Cache comment checker tree-sitter queries

[code-yeongyu]

Summary

• Cache tree-sitter queries per detector/language to avoid repeated query compilation.
• Add tests for cache reuse behavior.

Improvement table

Benchmark	Initial baseline	Final benchmark	Delta
BenchmarkCommentDetectorDetectPythonNoDocstrings-18	714618 ns/op, 1056 B/op, 22 allocs/op	11275 ns/op, 1024 B/op, 19 allocs/op	~63.4x faster, -32 B/op, -3 allocs/op
BenchmarkCommentDetectorDetectPythonWithDocstrings-18	3617623 ns/op, 2096 B/op, 43 allocs/op	14156 ns/op, 1800 B/op, 34 allocs/op	~255.6x faster, -296 B/op, -9 allocs/op
Unsupported extension	12.71 ns/op, 0 B/op, 0 allocs/op	unchanged fast path	Zero-allocation unsupported-file path preserved

Verification

• go build -o comment-checker ./cmd/comment-checker
• go test ./...
• go test -race ./pkg/core
• go test -bench='BenchmarkCommentDetectorDetectPython(NoDocstrings|WithDocstrings)$' -benchmem ./pkg/core
• manual CLI pass/warning QA

Review

• Security, QA, context, and goal-verifier lanes completed locally.
• Cubic review passed before merge.

[cubic-dev-ai]

2 issues found across 2 files

Confidence score: 3/5

• There is a concrete resource-management risk in pkg/core/detector.go: cached *sitter.Query objects are not closed, so C-backed query memory can leak for the detector lifetime in long-running use.
• The test in pkg/core/detector_test.go currently validates cache size only, which may miss a regression where queries are recompiled each call under the same key; this leaves cache-reuse behavior under-verified.
• Given the moderate severities and fairly strong confidence, this looks mergeable with caution, but fixing the query cleanup and strengthening reuse assertions would reduce regression risk.
• Pay close attention to pkg/core/detector.go and pkg/core/detector_test.go - query lifecycle cleanup and true cache-reuse validation need tightening.

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="pkg/core/detector.go">

<violation number="1" location="pkg/core/detector.go:171">
P2: Cached `*sitter.Query` instances are never closed, so query C-resources can leak for the lifetime of each detector.</violation>
</file>

<file name="pkg/core/detector_test.go">

<violation number="1" location="pkg/core/detector_test.go:120">
P2: This only checks cache size, so it won’t fail if the query is recompiled on every call and stored under the same key. Verify reuse by comparing the cached query pointer across calls (or by counting compilations) instead.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.
Re-trigger cubic}

[cubic-dev-ai]

P2: Cached *sitter.Query instances are never closed, so query C-resources can leak for the lifetime of each detector.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At pkg/core/detector.go, line 171:

<comment>Cached `*sitter.Query` instances are never closed, so query C-resources can leak for the lifetime of each detector.</comment>

<file context>
@@ -137,6 +131,47 @@ func (d *CommentDetector) detectDocstrings(sourceCode []byte, filePath string, l
+		return nil
+	}
+
+	cache[langName] = compiledQuery
+	return compiledQuery
+}
</file context>

[cubic-dev-ai]

P2: This only checks cache size, so it won’t fail if the query is recompiled on every call and stored under the same key. Verify reuse by comparing the cached query pointer across calls (or by counting compilations) instead.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At pkg/core/detector_test.go, line 120:

<comment>This only checks cache size, so it won’t fail if the query is recompiled on every call and stored under the same key. Verify reuse by comparing the cached query pointer across calls (or by counting compilations) instead.</comment>

<file context>
@@ -98,3 +98,25 @@ func main() {}`
+	// then
+	detector.queryMu.Lock()
+	defer detector.queryMu.Unlock()
+	assert.Len(t, detector.commentQueries, 1)
+	assert.Len(t, detector.docstringQueries, 1)
+}
</file context>