Summary

This PR reduces redundant allocations in hot paths by caching parameter type arrays, attribute lookups, and assembly lists.

Critical Issues

None found ✅

Suggestions

1. IsTrxEnabled static caching may be too aggressive (TestExtensions.cs:177-195)

The new _cachedIsTrxEnabled static field is cached globally and never invalidated. While TRX configuration is unlikely to change during a test run, the cache is shared across:

• Multiple test sessions (if the process is reused)
• Different TestContext instances that might have different service configurations

Recommendation: Consider making this cache per-TestContext or per-session, or document why global static caching is safe here. If TRX state truly never changes in a process lifetime, the current approach is fine.

2. Consider defensive null check in GetParameterTypes (MethodDataSourceAttribute.cs:269-281)

The new GetParameterTypes helper correctly handles null and empty arrays. However, the loop doesn't validate that parameters[i].Type is non-null. While this is likely guaranteed by the framework, a defensive check could prevent future null reference exceptions if the contract changes.

Optional: Add a null check or document the invariant that ParameterMetadata.Type is never null.

3. Field keyword usage for Assemblies property (ReflectionTestDataCollector.cs:36)

The field keyword (C# 13 semi-auto property) is modern and appropriate here, replacing the previous lock-based caching. Good use of modern C#!

Just verify that:

• The project targets C# 13+ (<LangVersion>preview</LangVersion> or <LangVersion>13</LangVersion>)
• The team is aligned on using C# 13 features

Performance Impact Assessment

All changes align with TUnit Rule 4: Performance First. The optimizations target hot paths:
✅ Data generation (MethodDataSourceAttribute) - avoids repeated LINQ allocations per iteration
✅ Test discovery (ReflectionAttributeExtractor) - caches combined attributes
✅ Assembly scanning (ReflectionTestDataCollector) - uses semi-auto property vs lock overhead

These are solid micro-optimizations. Great work!

Verdict

✅ APPROVE - No critical issues. Suggestions are optional improvements for consideration.