axiom-realm-migration-ref

Realm 到 SwiftData 迁移 — 参考指南

目的：提供从 Realm 迁移到 SwiftData 的完整路径 Swift 版本：Swift 5.9+（建议使用启用严格并发检查的 Swift 6） iOS 版本：iOS 17+（建议使用 iOS 26+） 背景：Realm Device Sync 将于 2025 年 9 月 30 日结束。对于需要在截止日期前迁移的 Realm 用户，本指南至关重要。

---

关键时间表

Realm Device Sync 弃用截止日期 = 2025 年 9 月 30 日

如果你的应用使用 Realm Sync：

• ⚠️ 你必须在 2025 年 9 月 30 日前完成迁移
• ✅ SwiftData 是推荐的替代方案
• ⏰ 剩余时间：取决于当前日期，但生产环境应用的迁移需要 2-8 周

本指南提供成功迁移所需的一切。

---

迁移策略概览

Phase 1 (Week 1-2): Preparation & Planning
├─ Audit current Realm usage
├─ Understand model relationships
├─ Plan data migration path
└─ Set up test environment

Phase 2 (Week 2-3): Development
├─ Create SwiftData models from Realm schemas
├─ Implement data migration logic
├─ Convert threading model to async/await
└─ Test with real data

Phase 3 (Week 3-4): Migration
├─ Migrate existing app users' data
├─ Run in parallel (Realm + SwiftData)
├─ Verify CloudKit sync works
└─ Monitor for issues

Phase 4 (Week 4+): Production
├─ Deploy update with parallel persistence
├─ Gradual cutover from Realm to SwiftData
├─ Deprecate Realm code
└─ Monitor CloudKit sync health

---

第 1 部分：模式等效项

模型定义转换

Realm → SwiftData：基本模型

// REALM
class RealmTrack: Object {
    @Persisted(primaryKey: true) var id: String
    @Persisted var title: String
    @Persisted var artist: String
    @Persisted var duration: TimeInterval
    @Persisted var genre: String?
}

// SWIFTDATA
@Model
final class Track {
    @Attribute(.unique) var id: String  // remove if using CloudKit sync
    var title: String
    var artist: String
    var duration: TimeInterval
    var genre: String?

    init(id: String, title: String, artist: String, duration: TimeInterval, genre: String? = nil) {
        self.id = id
        self.title = title
        self.artist = artist
        self.duration = duration
        self.genre = genre
    }
}

主要区别：

• Realm：@Persisted(primaryKey: true) → SwiftData：@Attribute(.unique)（与 CloudKit 同步时不支持 — 使用 CloudKit 时请移除）
• Realm：隐式初始化器 → SwiftData：需要显式初始化器
• Realm：Object 基类 → SwiftData：在 final class 上使用 @Model 宏

Realm → SwiftData：关系

// REALM: One-to-Many
class RealmAlbum: Object {
    @Persisted(primaryKey: true) var id: String
    @Persisted var title: String
    @Persisted var tracks: RealmSwiftCollection<RealmTrack>
}

// SWIFTDATA: One-to-Many
@Model
final class Album {
    @Attribute(.unique) var id: String
    var title: String

    @Relationship(deleteRule: .cascade, inverse: \Track.album)
    var tracks: [Track] = []
}

@Model
final class Track {
    @Attribute(.unique) var id: String
    var title: String
    var album: Album?  // Inverse automatically maintained
}

主要区别：

• Realm：显式使用 RealmSwiftCollection 类型 → SwiftData：原生 [Track] 数组
• Realm：手动管理关系 → SwiftData：自动处理反向关系
• Realm：没有删除规则 → SwiftData：deleteRule: .cascade / .nullify / .deny

Realm → SwiftData：索引

// REALM
class RealmTrack: Object {
    @Persisted(primaryKey: true) var id: String
    @Persisted(indexed: true) var genre: String
    @Persisted(indexed: true) var releaseDate: Date
}

// SWIFTDATA
@Model
final class Track {
    @Attribute(.unique) var id: String
    @Attribute(.indexed) var genre: String = ""
    @Attribute(.indexed) var releaseDate: Date = Date()
}

---

第 2 部分：线程模型转换

Realm 线程处理 → Swift 并发

Realm：手动处理线程

class RealmDataManager {
    func fetchTracksOnBackground() {
        DispatchQueue.global().async {
            let realm = try! Realm()  // Must get Realm on each thread
            let tracks = realm.objects(RealmTrack.self)

            DispatchQueue.main.async {
                self.updateUI(tracks: Array(tracks))
            }
        }
    }

    func saveTrackOnBackground(_ track: RealmTrack) {
        DispatchQueue.global().async {
            let realm = try! Realm()
            try! realm.write {
                realm.add(track)
            }
        }
    }
}

问题：

• 手动使用 DispatchQueue 处理线程容易出错
• 容易在错误的线程上访问对象
• 没有编译时保证

SwiftData：基于 Actor 的并发

actor SwiftDataManager {
    let modelContainer: ModelContainer

    func fetchTracks() async -> [Track] {
        let context = ModelContext(modelContainer)
        let descriptor = FetchDescriptor<Track>()
        return (try? context.fetch(descriptor)) ?? []
    }

    func saveTrack(_ track: Track) async {
        let context = ModelContext(modelContainer)
        context.insert(track)
        try? context.save()
    }
}

// Usage (automatic thread handling)
@MainActor
class ViewController: UIViewController {
    @State private var tracks: [Track] = []
    private let manager: SwiftDataManager

    func loadTracks() async {
        tracks = await manager.fetchTracks()
    }
}

优势：

• 无需手动使用 DispatchQueue
• 编译时线程安全
• 自动隔离 Actor
• 兼容 Swift 6 严格并发检查

常见线程处理模式

Realm 模式	SwiftData 模式
DispatchQueue.global().async	后台 Actor 中的 async/await
realm.write { }	context.insert() + context.save()
手动创建线程本地 Realm 实例	共享的 ModelContainer + 后台 ModelContext
检查 Thread.isMainThread	添加 @MainActor 注解

---

第 3 部分：模式迁移策略

简单模式迁移（直接转换）

对于模式简单的应用（少于 5 张表、少于 10 个字段），直接迁移很简单：

actor SchemaImporter {
    let realmPath: String
    let modelContainer: ModelContainer

    func migrateFromRealm() async throws {
        // 1. Open Realm database
        let realmConfig = Realm.Configuration(fileURL: URL(fileURLWithPath: realmPath))
        let realm = try await Realm(configuration: realmConfig)

        // 2. Create SwiftData context
        let context = ModelContext(modelContainer)

        // 3. Migrate each model type
        try migrateAllTracks(from: realm, to: context)
        try migrateAllAlbums(from: realm, to: context)
        try migrateAllPlaylists(from: realm, to: context)

        // 4. Save all at once
        try context.save()

        print("Migration complete!")
    }

    private func migrateAllTracks(from realm: Realm, to context: ModelContext) throws {
        let realmTracks = realm.objects(RealmTrack.self)

        for realmTrack in realmTracks {
            let sdTrack = Track(
                id: realmTrack.id,
                title: realmTrack.title,
                artist: realmTrack.artist,
                duration: realmTrack.duration,
                genre: realmTrack.genre
            )
            context.insert(sdTrack)
        }
    }

    private func migrateAllAlbums(from realm: Realm, to context: ModelContext) throws {
        let realmAlbums = realm.objects(RealmAlbum.self)

        for realmAlbum in realmAlbums {
            let sdAlbum = Album(
                id: realmAlbum.id,
                title: realmAlbum.title
            )
            context.insert(sdAlbum)

            // Connect relationships after creating all records
            for realmTrack in realmAlbum.tracks {
                if let sdTrack = findTrack(id: realmTrack.id, in: context) {
                    sdAlbum.tracks.append(sdTrack)
                }
            }
        }
    }

    private func findTrack(id: String, in context: ModelContext) -> Track? {
        let descriptor = FetchDescriptor<Track>(
            predicate: #Predicate { $0.id == id }
        )
        return try? context.fetch(descriptor).first
    }
}

复杂模式迁移（转换层）

对于模式复杂、包含许多计算属性或需要进行数据转换的应用：

// Step 1: Define transformation layer
struct TrackDTO {
    let realmTrack: RealmTrack

    var id: String { realmTrack.id }
    var title: String { realmTrack.title }
    var cleanTitle: String { realmTrack.title.trimmingCharacters(in: .whitespaces) }
    var durationFormatted: String {
        let minutes = Int(realmTrack.duration) / 60
        let seconds = Int(realmTrack.duration) % 60
        return String(format: "%d:%02d", minutes, seconds)
    }
}

// Step 2: Migrate through transformation layer
actor ComplexMigrator {
    let modelContainer: ModelContainer

    func migrateWithTransformation(from realm: Realm) throws {
        let context = ModelContext(modelContainer)

        let realmTracks = realm.objects(RealmTrack.self)
        for realmTrack in realmTracks {
            let dto = TrackDTO(realmTrack: realmTrack)

            // Transform data during migration
            let sdTrack = Track(
                id: dto.id,
                title: dto.cleanTitle,  // Cleaned version
                artist: realmTrack.artist,
                duration: realmTrack.duration
            )
            context.insert(sdTrack)
        }

        try context.save()
    }
}

---

第 4 部分：CloudKit 同步迁移

Realm Sync → SwiftData CloudKit

现已弃用的 Realm Sync 提供自动同步。SwiftData 直接使用 CloudKit：

// REALM SYNC: Automatic but deprecated
let config = Realm.Configuration(
    syncConfiguration: SyncConfiguration(user: app.currentUser!)
)

// SWIFTDATA: CloudKit (recommended replacement)
let schema = Schema([Track.self, Album.self])
let config = ModelConfiguration(
    schema: schema,
    cloudKitDatabase: .private("iCloud.com.example.MusicApp")
)

let container = try ModelContainer(for: schema, configurations: config)

同步状态监控

@MainActor
class CloudKitSyncMonitor: ObservableObject {
    @Published var isSyncing = false
    @Published var lastSyncDate: Date?
    @Published var syncError: Error?

    let modelContainer: ModelContainer

    func startMonitoring() {
        // Monitor CloudKit sync notifications
        NotificationCenter.default.addObserver(
            forName: NSNotification.Name("CloudKitSyncDidComplete"),
            object: nil,
            queue: .main
        ) { [weak self] _ in
            self?.isSyncing = false
            self?.lastSyncDate = Date()
        }
    }

    func syncNow() async {
        isSyncing = true

        do {
            let context = ModelContext(modelContainer)
            // SwiftData sync happens automatically
            // Manually fetch to trigger sync
            let descriptor = FetchDescriptor<Track>()
            _ = try context.fetch(descriptor)
        } catch {
            syncError = error
        }

        isSyncing = false
    }
}

迁移时机：Realm Sync → CloudKit

Timeline:
Week 1-2: Development & Testing
├─ Create SwiftData models
├─ Test migrations in non-CloudKit mode
└─ Prepare CloudKit configuration

Week 3: CloudKit Sync Testing
├─ Enable CloudKit in test build
├─ Verify sync works with small datasets
├─ Test multi-device sync
└─ Test conflict resolution

Week 4+: Production Rollout
├─ Deploy app with SwiftData + CloudKit
├─ Initially run parallel (Realm Sync + SwiftData CloudKit)
├─ Monitor both sync mechanisms
├─ Gradually deprecate Realm Sync
└─ Final cutoff before Sept 30, 2025

---

第 5 部分：实际迁移场景

场景 A：小型应用（少于 10,000 条记录）

时间线：1-2 周 数据大小：少于 10 MB

// 1. Export Realm data
let realmPath = Realm.Configuration.defaultConfiguration.fileURL!

// 2. Migrate in background task
actor SmallAppMigration {
    let modelContainer: ModelContainer

    func migrateSmallApp() async throws {
        let realmConfig = Realm.Configuration(fileURL: realmPath)
        let realm = try await Realm(configuration: realmConfig)

        let context = ModelContext(modelContainer)

        // All-at-once migration (safe for < 10k records)
        let allTracks = realm.objects(RealmTrack.self)
        for realmTrack in allTracks {
            let track = Track(from: realmTrack)
            context.insert(track)
        }

        try context.save()
        print("✅ Migrated \(allTracks.count) tracks")
    }
}

// 3. Deploy
// Option 1: Migrate on first launch (offline)
// Option 2: Provide manual "Migrate Data" button
// Option 3: Automatic migration in background

场景 B：中型应用（100,000 - 1,000,000 条记录）

时间线：3-4 周 数据大小：100 MB - 1 GB 挑战：进度报告、内存管理

actor MediumAppMigration {
    let modelContainer: ModelContainer
    let realmPath: String

    typealias ProgressCallback = (Int, Int) -> Void

    func migrateMediumApp(onProgress: @MainActor ProgressCallback) async throws {
        let realmConfig = Realm.Configuration(fileURL: URL(fileURLWithPath: realmPath))
        let realm = try await Realm(configuration: realmConfig)

        let context = ModelContext(modelContainer)
        let allTracks = realm.objects(RealmTrack.self)
        let totalCount = allTracks.count

        // Chunk-based migration for memory efficiency
        var count = 0
        for chunk in Array(allTracks).chunked(into: 5000) {
            for realmTrack in chunk {
                let track = Track(from: realmTrack)
                context.insert(track)
            }

            // Save periodically
            try context.save()

            count += chunk.count
            await onProgress(count, totalCount)

            // Check for cancellation
            if Task.isCancelled {
                throw CancellationError()
            }
        }
    }
}

// 4. Show progress UI
@MainActor
class MigrationViewController: UIViewController {
    @IBOutlet weak var progressView: UIProgressView!
    @IBOutlet weak var statusLabel: UILabel!

    func startMigration() {
        Task {
            do {
                try await migrator.migrateMediumApp { current, total in
                    self.progressView.progress = Float(current) / Float(total)
                    self.statusLabel.text = "Migrated \(current) of \(total)..."
                }

                self.statusLabel.text = "✅ Migration complete!"
            } catch {
                self.statusLabel.text = "❌ Migration failed: \(error)"
            }
        }
    }
}

场景 C：大型应用（企业级，超过 1 Million 条记录）

时间线：6-8 周 数据大小：超过 1 GB 挑战：尽量减少停机时间、确保数据完整性、制定回滚计划

class EnterpriseGradualMigration {
    let coreDataStack: CoreDataStack  // Existing Realm
    let modelContainer: ModelContainer
    let batchSize = 10000

    // Phase 1: Parallel migration
    func startGradualMigration() async {
        var offset = 0
        let totalRecords = countAllRecords()

        while offset < totalRecords {
            let batch = fetchRealmBatch(limit: batchSize, offset: offset)
            try? await migrateBatch(batch)

            offset += batchSize
            await reportProgress(offset, totalRecords)
        }
    }

    private func migrateBatch(_ batch: [RealmTrack]) async throws {
        let context = ModelContext(modelContainer)

        for realmTrack in batch {
            let track = Track(from: realmTrack)
            context.insert(track)
            track.migrationStatus = .completedPhase1
        }

        try context.save()

        // Give main thread time to breathe
        try await Task.sleep(nanoseconds: 100_000_000)  // 100ms
    }

    // Phase 2: Verify all migrated
    func verifyMigrationComplete() async throws {
        let sdContext = ModelContext(modelContainer)
        let sdCount = try sdContext.fetch(FetchDescriptor<Track>())

        let realmCount = countAllRealmRecords()

        guard sdCount.count == realmCount else {
            throw MigrationError.countMismatch(sd: sdCount.count, realm: realmCount)
        }

        print("✅ Verified: \(sdCount.count) records migrated")
    }

    // Phase 3: Rollback plan
    func rollbackToRealm() {
        // Keep Realm database intact until 100% confident
        // Only delete Realm after running stable on SwiftData for 2+ weeks
    }
}

---

第 6 部分：测试与验证

数据完整性检查清单

在正式启用 SwiftData 前：

@MainActor
class MigrationVerifier {
    func verifyMigration() async throws {
        print("🔍 Running migration verification...")

        // 1. Count verification
        let sdCount = try await countSwiftDataRecords()
        let realmCount = countRealmRecords()
        print("✓ Record count: SD=\(sdCount), Realm=\(realmCount)")

        guard sdCount == realmCount else {
            throw VerificationError.countMismatch
        }

        // 2. Data integrity sampling (spot checks)
        try await verifySampleRecords(count: min(100, sdCount / 10))
        print("✓ Spot checked 100 records - all valid")

        // 3. Relationship integrity
        try await verifyRelationships()
        print("✓ All relationships intact")

        // 4. CloudKit sync test
        try await verifyCloudKitSync()
        print("✓ CloudKit sync working")

        // 5. Performance test
        try await verifyPerformance()
        print("✓ Query performance acceptable")

        print("✅ All verifications passed!")
    }

    private func verifySampleRecords(count: Int) async throws {
        let sdContext = ModelContext(modelContainer)
        let descriptor = FetchDescriptor<Track>()

        let tracks = try sdContext.fetch(descriptor)
        let sample = Array(tracks.prefix(count))

        for track in sample {
            // Verify fields populated
            assert(!track.id.isEmpty, "Track has empty ID")
            assert(!track.title.isEmpty, "Track has empty title")
            assert(track.duration > 0, "Track has invalid duration")
        }
    }

    private func verifyRelationships() async throws {
        let sdContext = ModelContext(modelContainer)

        let albumDescriptor = FetchDescriptor<Album>()
        let albums = try sdContext.fetch(albumDescriptor)

        for album in albums {
            // Verify inverse relationships
            for track in album.tracks {
                assert(track.album?.id == album.id, "Relationship broken")
            }
        }
    }

    private func verifyCloudKitSync() async throws {
        let sdContext = ModelContext(modelContainer)

        // Insert test record
        let testTrack = Track(
            id: "test-" + UUID().uuidString,
            title: "Test Track",
            artist: "Test Artist",
            duration: 240
        )
        sdContext.insert(testTrack)
        try sdContext.save()

        // Verify CloudKit sync initiated
        // (Check iCloud → iPhone → Settings → iCloud for sync status)
        print("ℹ️  Check iCloud app to verify sync initiated")
    }

    private func verifyPerformance() async throws {
        let sdContext = ModelContext(modelContainer)

        let start = Date()

        let descriptor = FetchDescriptor<Track>(
            sortBy: [SortDescriptor(\.title)]
        )
        _ = try sdContext.fetch(descriptor)

        let elapsed = Date().timeIntervalSince(start)
        print("Fetch time: \(String(format: "%.2f", elapsed))s")

        guard elapsed < 2.0 else {
            throw VerificationError.performanceIssue
        }
    }
}

---

第 7 部分：故障排查

常见迁移问题

问题	原因	解决方案
“Property must have default”	CloudKit 限制	添加默认值：var title: String = ""
关系未同步	缺少反向关系	添加 inverse: \Track.album
同步卡住	CloudKit 认证问题	检查“设置 → iCloud → CloudKit”
导入期间内存膨胀	未分块处理	实现批量导入（每次 1000 条）
数据丢失	没有备份	迁移后保留 Realm 副本 2 周

---

第 8 部分：成功标准

满足以下条件即表示迁移成功：

• 所有数据均已正确迁移（数量相符）
• 样本记录验证通过（抽查 100 条以上记录）
• 关系完整（反向关系正常工作）
• 已启用 CloudKit 同步且运行正常
• 性能可接受（查询耗时 < 1 秒）
• 没有数据竞争（Swift 6 严格并发）
• 已在真实设备上测试（不只是模拟器）
• 已记录并测试回滚计划
• Realm 数据库保留为备份，持续 2 周
• 上线 1 周后生产环境中零崩溃

---

快速参考：命令检查清单

# 1. Audit Realm usage
grep -r "RealmTrack\|RealmAlbum" . --include="*.swift"

# 2. Count Realm records (in app)
let realm = try! Realm()
let count = realm.objects(RealmTrack.self).count

# 3. Export Realm database
cp ~/Library/Developer/Realm/my_realm.realm ~/Downloads/backup.realm

# 4. Test SwiftData models
// Create in-memory test container
let config = ModelConfiguration(isStoredInMemoryOnly: true)
let container = try ModelContainer(for: Track.self, configurations: config)

# 5. Verify CloudKit
Settings → [Your Name] → iCloud → Check CloudKit status

---

资源

WWDC：2024-10137

文档：/swiftdata

技能：axiom-swiftdata、axiom-swift-concurrency、axiom-database-migration

---

创建时间：2025-11-30
状态：可用于生产环境的迁移指南
紧急程度：Realm Device Sync 将于 2025 年 9 月 30 日停止服务
预计迁移时间：2-8 周，具体取决于应用复杂度