I currently have a schema in production (cloudKit and local files) containing non-optional transformable values, e.g. @Attribute(.transformable(by: TestTransformer.self)) var number: TestTransformable = TestTransformable.init(value: 100) Unfortunately, this is preventing any migration from succeeding (documented at length in FB22151570). Briefly summarized, any migration from a Schema containing non-optional transformable values fails between willMigrate and didMigrate with the error "Can't find model for source store". This occurs for all migrations, including lightweight with a migration plan, lightweight without a plan, and custom migrations. Worst of all, this also prevents migration to optional transformable values, or the elimination of the transformable value entirely, leaving us completely stuck. (note: optional transformable values only work when they have a default value set to nil, otherwise even these have issues migrating) We already have features being blocked by this issue, and would like to preserve user-data while restoring our ability to move forwards with database. Are there any known workarounds for using SwiftData (+CloudKit) when schema migration is non-operational?

When deleting a SwiftData entity, I sometimes encounter the following error in a document based SwiftUI app: Fatal error: Unexpected backing data for snapshot creation: SwiftData._FullFutureBackingData<MyEntityClass> The deletion happens in a SwiftUI View and the code used to retrieve the entity is standard (the ModelContext is injected from the @Environment): let myEntity = modelContext.model(for: entityIdToDelete) modelContext.delete(myEntity) Unfortunately, I haven't yet managed to isolate this any further in order to come up with a reproducible PoC. Could you give me further information about what this error means?

I am trying out the new AttributedString binding with SwiftUI’s TextEditor in iOS26. I need to save this to a Core Data database. Core Data has no AttributedString type, so I set the type of the field to “Transformable”, give it a custom class of NSAttributedString, and set the transformer to NSSecureUnarchiveFromData When I try to save, I first convert the Swift AttributedString to NSAttributedString, and then save the context. Unfortunately I get this error when saving the context, and the save isn't persisted: CoreData: error: SQLCore dispatchRequest: exception handling request: <NSSQLSaveChangesRequestContext: 0x600003721140> , <shared NSSecureUnarchiveFromData transformer> threw while encoding a value. with userInfo of (null) Here's the code that tries to save the attributed string: struct AttributedDetailView: View { @ObservedObject var item: Item @State private var notesText = AttributedString() var body: some View { VStack { TextEditor(text: $notesText) .padding() .onChange(of: notesText) { item.attributedString = NSAttributedString(notesText) } } .onAppear { if let nsattributed = item.attributedString { notesText = AttributedString(nsattributed) } else { notesText = "" } } .task { item.attributedString = NSAttributedString(notesText) do { try item.managedObjectContext?.save() } catch { print("core data save error = \(error)") } } } } This is the attribute setup in the Core Data model editor: Is there a workaround for this? I filed FB17943846 if someone can take a look. Thanks.

SwiftData crashes 100% when fetching history of a model that contains an optional codable property that's updated: SwiftData/Schema.swift:389: Fatal error: Failed to materialize a keypath for someCodableID.someID from CrashModel. It is possible that this path traverses a type that does not work with append(), please file a bug report with a test. Would really appreciate some help or even a workaround. Code: import Foundation import SwiftData import Testing struct VaultsSwiftDataKnownIssuesTests { @Test func testCodableCrashInHistoryFetch() async throws { let container = try ModelContainer( for: CrashModel.self, configurations: .init( isStoredInMemoryOnly: true ) ) let context = ModelContext(container) try SimpleHistoryChecker.hasLocalHistoryChanges(context: context) // 1: insert a new value and save let model = CrashModel() model.someCodableID = SomeCodableID(someID: "testid1") context.insert(model) try context.save() // 2: check history it's fine. try SimpleHistoryChecker.hasLocalHistoryChanges(context: context) // 3: update the inserted value before then save model.someCodableID = SomeCodableID(someID: "testid2") try context.save() // The next check will always crash on fetchHistory with this error: /* SwiftData/Schema.swift:389: Fatal error: Failed to materialize a keypath for someCodableID.someID from CrashModel. It is possible that this path traverses a type that does not work with append(), please file a bug report with a test. */ try SimpleHistoryChecker.hasLocalHistoryChanges(context: context) } } @Model final class CrashModel { // optional codable crashes. var someCodableID: SomeCodableID? // these actually work: //var someCodableID: SomeCodableID //var someCodableID: [SomeCodableID] init() {} } public struct SomeCodableID: Codable { public let someID: String } final class SimpleHistoryChecker { static func hasLocalHistoryChanges(context: ModelContext) throws { let descriptor = HistoryDescriptor<DefaultHistoryTransaction>() let history = try context.fetchHistory(descriptor) guard let last = history.last else { return } print(last) } }

I have several macOS applications that use CloudKit. I need to test and finds out what happens when the user signs out of their iCloud account. That's because the application may lose data after signing out and then signing in again. Every time I do that, it'll take 15, 20 minutes... I don't time it, but it takes quite a gigantic time to sign out as the spinner keeps rolling. Why does it take so long to just sign out? This sign out effect is untestable because it takes a long time to sign out of an iCloud account and then make changes to the code and then test again. In case you need to know, my system version is Sequoia 15.7.

Hi all, I am using SwiftData and cloudkit and I am having an extremely persistent bug. I am building an education section on a app that's populated with lessons via a local JSON file. I don't need this lesson data to sync to cloudkit as the lessons are static, just need them imported into swiftdata so I've tried to use the modelcontainer like this: static func createSharedModelContainer() -> ModelContainer { // --- Define Model Groups --- let localOnlyModels: [any PersistentModel.Type] = [ Lesson.self, MiniLesson.self, Quiz.self, Question.self ] let cloudKitSyncModels: [any PersistentModel.Type] = [ User.self, DailyTip.self, UserSubscription.self, UserEducationProgress.self // User progress syncs ] However, what happens is that I still get Lesson and MiniLesson record types on cloudkit and for some reason as well, whenever I update the data models or delete and reinstall the app on simulator, the lessons duplicate (what seems to happen is that a set of lessons comes from the JSON file as it should), and then 1-2 seconds later, an older set of lessons gets synced from cloudkit. I can delete the old set of lessons if I just delete the lessons and mini lessons record types, but if I update the data model again, this error reccurrs. Sorry, I don't know if I managed to explain this well but essentially I just want to stop the lessons and minilessons from being uploaded to cloudkit as I think this will fix the problem. Am I doing something wrong with the code?

I have the following code running on macOS and iOS: CKQuerySubscription *zsub = [[CKQuerySubscription alloc] initWithRecordType:ESS_CLOUDCONTROLLER_RECORDTYPE_PUSHNOTE predicate:[NSPredicate predicateWithFormat:@"TRUEPREDICATE"] subscriptionID:@"pushZSub" options:CKQuerySubscriptionOptionsFiresOnRecordUpdate|CKQuerySubscriptionOptionsFiresOnRecordCreation|CKQuerySubscriptionOptionsFiresOnRecordDeletion]; zsub.zoneID = zid; CKNotificationInfo *inf = [[CKNotificationInfo alloc] init]; inf.shouldSendContentAvailable = YES; inf.desiredKeys = @[ESS_PN_RECORDFIELD_KEY_OVERALLDATE]; zsub.notificationInfo = inf; CKModifySubscriptionsOperation *msop = [[CKModifySubscriptionsOperation alloc] initWithSubscriptionsToSave:@[zsub] subscriptionIDsToDelete:nil]; msop.qualityOfService = NSQualityOfServiceUserInitiated; msop.modifySubscriptionsCompletionBlock = ^(NSArray<CKSubscription *> * _Nullable savedSubscriptions, NSArray<CKSubscriptionID> * _Nullable deletedSubscriptionIDs, NSError * _Nullable operationError) { dispatch_async(dispatch_get_main_queue(), ^{ if (savedSubscriptions.count == 1) { //works also when already created. compH(YES, nil); } else { compH(NO, nil); } }); }; [self.database addOperation:msop]; (code synopsis: after i create a custom zone (not shown in code), I add a ckquerysubscription to it for a specific record type, configured as a silent notification) When I change the according record in my Mac app, I get an immediate silent push on iOS. On macOS, however, after I change the record in my iOS app, I don't get one. Sometimes, one silent push makes it through every now and then a minute+ late or so, and after that, it's going missing again. What's the deal? Everything's set up correctly (com.apple.developer.aps-environment is set, container-identifiers are the same, icloud services are the same, ubiquity-kvstore-identifier are the same). I obviously register for remote notifications in both apps. I see all the records and subscriptions and zones in both the Mac and iOS app. I tried setting alertBody to an empty string, or soundName to an empty string, or both to an empty string: no difference I tried having different subscriptions for my Mac and iOS app, since they use different bundle ids, but that was merged into one subscription server-side, so I'm thinking that's not it I tried making it not-silent by setting contentAvailable to NO and adding a full alertBody, title and subtitle. Again, worked on iOS, not on macOS. This has been going on since macOS 14 Sonoma (when I first got reports of this. Now running on macOS 26.3). Before Sonoma, it worked just fine. Now I thought perhaps it's because I had a subscription on the default zone, and not a custom one, so I tried subscribing to changes on a record in a custom zone (see code above), but that did not change anything either. It's all working fine, only the push notifications are not making it through to the Mac app. If I sudo killall apsd (kill the push service daemon), the last push notification suddenly miraculously makes it through, by the way. At this point, I'm out of ideas and would very much appreciate pointers as to how to debug this. Polling every 30 seconds for changes is so 1990s. Speaking of which, this is a rather long-time-running app (started in 2011). Could my CloudKit database be “too old” or “corrupted” or whatever? Thank you kindly, – Matthias

Greetings i have an app that uses three different SwiftData models and i want to know what is the best way to use the them accross the app. I though a centralized behaviour and i want to know if it a correct approach.First let's suppose that the first view of the app will load the three models using the @Enviroment that work with @Observation. Then to other views that add data to the swiftModels again with the @Environment. Another View that will use the swiftData models with graph and datas for average and min and max.Is this a corrent way? or i should use @Query in every view that i want and ModelContext when i add the data. @Observable class CentralizedDataModels { var firstDataModel: [FirstDataModel] = [] var secondDataModel: [SecondDataModel] = [] var thirdDataModel: [ThirdDataModel] = [] let context: ModelContext init(context:ModelContext) { self.context = context } }

Hello, I have 3 model versions and I'm trying to step through migration. Version 2 makes significant changes to v1. As a result, I've renamed the entities in question by appending _v2 to their name, as the data isn't important to retain. v3, remove's the appended version number from v2. Setting the .xcdatamodeld to v3 and the migrations steps array as follows causes the app to error [ NSLightweightMigrationStage([v1]), NSLightweightMigrationStage([v2]), NSLightweightMigrationStage([v3]), ] CoreData: error: <NSPersistentStoreCoordinator: 0x10740d680>: Attempting recovery from error encountered during addPersistentStore: 0x10770f8a0 Error Domain=NSCocoaErrorDomain Code=134110 "An error occurred during persistent store migration." An error occurred during persistent store migration. Cannot merge multiple root entity source tables into one destination entity root table. I find this odd because if I run the migration independently across app launches, the migration appears to drop the no longer used tables in v2, then re-add them back in v3. So it seems to me that something is not finishing completely with the fully stepped through migration. -- I'm also unable to understand how to use NSCustomMigrationStage I've tried setting it to migrate from v1, to v2, but I'm getting a crash with error Duplicate version checksums across stages detected

Apple's iCloud File Management documentation says to "avoid special punctuation or other special characters" in filenames, but doesn't specify which characters. I need a definitive list to implement filename sanitization in my shipping app. Confirmed issues Our iOS app (CyberTuner, App Store, 15 years shipping on App Store) manages .rcta files in the iCloud ubiquity container via NSFileManager APIs. We've confirmed two characters causing sync failures: Ampersand (&): A file named Yamaha CP70 & CP80.rcta caused repeated "couldn't be backed up" dialogs. ~12 users reported this independently. Replacing & resolved it immediately. No other files in the same directory were affected. Percent (%): A file with % in the filename was duplicated by iCloud sync (e.g., filename% 1.rcta, filename% 2.rcta), and the original was lost. Currently reproducing across multiple devices. Both characters have special meaning in URL encoding (% is the escape character, & is the query parameter separator), which suggests the issue may be in URL handling within the sync pipeline. What I'm looking for: A definitive list of characters that cause problems in the iCloud sync pipeline specifically — not APFS restrictions, but CloudDocs/FileProvider/server-side issues. Confirmation whether these characters are problematic: & % # ? + / : * " < > | Is there a system API for validating or sanitizing filenames for iCloud compatibility before writing to the ubiquity container? Our users are piano technicians who naturally name files "Steinway & Sons" — we need to know exactly what to sanitize rather than guessing. Environment: iOS 17–26, Xcode 26.1, APFS, NSFileManager ubiquity container APIs Bundle FEEDBACK ASSISTANT ID FB21900837

We are currently implementing a custom iCloud sync for our macOS and iOS apps using CloudKit. Syncing works fine as long as the number of record sends is relatively small. But when we test with a large number of changes ( 80,000+ CKRecords ) we start running into problems. Our sending strategy is very conservative to avoid rate limits: We send records sequentially in batches of 250 records With about 2 seconds pause between operations Records are small and contain no assets (assets are uploaded separately) At some point we start receiving: “Database commit size exceeds limit” After that, CloudKit begins returning rate-limit errors with retryAfter-Information in the error. We wait for the retry time and try again, but from this moment on, nothing progresses anymore. Every subsequent attempt fails. We could not find anything in the official documentation regarding such a “commit size” limit or what triggers this failure state. So my questions are: Are there undocumented limits on the total number of records that can exist in an iCloud database (private or shared)? Is there a maximum volume of record modifications a container can accept within a certain timeframe, even if operations are split into small batches with pauses? Is it possible that sending large numbers of records in a row can temporarily or permanently “stall” a CloudKit container? Any insights or experiences would be greatly appreciated. Thank you!

Hi everyone I would like to achieve having unidirectional relationships in my SwiftData project (which I believe is possible: https://developer.apple.com/documentation/updates/swiftdata?changes=_9) but I'm afraid I'm struggling to overcome the errors I'm experiencing. For example, I have the following models: @Model final class Quota { @Attribute(.unique) var id: UUID var allowance: Int @Relationship(inverse: nil) var fish: Fish init(id: UUID = UUID(), fish: Fish, allowance: Int) { self.id = id self.fish = fish self.allowance = allowance } } @Model final class Fish { @Attribute(.unique) var id: Int var name: String init(id: Int, name: String) { self.id = id, self.name = name } } However, when I attempt to save a quota as so: let quota: Quota = .init(fish: Fish(id: 2, name: "Salmon"), allowance: 50) modelContext?.insert(quota) try save() I keep getting the following error: SwiftData.DefaultStore save failed with error: Error Domain=NSCocoaErrorDomain Code=1570 "%{PROPERTY}@ is a required value." UserInfo={NSValidationErrorObject=<NSManagedObject: 0x600002217390> (entity: Fish; id: 0x83319d001151328d <x-coredata://C76A2A64-146E-432F-A565-319B5A2F23F5/Fish/p12>; data: { id = nil; }), NSLocalizedDescription=%{PROPERTY}@ is a required value., NSValidationErrorKey=id, NSValidationErrorValue=null} %{PROPERTY}@ is a required value. However, if I set up Quota and Fish with an inverse relationship then the data saves as expected, so I'm a little confused. Is there anyone out there who can provide some guidance as to why I'm seeing this error when I try to save a record in SwiftData with no inverse relationship? I do fully understand about unidirectional vs bidirectional relationships but I have a scenario where I need the relationship to be unidirectional. Also, as a side note, the Fish record already exists in my database, but if I delete it and try to save the record I still see this error. Thank you so much in advance for any help.

I'm working on a macOS app with a Safari web extension. I'm trying to share a SwiftData model between devices using CloudKit synchronization. I am able to get synchronization in the main app on the same device, CloudKit sync works correctly — changes appear in the CloudKit Dashboard under com.apple.coredata.cloudkit.zone. However, in the Safari App Extension, data is saved locally and persists across launches, but never syncs to CloudKit. I have followed the recommended practices for configuring the App Group and entitlements, but the issue persists. Questions: Is there an official limitation preventing Safari App Extensions from connecting to the CloudKit daemon (cloudd)? If not, what entitlements or configuration changes are required for a Safari App Extension to successfully sync with CloudKit? Is the xpc_error=159 from bootstrap_look_up() a known sandbox restriction for this extension type? Any guidance from Apple engineers or others who have successfully used CloudKit from a Safari App Extension would be appreciated. What I’ve confirmed: The extension’s .entitlements includes: com.apple.security.app-sandbox com.apple.developer.icloud-services CloudKit com.apple.developer.icloud-container-identifiers iCloud.dev.example.myapp Same iCloud container ID for both app and extension CloudKit container exists and is initialized in CloudKit Console Running in :Sandbox environment during development Database name in SwiftData matches container identifier (without the iCloud. prefix) The extension’s codesign output shows correct entitlements App Group is configured (although in this case, extension and app use separate stores intentionally) Observed behavior in Console.app logs: CloudKit sync engine initializes in the extension XPC activities are registered for import/export: _xpc_activity_register: com.apple.coredata.cloudkit.activity.export. xpc_activity_set_criteria: ... import. Then a bootstrap lookup fails: failed to do a bootstrap look-up: xpc_error=[159: Unknown error: 159] CloudKit daemon connection error: CKErrorDomain Code=6 "Error connecting to CloudKit daemon" NSCocoaErrorDomain Code=4099 There is no “Will attempt to upload transactions” or “Upload succeeded” logs are ever seen. Symptoms When the extension is run, I see logs like the following in Console.app: [0x13e215820] failed to do a bootstrap look-up: xpc_error=[159: Unknown error: 159] CoreData+CloudKit: -[PFCloudKitSetupAssistant _checkAccountStatus:]_block_invoke(342): Fetched account info for store : (null) Error Domain=CKErrorDomain Code=6 "Error connecting to CloudKit daemon. This could happen for many reasons..."

I have an iOS app using SwiftData with VersionedSchema. The schema is synchronized with an CloudKit container. I previously introduced some model properties that I have now removed, as they are no longer needed. This results in the current schema version being identical to one of the previous ones (except for its version number). This results in the following exception: 'NSInvalidArgumentException', reason: 'Duplicate version checksums across stages detected.' So it looks like we cannot have a newer schema version with an identical content to an older schema version. The intuitive way would be to re-add the old (identical) schema version to the end of the "schemas" list property in the SchemaMigrationPlan, in order to signal that it is the newest one, and to add a migration stage back to it, thus: public enum MySchemaMigrationPlan: SchemaMigrationPlan { public static var schemas: [any VersionedSchema.Type] { [ SchemaV100.self, SchemaV101.self, SchemaV100.self ] } public static var stages: [MigrationStage] { [ migrateV100toV101, migrateV101toV100 ] } However, I am not sure if this is the right way to go, as previously, as I wanted to write unit tests for schema migration and rollback, I tried defining an inverse for each migration stage, so that I could trigger a migration and a rollback from a unit test, which resulted in an exception saying that it is not supported to downgrade a VersionedSchema. I must admit that I solved the original problem by introducing a dummy model property that I will later remove. What would have been the correct approach?

When creating a new project in Xcode 26, the default for defaultIsolation is MainActor. Core Data creates classes for each entity using code gen, but now those classes are also internally marked as MainActor, which causes issues when accessing managed object from a background thread like this. Is there a way to fix this warning or should Xcode actually mark these auto generated classes as nonisolated to make this better? Filed as FB13840800. nonisolated struct BackgroundDataHandler { @concurrent func saveItem() async throws { let context = await PersistenceController.shared.container.newBackgroundContext() try await context.perform { let newGame = Item(context: context) newGame.timestamp = Date.now // Main actor-isolated property 'timestamp' can not be mutated from a nonisolated context; this is an error in the Swift 6 language mode try context.save() } } } Turning code gen off inside the model and creating it manually, with the nonisolated keyword, gets rid of the warning and still works fine. So I guess the auto generated class could adopt this as well? public import Foundation public import CoreData public typealias ItemCoreDataClassSet = NSSet @objc(Item) nonisolated public class Item: NSManagedObject { }

The NSMetadataUbiquitousItemDownloadingStatusKey indicates the status of a ubiquitous (iCloud Drive) file. A key value of NSMetadataUbiquitousItemDownloadingStatusDownloaded is defined as indicating there is a local version of this file available. The most current version will get downloaded as soon as possible . However this no longer occurs since iOS 18.4. A ubiquitous file may remain in the NSMetadataUbiquitousItemDownloadingStatusDownloaded state for an indefinite period. There is a workaround: call [NSFileManager startDownloadingUbiquitousItemAtURL: error:] however this shouldn't be necessary, and introduces delays over the previous behaviour. Has anyone else seen this behaviour? Is this a permanent change? FB17662379

I’m seeing an issue with CloudKit sharing in iOS 26.2.1: When I call CKShare(rootRecord:) with a brand-new record in a fresh custom zone, the share is created with no root attached (rootFromShare == nil). After saving both the root and share in a single CKModifyRecordsOperation, fetching the share from the server still shows no root reference (rootRecordID == nil). No errors are thrown, but sharing simply fails. Key facts: • Custom zone created and confirmed (sharing enabled, capsRaw=7/15) • Brand-new record type and fresh IDs each run • Never reusing records or shares • Saving both root and share together in one operation • No default zone usage; always custom private zone Tested: • Multiple devices, iCloud accounts, and app versions • Both simulator and physical device Debug logs consistently show: • SHARE_CREATE_SHARE_LOCAL ... rootFromShare=nil • After save/fetch: rootRecordID=nil on server Has anyone seen this? Is there a new CloudKit regression in iOS 26.x, or am I missing something subtle? Minimal sample project and full debug logs available if needed. Any insights or workarounds would be hugely appreciated!

"No records found" If I create a new record on the console, I can copy the record name. I can then query for recordName and get that individual record back. BUT no other queries work. I cannot query all records. I cannot query by individual property. Just returns "no records found" Seems like my indexes got messed up. Is there a way to reset indexes on prod? This is on a coredata.cloudkit managed zone.

Since running on iOS 14b1, I'm getting this in my log (I have Core Data logging enabled): error: Store opened without NSPersistentHistoryTrackingKey but previously had been opened with NSPersistentHistoryTrackingKey - Forcing into Read Only mode store at 'file:///private/var/mobile/Containers/Shared/AppGroup/415B75A6-92C3-45FE-BE13-7D48D35909AF/StoreFile.sqlite' As far as I can tell, it's impossible to open my store without that key set - it's in the init() of my NSPersistentContainer subclass, before anyone calls it to load stores. Any ideas?

In a document based SwiftData app for macOS, how do you go about opening a (modal) child window connected to the ModelContainer of the currently open document? Using .sheet() does not really result in a good UX, as the appearing view lacks the standard window toolbar. Using a separate WindowGroup with an argument would achieve the desired UX. However, as WindowGroup arguments need to be Hashable and Codable, there is no way to pass a ModelContainer or a ModelContext there: WindowGroup(id: "myWindowGroup", for: MyWindowGroupArguments.self) { $args in ViewThatOpensInAWindow(args: args) } Is there any other way?