Modernize tests

Modernize Tests

Test modernization refers to two potential actions: migrating from XCTest to Swift Testing, and updating existing Swift Testing tests to use recommended patterns.

XCTests should be migrated to Swift Testing when possible. However, not all XCTests can be migrated to Swift Testing.

• UI tests (those that use XCUIAutomation) cannot be written with Swift Testing, and must remain XCTests.
• XCTests that use the measure { ... } family of APIs for performance measurement cannot be migrated. However, other test methods within an XCTestCase that do not use XCTest performance APIs can be migrated.

Migration Reference

Imports

Replace import XCTest with import Testing. A file can import both if it contains mixed test content during incremental migration.

When removing import XCTest, check whether the file uses Foundation types (URL, CharacterSet, ProcessInfo, Data, etc.). XCTest re-exports Foundation, so add import Foundation if needed.

Test Classes to Suites

Remove XCTestCase inheritance. Prefer structs over classes:

• final class FoodTruckTests: XCTestCase { ... } -> struct FoodTruckTests { ... }

Move setUp/tearDown code to init/deinit

Replace override func setUp() with init() (can be async throws). Replace override func tearDown() with deinit. If deinit is needed, use actor or final class instead of struct (since structs have no deinit). Change stored properties to not use implicitly-unwrapped optional types, and move their initial assignment from setUp to either be initialized inline or, if the initialization is complex, in an initializer.

struct MyTests {
    var fixture = Fixture()
    mutating func `Fixture behaves as expected`() {
        #expect(fixture.doSomething())
    }
}

Avoid pulling instance variables into function bodies; this can cause noise. Swift Testing reinvokes the initializer fresh before each test runs. If the test mutates an instance variable with value semantics, you may need to mark the test function mutating.

Test Methods

Replace the test name prefix with the @Test attribute. If the resulting test name includes multiple camelCase words, use a raw identifier with the test name in sentence case.

• func testEngineDoesNotStall() { ... } -> @Test func Engine does not stall() { ... }
• func testIgnition() { ... } -> @Test func ignition() { ... }

Test functions can be async, throws, or async throws, and can be isolated to a global actor with @MainActor.

Assertions to Expectations

When migrating a test from XCTest to Swift Testing, apply these mappings:

XCTAssert(x), XCTAssertTrue(x) -> #expect(x) XCTAssertFalse(x) -> #expect(!x) XCTAssertNil(x) -> #expect(x == nil) XCTAssertNotNil(x) -> #expect(x != nil) XCTAssertEqual(x, y) -> #expect(x == y) XCTAssertNotEqual(x, y) -> #expect(x != y) XCTAssertIdentical(x, y) -> #expect(x === y) XCTAssertNotIdentical(x, y) -> #expect(x !== y) XCTAssertGreaterThan(x, y) -> #expect(x > y) XCTAssertGreaterThanOrEqual(x, y) -> #expect(x >= y) XCTAssertLessThanOrEqual(x, y) -> #expect(x <= y) XCTAssertLessThan(x, y) -> #expect(x < y) try XCTUnwrap(x) -> try #require(x)

There is no direct equivalent for XCTAssertEqual(_:_:accuracy:); use floating point math directly.

Errors

When the error type is Equatable and the exact value is known, prefer to check the specific error value.

XCTAssertThrowsError(try f())

->

#expect(throws: (any Error).self) {
    try f()
}

XCTAssertThrowsError(try f()) { error in
    XCTAssertEqual(error, specificError)
}

->

#expect(throws: specificError) {
    try f()
}

XCTAssertThrowsError(try f()) { error in
    // Check error
}

->

let error = #expect(throws: (any Error).self) {
    try f()
}
// Check error

XCTAssertNoThrow(try f())

->

#expect(throws: Never.self) {
    try f()
}

continueAfterFailure

By default continueAfterFailure is true, which means expectations do not halt the test run. Some XCTestCases set continueAfterFailure = false, which means the XCTAssert family of functions will throw Objective-C exceptions that halt the test execution.

When a test method sets continueAfterFailure = false, all subsequent assertions need to be try #require(x) instead of #expect(x) to preserve this behavior. When adding try #require(x), add throws to the affected methods.

When continueAfterFailure = false is set in setUp, the conversion to try #require(x) must apply to all assertions in all test methods in that class.

Promote Issue.record/XCTFail to expectations

Wherever it is not disruptive, convert usage of Issue.record or XCTFail to #expect or #require, depending if the test exits after (taking continueAfterFailure into account).

In some cases, the source of the expectation itself is sufficient to explain the failure, and the comment would be redundant.

For example, the following structures should be converted as such:

guard let object = somethingOptional() else {
    Issue.record("Could not get object")
    return
}

guard object.isAvailable() else {
    Issue.record("Object not available")
    return
}

if !object.performOperation() {
    Issue.record("Failed to perform operation")
}

->

let object = try #require(somethingOptional(), "Could not get object")
try #require(object.isAvailable())
#expect(object.performOperation())

Asynchronous Expectations to Confirmations

Replace XCTestExpectation + fulfill() + await fulfillment(of:) with confirmation():

// Before
let exp = expectation(description: "...")
handler = { exp.fulfill() }
doWork()
await fulfillment(of: [exp])

// After
await confirmation("...") { confirm in
    handler = { confirm() }
    doWork()
}

For assertForOverFulfill = false with an expectedFulfillmentCount, use a range: await confirmation("...", expectedCount: 10...) { confirm in ... }

Skipping Tests

Replace XCTSkipIf/XCTSkipUnless with traits on the test or suite:

• try XCTSkipIf(condition) -> @Test(.disabled(if: condition))
• try XCTSkipUnless(condition) -> @Test(.enabled(if: condition))

Replace throw XCTSkip("reason") mid-test with try Test.cancel("reason").

When a skip checks OS version or platform availability, replace it with an @available attribute on the test function instead of .enabled(if:).

Known Issues

Replace XCTExpectFailure("...", ...) { ... } with withKnownIssue("...") { ... }.

For intermittent failures, replace .nonStrict() option (or the shorthand strict: false parameter) with isIntermittent: true.

For conditional/matching: use when: and matching: parameters:

withKnownIssue("...") {
    try riskyOperation()
} when: {
    shouldExpectFailure
} matching: { issue in
    issue.error != nil
}

Concurrency and Serial Execution

XCTest runs synchronous tests on the main actor and sequentially within a suite by default. Swift Testing runs all test functions on an arbitrary task and in parallel. Add @MainActor only if a test explicitly relied on main-actor isolation in its XCTest form, and add @Suite(.serialized) if tests depend on shared state.

Attachments

Replace XCTAttachment + self.add(attachment) with Attachment.record(value). The attached type must conform to Attachable (automatic for Codable and NSSecureCoding types when Foundation is imported).

Modernization Guidelines

• When migrating from XCTest, migrate one test class at a time. A file can contain both XCTest and Swift Testing tests during migration.
• Prefer struct for suites unless deinit (tearDown) is needed, in which case use actor or final class.
• Remove the test prefix from method names when adding @Test. For lengthier test names which read like a sentence, use raw identifier syntax to improve readability, e.g. @Test func Authenticate, fetch summary, then check count() { ... }.
• Use raw identifier syntax only for multi-word names that read like a sentence.
• When migrating setUp, convert implicitly-unwrapped optional properties to non-optional properties initialized in-place, or in init if initialization is complex, may throw, or is async.
• Look for explicit XCTFail/Issue.record calls that could be converted to #expect or #require
• Do not change try #require calls into #expect; this changes the behavior of tests.
• Add @MainActor only to tests that explicitly relied on XCTest's implicit main-actor isolation. Do not add it unnecessarily.
• Look for tests that loop over inputs or many repeated tests with the same logic and convert them to parameterized tests using @Test(arguments:).
• For suites with shared mutable state between tests, add @Suite(.serialized) and consider using actor or class instead of struct.
• Do not introduce usage of underscore-prefixed symbols such as #_sourceLocation; only use public API. For source locations, always use the full SourceLocation(fileID:filePath:line:column:) initializer.
• If the test suite already uses #_sourceLocation, do not replace the existing usage as part of modernization.
• Split this work over multiple agents if necessary if the modernization task is complex