x = coalesce(s)	500, Integer cannot be cast to Short	200, smallint
eval x = coalesce(b)	500, Integer cannot be cast to Byte	200, tinyint
eval x = coalesce(f, 2.5F)	500, Float cannot be cast to Double	200, float
eval x = mvappend(f, 2.5F)	500, Float cannot be cast to Double	200, array
eval x = array(f, 2.5F)	500, Float cannot be cast to Double	200, array
stats percentile_approx(s, 50), max(lower(host))	500, Double cannot be cast to Short	200, smallint
stats median(s), max(lower(host))	500, Double cannot be cast to Short	200, smallint
stats percentile_approx(f, 50), max(lower(host))	500, Double cannot be cast to Float	200, float
eval x = coalesce(f)	200, float	200, float

A matrix crossed every numeric field type with each literal kind and every field pair. On main, 33 of 126 coalesce calls failed, 15 of 108 mvappend calls, 16 float-literal array calls and 10 of 24 percentile_approx calls. On this branch each of those returns 200, apart from coalesce calls on unsigned_long, which PPL does not support and which fail the same way on main. The fixed values match the first present argument for coalesce, the elements for mvappend and array, and the pushed-down result for percentile_approx.

Notes

1. Test posture - issues/5873.yml adds 15 cases. The casts happen in generated code and in Avatica, so a plan-level unit test cannot reach them and a YAML REST test can. ArrayFunctionImplTest pinned the old Float result for FLOAT, so it now expects doubles.
2. Why FLOAT sits with DOUBLE - It looks like a typo next to REAL, but it matches Calcite's own mapping, which ExprSqlType already copies as case DOUBLE, FLOAT. Each switch carries a one-line comment, since regrouping FLOAT with REAL would bring the 500 back. A float field widened to FLOAT shows float noise, so 0.1 becomes 0.10000000149011612, the same as ifnull returns.

Related Issues

Resolves #5873

Testing

• issues/5873.yml, 15 cases. Without the fix, every case except the coalesce(f) and coalesce(s, 0) controls fails with HTTP 500. Reverting one function's hunk fails only that function's cases.
• :core:test ran 5334 tests and :ppl:test ran 2145, both with 0 failures.
• The before and after table and the matrices ran on a local single-node 3.10.0-SNAPSHOT build of main and of this branch.

Check List

• New functionality includes testing.
• New functionality has been documented.
• New functionality has javadoc added.
• New functionality has a user manual doc added.
• New PPL command checklist all confirmed.
• API changes companion pull request created.
• Commits are signed per the DCO using --signoff or -s.
• Public documentation issue/PR created.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🧪 PR contains tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
General	Add bounds checking for numeric casts Direct casting from double to short and byte may cause overflow or precision loss without validation. Consider adding range checks or documenting the expected behavior when retValue exceeds the target type's bounds. core/src/main/java/org/opensearch/sql/calcite/udf/udaf/PercentileApproxFunction.java [61-66] case SMALLINT: - short shortRet = (short) retValue; + short shortRet = (short) Math.max(Short.MIN_VALUE, Math.min(Short.MAX_VALUE, retValue)); return shortRet; case TINYINT: - byte byteRet = (byte) retValue; + byte byteRet = (byte) Math.max(Byte.MIN_VALUE, Math.min(Byte.MAX_VALUE, retValue)); return byteRet; Suggestion importance[1-10]: 5 __ Why: The suggestion addresses potential overflow when casting double to short and byte, which is a valid concern. However, the existing code follows the same pattern as the INTEGER and BIGINT cases (lines 55-60), suggesting this is the intended behavior. The suggestion adds defensive programming but may not be necessary if the percentile approximation is expected to stay within bounds. The score reflects a moderate improvement in robustness without being critical.	Low

[codecov]

Codecov Report

❌ Patch coverage is 0% with 13 lines in your changes missing coverage. Please review.
✅ Project coverage is 63.86%. Comparing base (0026279) to head (8515f96).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
...sql/calcite/udf/udaf/PercentileApproxFunction.java	0.00%	5 Missing ⚠️
...nction/udf/condition/EnhancedCoalesceFunction.java	0.00%	4 Missing ⚠️
...xpression/function/CollectionUDF/MVAppendCore.java	0.00%	2 Missing ⚠️
...n/function/CollectionUDF/MVAppendFunctionImpl.java	0.00%	2 Missing ⚠️

❌ Your project check has failed because the head coverage (63.86%) is below the target coverage (99.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #5874      +/-   ##
============================================
+ Coverage     63.80%   63.86%   +0.06%     
- Complexity     9070     9088      +18     
============================================
  Files           962      962              
  Lines         40989    41034      +45     
  Branches       4660     4668       +8     
============================================
+ Hits          26152    26206      +54     
+ Misses        13960    13950      -10     
- Partials        877      878       +1     

Flag	Coverage Δ
sql-engine	63.86% <0.00%> (+0.06%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.