Add unit test for not-ready Set detection in filterResultForMatchedRows

[tiandiwonder]

Changelog category (leave one):

• Build/Testing/Packaging Improvement

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

Not for changelog (test improvement)

Documentation entry for user-facing changes

• Documentation is not required (test improvement only)

---

Summary

This PR adds regression protection for the fix in PR #103029, which addressed incomplete not-ready Set detection in filterResultForMatchedRows.

Background

PR #103029 fixed a bug where filterResultForMatchedRows only checked filter_dag for not-ready Sets, but evaluated combined_dag = merge(pre_actions_dag, filter_dag). If a not-ready Set was in pre_actions_dag, the check would be bypassed, causing FunctionIn to throw a LOGICAL_ERROR exception when encountering the not-ready Set during evaluation.

This PR

Adds three unit tests in src/Processors/QueryPlan/Optimizations/tests/gtest_convert_any_join.cpp:

1. RealInOnNotReadySetWithoutCombinedGuardDiesOrThrows (death test)

   • Verifies that without the fix, real FunctionIn encounters not-ready Set and throws exception
   • Uses EXPECT_DEATH to prove the exception path exists

2. PreActionsDagNotReadySetReturnsBeforeEvaluatingFilter ⭐ (path-sensitive regression test)

   • Uses probe mechanism to verify early return before evaluation
   • If fix is reverted, probe would be triggered → test fails
   • Provides strong regression protection

3. FixedFilterResultForMatchedRows_ReturnsUnknown (positive test)

   • Verifies the fix returns UNKNOWN when pre_actions_dag contains not-ready Set

Implementation Details

• Exposes filterResultForMatchedRows in Utils.h for testing (follows pattern from commit 70cc74c which moved filterResultForNotMatchedRows from anonymous namespace to Utils.h)
• Adds MockNotReadyFutureSet test fixture
• Adds FunctionTestNotReadySetProbe with disabled constant folding for path-sensitive testing

Related

• PR [CI Fix] Fix not-ready Set when filter depends on the left-side column #103029: [CI Fix] Fix not-ready Set when filter depends on the left-side column #103029 (original fix, already merged)

[clickhouse-gh]

Workflow [PR], commit [e434b1f]

Summary: ❌

job_name	test_name	status	info	comment
Docs check		FAIL
	Generate system tables documentation	FAIL	cidb
Unit tests (amd_llvm_coverage)		FAIL
	ConvertAnyJoin.RealInOnNotReadySetWithoutCombinedGuardDiesOrThrows	FAIL	cidb
	AllTests	FAIL	cidb
LLVM Coverage		DROPPED

---

AI Review

Summary

This PR exposes filterResultForMatchedRows for unit testing and adds regression tests for not-ready Set handling in ANY JOIN optimization. The core code/test direction is reasonable, but the PR currently includes unrelated personal notes and has changelog metadata that does not match the stated intent of “not for changelog”.

PR Metadata

• ⚠️ Changelog category is inconsistent with the provided entry. The PR uses Build/Testing/Packaging Improvement, but the entry says Not for changelog (test improvement).
• ⚠️ For this PR content (test-only regression coverage), the category should be Not for changelog (or CI Fix or Improvement if that is the intent), with no user-facing changelog line.
• Exact replacement text:
  • ### Changelog category (leave one):
  • - Not for changelog (changelog entry is not required)
  • ### Changelog entry ...
  • Not for changelog (test improvement)

Findings

• ❌ Blockers
  • [tasks/lessons.md:1] The PR adds an unrelated personal working-notes file (tasks/lessons.md) with local-path/process notes, which is out of scope for the optimizer regression-test change and should not be merged as product code.
  • Suggested fix: remove tasks/lessons.md from the PR.

ClickHouse Rules

Item	Status	Notes
Deletion logging	➖
Serialization versioning	➖
Core-area scrutiny	✅
No test removal	✅
Experimental gate	➖
No magic constants	✅
Backward compatibility	✅
SettingsChangesHistory.cpp	➖
PR metadata quality	⚠️	Category/entry mismatch: marked Build/Testing/Packaging Improvement but entry says Not for changelog
Safe rollout	✅
Compilation time	✅
No large/binary files	✅

Final Verdict

• Status: ⚠️ Request changes
• Minimum required actions:
  1. Remove tasks/lessons.md from the PR.
  2. Fix PR changelog metadata so category and entry are consistent (use Not for changelog for this test-only change).

[clickhouse-gh]

This file looks like an accidental personal working note rather than part of the ANY JOIN regression test change. It adds unrelated process notes (including local absolute paths) and significantly increases PR noise.

Please remove tasks/lessons.md from this PR so the change stays focused on the intended optimizer/unit-test fix.