Migrate Nunit To Mstest

4. Resolve semantic mappings

Load the mapping cheatsheet for every high-risk construct found in Step 1. These rules are mandatory:

• NUnit's default LifeCycle.SingleInstance differs from MSTest's per-test class instances. Preserve class-shared fields explicitly.
• [OneTimeSetUp] and [OneTimeTearDown] become static MSTest methods. Move any instance state they use to a static holder rather than merely adding static.
• NUnit Assert.Throws<T> and Throws.TypeOf<T> require an exact exception type and map to Assert.ThrowsExactly<T>. NUnit Assert.Catch<T> and Throws.InstanceOf<T> permit derived types and map to Assert.Throws<T>.
• NUnit Is.TypeOf<T> maps to Assert.IsExactInstanceOfType<T>; Is.InstanceOf<T> maps to Assert.IsInstanceOfType<T>.
• NUnit equality constraints compare sequences element-by-element. Use Assert.AreSequenceEqual on MSTest 4.3+ or CollectionAssert.AreEqual with materialized lists; never use reference-based Assert.AreEqual for sequences.
• TestCase(ExpectedResult=...) and TestCaseData.Returns(...) require rewriting the target test to assert the expected result.
• Preserve per-row names, categories, and ignores by returning TestDataRow<T> from DynamicData when needed. NUnit row properties have no direct TestDataRow<T> equivalent and require a custom data source or an explicit metadata decision.
• MSTest 4.4 combinatorial attributes can map NUnit [Combinatorial], [Values], [Range], and [Random]; add using Microsoft.VisualStudio.TestTools.UnitTesting.Combinatorial;. Constructor semantics differ, so translate NUnit ranges and random bounds explicitly. [Sequential] still requires explicit rows or a custom data source.
• NUnit [Pairwise] has no MSTest built-in. Preserve the exact generated rows in DynamicData or use a verified pairwise ITestDataSource; full Cartesian combinations change test counts and execution cost.
• Convert [Theory], [Datapoint], and [DatapointSource] into explicit DynamicData or combinatorial sources. Preserve automatic bool/enum values and the rule that a theory fails when every row violates its assumptions.
• [Explicit], [Repeat], parameterized fixtures, fixture sources, and Assert.Multiple have no behavior-identical mechanical mapping. Rewrite deliberately or report manual follow-up; never approximate silently.
• NUnit [Retry(n)] counts the initial attempt, while MSTest [Retry(n)] counts retries after the initial attempt. Subtract one and review NUnit RetryExceptions filters separately.
• [CancelAfter(ms)] maps to [Timeout(ms, CooperativeCancellation = true)]. Under VSTest, remove NUnit's injected CancellationToken parameter and use an injected or property-based TestContext.CancellationToken; VSTest otherwise treats it as missing data. Retain a method token only when the preserved runner is proven to support injection. Expand fixture-level defaults to every affected test and lifecycle method.
• [Apartment(ApartmentState.STA)] maps to [STATestClass] or [STATestMethod]. Set UseSTASynchronizationContext = true when async continuations must remain on the STA thread. MTA is the MSTest default. [RequiresThread] and [SingleThreaded] guarantee thread identity that MSTest attributes do not generally preserve; use a custom executor or report manual follow-up.
• Map NUnit platform and culture gates to OSCondition, ArchitectureCondition, or a MemberCondition helper. Map SetCulture and SetUICulture to setup/cleanup that saves and restores the original culture.
• Expand [DefaultFloatingPointTolerance] into explicit deltas on every affected assertion. Preserve method-over-fixture-over-assembly precedence.

5. Preserve lifecycle and namespace setup

• [SetUp] and [TearDown] remain per-test lifecycle methods.
• For NUnit's default single fixture instance, audit every mutable instance field. If tests depend on sharing, use static state created by [ClassInitialize] and released by [ClassCleanup].
• [FixtureLifeCycle(LifeCycle.InstancePerTestCase)] already matches MSTest class instantiation; remove the attribute and keep per-test instance state.
• [SetUpFixture] is namespace-scoped in NUnit. MSTest assembly initialization is assembly-wide. Use [AssemblyInitialize] only when the NUnit setup already covers the whole assembly; otherwise move setup into the affected classes or a shared helper without widening scope.
• Preserve setup/cleanup inheritance order. Do not merge base and derived methods unless the resulting order is proven equivalent.
• Map NUnit TestContext.WriteLine to an injected or property-based MSTest TestContext.WriteLine. Translate directory, test-name, and attachment APIs individually.

6. Preserve parallelization and ordering

NUnit and MSTest both run serially by default. Do not add parallelization for an unconfigured NUnit project.

When NUnit explicitly opts in:

• assembly [Parallelizable(ParallelScope.Fixtures)] -> [assembly: Parallelize(Workers = N, Scope = ExecutionScope.ClassLevel)]
• fixture [Parallelizable(ParallelScope.Children)] or method-level parallelism -> method-level MSTest parallelization only after confirming shared instance state is safe
• [NonParallelizable] -> [DoNotParallelize] when parallelization is enabled
• [LevelOfParallelism(N)] -> Workers = N
• resource-specific serialization can use MSTest 4.4 [ResourceLock("key")] when it preserves a narrower lock than [DoNotParallelize]

NUnit [Order] is not general dependency semantics. Use MSTest 4.4 [DependsOn] only when the source truly expresses a prerequisite; otherwise remove ordering by making tests independent or report manual follow-up.

7. Verify parity

1. Run tests with the same platform, filter, and configuration used for the baseline.
2. Compare discovered, passed, failed, and skipped counts.
3. Investigate every difference:
   • missing rows -> DataRow, DynamicData, combinatorial, or row metadata conversion
   • changed exceptions/types -> exact-vs-derived mapping
   • state failures -> NUnit single-instance fixture semantics or setup order
   • concurrency failures -> Parallelize, DoNotParallelize, worker count, or resource locks
   • changed skips -> Ignore, Explicit, row-level ignore, or retry behavior
4. Confirm no NUnit package, namespace, attribute, constraint, adapter setting, or custom NUnit extension remains unless documented for follow-up.
5. Read back high-risk changed files and name the exact target APIs in the result.

Use this final response shape:

• Changed: files and exact high-risk mappings.
• Verified: final command and discovered/passed/failed/skipped counts.
• Preserved: target framework, test platform, fixture scope, and concurrency choice.
• Remaining: manual follow-up, or none.

Completion Criteria

• NUnit version and test platform were identified
• NUnit packages and source constructs were converted
• Target framework and test platform stayed unchanged
• Default single-instance fixture semantics were audited and preserved where observable
• Data-row metadata and assertion semantics were preserved
• Namespace setup and parallelization scope were not silently widened
• Every attribute on NUnit's official attribute index was classified as direct, rewritten, removed, or manual
• Build succeeds and test results match the baseline
• Unsupported custom extensions or behavior are called out

Follow-up

Run migrate-vstest-to-mtp separately if the user also wants MTP. Use writing-mstest-tests only after parity is established.