Hi, I'm trying to sign in with Apple CloudKit. I'm using the following code: 'use client'; import { CLOUDKIT_CONSTANTS } from '@/constants/cloudkit'; import { setCloudKitConfigured } from '@/lib/cloudkitSingleton'; import { CloudKitStatic } from '@/types/cloudkit'; import Script from 'next/script'; declare global { interface Window { CloudKit: CloudKitStatic; } } export default function Home() { const initializeCloudKit = async () => { console.info('⭐️ initializeCloudKit - start'); // 古い認証情報を削除 try { // LocalStorageから古い認証情報を削除 const keysToRemove = []; for (let i = 0; i < localStorage.length; i++) { const key = localStorage.key(i); if (key && (key.includes('cloudkit') || key.includes('CloudKit'))) { keysToRemove.push(key); } } keysToRemove.forEach(key => localStorage.removeItem(key)); // SessionStorageからも削除 const sessionKeysToRemove = []; for (let i = 0; i < sessionStorage.length; i++) { const key = sessionStorage.key(i); if (key && (key.includes('cloudkit') || key.includes('CloudKit'))) { sessionKeysToRemove.push(key); } } sessionKeysToRemove.forEach(key => sessionStorage.removeItem(key)); console.log('古い認証情報を削除しました'); } catch (cleanupError) { console.warn('認証情報のクリーンアップ中にエラー:', cleanupError); } try { const cloudKit = window.CloudKit.configure({ containers: [ { containerIdentifier: 'XXXXXX', apiTokenAuth: { apiToken: 'XXXX-XXXX-XXXX-XXXX-XXXX-XXXX-XXXX-XXXX', persist: false, signInButton: { id: 'cloudkit-sign-in-button', theme: 'black', }, signOutButton: { id: 'cloudkit-sign-out-button', theme: 'black', }, }, environment: 'development', }, ], }); console.info('⭐️ cloudKit', cloudKit); setCloudKitConfigured(true); const container = cloudKit.getDefaultContainer(); console.info('⭐️ CloudKit configured, setting up auth...'); // 初期認証状態をチェック try { const initialUser = await container.setUpAuth(); console.info('⭐️ setUpAuth result:', initialUser); } catch (authError) { console.info('⭐️ setUpAuth error (expected for unauthenticated):', authError); } // CloudKitの標準コールバックも併用（念のため） try { container.whenUserSignsIn().then((userInfo: any) => { console.info('⭐️ CALLBACK: whenUserSignsIn fired!', userInfo); }); container.whenUserSignsOut().then(() => { console.info('⭐️ CALLBACK: whenUserSignsOut fired!'); }); } catch (callbackError) { console.info('⭐️ Callback setup error (non-critical):', callbackError); } console.info('⭐️ initializeCloudKit - completed'); } catch (error) { console.error('⭐️ Critical CloudKit initialization error:', error); } }; return ( <> <Script src="https://cdn.apple-cloudkit.com/ck/2/cloudkit.js" strategy="afterInteractive" onLoad={() => { initializeCloudKit(); }} onError={error => { console.error('⭐️ CloudKit initialization error:', error); }} /> <div id="cloudkit-sign-in-button" /> <div id="cloudkit-sign-out-button" /> </> ); } In Chrome secret tab, I can sign in successfully. But in Chrome normal tab, I can't sign in. In normal tab, following error occurs on sign in button click: cloudkit.js:14 Uncaught (in promise) Error: UNKNOWN_ERROR cloudkit.js:14 GET https://api.apple-cloudkit.com/database/1/XXXXXX/XXXXXX/public/users/caller?ckjsBuildVersion=2420ProjectDev22&ckjsVersion=2.6.4&clientId=XXXXX-XXXXXXX-XXXX-XXXXX& ckAPIToken=XXXX-XXXX-XXXX-XXXX-XXXX-XXXX-XXXX-XXXX 421 (Misdirected Request) I think, cloudkit instance has re-initialized when I click the sign in button only in normal tab. So I can't sign in. Do you have any idea what might be causing the error ? Thanks in advance for your help!

What have people's experience with converting locally stored app data to a more browser based accessible format? Firebase seems expensive, Subabase a bit more challenging, and CloudKit too restrictive.

Hello, thank you Apple for supporting custom store with SwiftData and the Schema type is superb to work with. I have successfully set one up with SQL and have some feedback and issues regarding its APIs. There’s a highlighted message in the documentation about not using internal restricted symbols directly, but they contradict with the given protocols and I am concerned about breaking any App Store rules. Are we allowed to use these? If not, they should be opened up as they’re useful. BackingData is required to set up custom snapshots, initialization, and getting/setting values. And I want to use it with createBackingData() to directly initialize instances from snapshots when transferring them between server and client or concurrency. RelationshipCollection for casting to-many relationships from backing data or checking if an array contains a PersistentModel. SchemaProperty for type erasure in a collection. Schema.Relationship has KeyPath properties, but it is missing for Schema.Attribute and Schema.CompositeAttribute. Which means you can’t purely depend on the schema to map data. I am unable to access the properties of a custom struct type in a predicate unless I use Mirror with schemaMetadata() or CustomStringConvertible on the KeyPath directly to extract it. Trivial, but… the KeyPath property name is inconsistent (it’s all lowercase). It would be nice to retrieve property names from custom struct types, since you are unable access CodingKeys that are auto synthesized by Codable for structs. But I recently realized they’re a part Schema.CompositeAttribute, however I don’t know how to match these without the KeyPath… I currently map my entities using CodingKeys to their PredicateCodableKeyPathProviding.… but I wish for a simpler alternative! It’s unclear how to provide the schema to the snapshot before new models are created. I currently use a static property, but I want to make it flexible if more schemas and configurations are added later on. I considered saving and loading the schema in a temporary location, but doubtful that the KeyPath values will be available as they are not Codable. I suspect schemaMetadata() has the information I need to map the backing data without a schema for snapshots, but as mentioned previously, properties are inaccessible… Allow access to entity metatypes, like value types from SchemaProperty. They’re useful for getting data out of snapshots and casting them to CodingKeys and PredicateCodableKeyPathProviding. They do not carry over when you provide them in the Schema. I am unable to retrieve the primary key from PersistentIdentifier. It seems like once you create one, you can’t get it out, like the DataStoreConfiguration in ModelContainer is not the one you used to set it up. I cannot cast it, it is an entirely different struct? I have to use JSONSerialization to extract it, but I want to get it directly since it is not a column in my database. It is transformed when it goes to/from my tables. It’s unknown how to support some schema options, such as Spotlight and CloudKit. Allow for extending macro options, such as adding options to set as primary key, whether to auto increment, etc… You can create a schema for super and sub entities, but it doesn’t appear you can actually set them up from the @Model macro or use inheritance on these models… SwiftData history tracking seems incomplete for HistoryDelete, because that protocol requires HistoryTombstone, but this type cannot be instantiated, nor does it contain anything useful to infer from. As an aside, I want to create my own custom ModelActor that is a global actor. However, I’m unable to replicate the executor that Apple provides where the executor has a ModelContext, because this type does not conform to Sendable. So how did Apple do this? The documentation doesn’t mention unchecked Sendable, but I figure if the protocol is available then we would be able to set up our own. And please add concurrency features! Anyway, I hope for more continued support in the future and I am looking forward to what’s new this WWDC! 😊

About 4 months ago, I shipped the first version of my app with 4 versioned schemas that, unintentionally, had the same versionIdentifier of 1.2.0 in 2 of them: V1: 1.0.0 V2: 1.1.0 V3: 1.2.0 V4: 1.2.0 They are ordered correctly in the MigrationPlan, and they are all lightweight. Migration works, SwiftData doesn't crash on init and I haven't encountered any issues related to this. The app syncs with iCloud. Questions, preferable for anybody with knowledge of SwiftData internals: What will break in SwiftData when there are 2 duplicate numbers? Not that I would expect it to be safe, but does it happen to be safe to ship an update that changes V4's version to 1.3.0, what was originally intended?

We're in the process of migrating our app to the Swift 6 language mode. I have hit a road block that I cannot wrap my head around, and it concerns Core Data and how we work with NSManagedObject instances. Greatly simplied, our Core Data stack looks like this: class CoreDataStack { private let persistentContainer: NSPersistentContainer var viewContext: NSManagedObjectContext { persistentContainer.viewContext } } For accessing the database, we provide Controller classes such as e.g. class PersonController { private let coreDataStack: CoreDataStack func fetchPerson(byName name: String) async throws -> Person? { try await coreDataStack.viewContext.perform { let fetchRequest = NSFetchRequest<Person>() fetchRequest.predicate = NSPredicate(format: "name == %@", name) return try fetchRequest.execute().first } } } Our view controllers use such controllers to fetch objects and populate their UI with it: class MyViewController: UIViewController { private let chatController: PersonController private let ageLabel: UILabel func populateAgeLabel(name: String) { Task { let person = try? await chatController.fetchPerson(byName: name) ageLabel.text = "\(person?.age ?? 0)" } } } This works very well, and there are no concurrency problems since the managed objects are fetched from the view context and accessed only in the main thread. When turning on Swift 6 language mode, however, the compiler complains about the line calling the controller method: Non-sendable result type 'Person?' cannot be sent from nonisolated context in call to instance method 'fetchPerson(byName:)' Ok, fair enough, NSManagedObject is not Sendable. No biggie, just add @MainActor to the controller method, so it can be called from view controllers which are also main actor. However, now the compiler shows the same error at the controller method calling viewContext.perform: Non-sendable result type 'Person?' cannot be sent from nonisolated context in call to instance method 'perform(schedule:_:)' And now I'm stumped. Does this mean NSManageObject instances cannot even be returned from calls to NSManagedObjectContext.perform? Ever? Even though in this case, @MainActor matches the context's actor isolation (since it's the view context)? Of course, in this simple example the controller method could just return the age directly, and more complex scenarios could return Sendable data structures that are instantiated inside the perform closure. But is that really the only legal solution? That would mean a huge refactoring challenge for our app, since we use NSManageObject instances fetched from the view context everywhere. That's what the view context is for, right? tl;dr: is it possible to return NSManagedObject instances fetched from the view context with Swift 6 strict concurrency enabled, and if so how?

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

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

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?

Our app saves its data to iCloud by default. In most cases, this is working as intended & the data can be synced across devices with no problems. But recently, in testing, we discovered a situation where it's possible to save data before the NSMetadataQuery finishes & starts downloading the cloud files. When this happens, the query will then finish, and return the NEW file (with no other versions or conflicts). Is there a way to ensure that writing a file (version A) to ubiquitous storage when another version (version B) exists in the cloud is treated as a conflict, rather than just stomping all over the other version? I've tried querying the file metadata for the file URL (NSURLIsUbiquitousItemKey, NSMetadataUbiquitousItemDownloadingStatusKey, NSURLUbiquitousItemDownloadRequestedKey, NSURLUbiquitousItemHasUnresolvedConflictsKey) before saving, but it just returns nil.

Hey everyone, I have a question. When creating an app, how should I design a message table that involves personal privacy? The content is stored locally on the user's device, and then encrypted in the server database? How should I design it?

I'm building a photo editing app with a token-based subscription system using RevenueCat and StoreKit. Users purchase subscriptions that grant tokens for AI generations. There are no user accounts, the app is fully anonymous. Currently, I generate an anonymous account ID via RevenueCat SDK and store it in iCloud Keychain. This allows users on the same iCloud account to restore both their subscription and token balance across devices. However, users on a different iCloud account can restore their subscription via Apple, but their token balance is lost because there's no way to link the anonymous IDs. The problem is that if a user switches iCloud accounts or gets a new device without the same iCloud, their purchased tokens are orphaned. The subscription restores fine through Apple, but the token balance tied to the old anonymous ID becomes inaccessible. I have a few constraints: no user accounts, no email or phone sign-in, must work across devices owned by the same person, and must comply with App Store guidelines. My questions are: Is iCloud Keychain the right tool for this, or is there a better approach? Would CloudKit with an anonymous record zone be more appropriate? Are there any recommended patterns for persisting consumable balances tied to anonymous users across device migrations? Any guidance would be appreciated.

Hello, I am building a pretty large database (~40MB) to be used in my SwiftData iOS app as read-only. While inserting and updating the data, I noticed a substantial increase in size (+ ~10MB). A little digging pointed to ACHANGE and ATRANSACTION tables that apparently are dealing with Persistent History Tracking. While I do appreciate the benefits of that, I prefer to save space. Could you please point me in the right direction?

Hello everyone, I am experiencing a persistent authentication error when querying a custom user profile record, and the error message seems to be a red herring. My Setup: I have a custom CKRecord type called ColaboradorProfile. When a new user signs up, I create this record and store their hashed password, salt, nickname, and a custom field called loginIdentifier (which is just their lowercase username). In the CloudKit Dashboard, I have manually added an index for loginIdentifier and set it to Queryable and Searchable. I have deployed this schema to Production. The Problem: During login, I run an async function to find the user's profile using this indexed loginIdentifier. Here is the relevant authentication code: func autenticar() async { // ... setup code (isLoading, etc.) let lowercasedUsername = username.lowercased() // My predicate ONLY filters on 'loginIdentifier' let predicate = NSPredicate(format: "loginIdentifier == %@", lowercasedUsername) let query = CKQuery(recordType: "ColaboradorProfile", predicate: predicate) // I only need these specific keys let desiredKeys = ["password", "passwordSalt", "nickname", "isAdmin", "isSubAdmin", "username"] let database = CKContainer.default().publicCloudDatabase do { // This is the line that throws the error let result = try await database.records(matching: query, desiredKeys: desiredKeys, resultsLimit: 1) // ... (rest of the password verification logic) } catch { // The error always lands here logDebug("Error authenticating with CloudKit: \(error.localizedDescription)") await MainActor.run { self.errorMessage = "Connection Error: \(error.localizedDescription)" self.isLoading = false self.showAlert = true } } } The Error: Even though my query predicate only references loginIdentifier, the catch block consistently reports this error: Error authenticating with CloudKit: Field 'createdBy' is not marked queryable. I know createdBy (the system creatorUserRecordID) is not queryable by default, but my query isn't touching that field. I already tried indexing createdBy just in case, but the error persists. It seems CloudKit cannot find or use my index for loginIdentifier and is incorrectly reporting a fallback error related to a system field. Has anyone seen this behavior? Why would CloudKit report an error about createdBy when the query is explicitly on an indexed, custom field? I'm new to Swift and I'm struggling quite a bit. Thank you,

Hi ! Would anyone know (if possible) how to create backup files to export and then import from the data recorded by SwiftData? For those who wish, here is a more detailed explanation of my case: I am developing a small management software with customers and events represented by distinct classes. I would like to have an "Export" button to create a file with all the instances of these 2 classes and another "Import" button to replace all the old data with the new ones from a previously exported file. I looked for several solutions but I'm a little lost...

Testing Environment: iOS 18.4.1 / macOS 15.4.1 I am working on an iOS project that aims to utilize the user's iCloud Drive documents directory to save a specific directory-based file structure. Essentially, the app would create a root directory where the user chooses in iCloud Drive, then it would populate user generated files in various levels of nested directories. I have been attempting to use NSMetadataQuery with various predicates and search scopes but haven't been able to get it to directly monitor changes to files or directories that are not in the root directory. Instead, it only monitors files or directories in the root directory, and any changes in a subdirectory are considered an update to the direct children of the root directory. Example iCloud Drive Documents (Not app's ubiquity container) User Created Root Directory (Being monitored) File A Directory A File B An insertion or deletion within Directory A would only return a notification with userInfo containing data for NSMetadataQueryUpdateChangedItemsKey relating to Directory A, and not the file or directory itself that was inserted or deleted. (Query results array also only contain the direct children.) I have tried all combinations of these search scopes and predicates with no luck: query.searchScopes = [ rootDirectoryURL, NSMetadataQueryUbiquitousDocumentsScope, NSMetadataQueryAccessibleUbiquitousExternalDocumentsScope, ] NSPredicate(value: true) NSPredicate(format: "%K LIKE '*.md'", NSMetadataItemFSNameKey) NSPredicate(format: "%K BEGINSWITH %@", NSMetadataItemPathKey, url.path(percentEncoded: false)) I do see these warnings in the console upon starting my query: [CRIT] UNREACHABLE: failed to get container URL for com.apple.CloudDocs [ERROR] couldn't fetch remote operation IDs: NSError: Cocoa 257 "The file couldn’t be opened because you don’t have permission to view it." "Error returned from daemon: Error Domain=com.apple.accounts Code=7 "(null)"" But I am not sure what to make of that, since it does act normally for finding updates in the root directory. Hopefully this isn't a limitation of the API, as the only alternative I could think of would be to have multiple queries running for each nested directory that I needed updates for.

I have a document based SwiftData app in which I would like to implement a persistent cache. For obvious reasons, I would not like to store the contents of the cache in the documents themselves, but in my app's data directory. Is a use case, in which a document based SwiftData app uses not only the ModelContainers from the currently open files, but also a ModelContainer writing a database file in the app's documents directory (for cache, settings, etc.) supported? If yes, how can you inject two different ModelContexts, one tied to the currently open file and one tied to the local database, into a SwiftUI view?

In core-data I have a contact and location entity. I have one-to-many relationship from contact to locations and one-to-one from location to contact. I create contact in a seperate view and save it. Later I create a location, fetch the created contact, and save it while specifying the relationship between location and contact contact and test if it actually did it and it works. viewContext.perform { do { // Set relationship using the generated accessor method currentContact.addToLocations(location) try viewContext.save() print("Saved successfully. Locations count:", currentContact.locations?.count ?? 0) if let locs = currentContact.locations { print("📍 Contact has \(locs.count) locations.") for loc in locs { print("➡️ Location: \(String(describing: (loc as AnyObject).locationName ?? "Unnamed"))") } } } catch { print("Failed to save location: \(error.localizedDescription)") } } In my NSManagedObject class properties I have this : for Contact: @NSManaged public var locations: NSSet? for Location: @NSManaged public var contact: Contact? in my persistenceController I have: for desc in [publicStore, privateStore] { desc.setOption(true as NSNumber, forKey: NSPersistentStoreRemoteChangeNotificationPostOptionKey) desc.setOption(true as NSNumber, forKey: NSPersistentHistoryTrackingKey) desc.setOption(true as NSNumber, forKey: NSMigratePersistentStoresAutomaticallyOption) desc.setOption(true as NSNumber, forKey: NSInferMappingModelAutomaticallyOption) desc.setOption(true as NSNumber, forKey: "CKSyncCoreDataDebug") // Optional: Debug sync // Add these critical options for relationship sync desc.setOption(true as NSNumber, forKey: "NSPersistentStoreCloudKitEnforceRecordExistsKey") desc.setOption(true as NSNumber, forKey: "NSPersistentStoreCloudKitMaintainReferentialIntegrityKey") // Add this specific option to force schema update desc.setOption(true as NSNumber, forKey: "NSPersistentStoreRemoteStoreUseCloudKitSchemaKey") } When synchronization happens on CloudKit side, it creates CKRecords: CD_Contact and CD_Location. However for CD_Location it creates the relationship CD_contact as a string and references the CD_Contact. This I thought should have come as REFERENCE On the CD_Contact there is no CD_locations field at all. I do see the relationships being printed on coredata side but it does not come as REFERENCE on cloudkit. Spent over a day on this. Is this normal, what am I doing wrong here? Can someone advise?

Hi all, I recently discovered that I forgot to deploy my CloudKit schema changes from development to production - an oversight that unfortunately went unnoticed for 2.5 months. As a result, any data created during that time was never synced to iCloud and remains only in the local CoreData store. Once I pushed the schema to production, CloudKit resumed syncing new changes as expected. However, this leaves me with a gap: there's now a significant amount of data that would be lost if users delete or reinstall the app. Before I attempt to implement a manual backup or migration strategy, I was wondering: Does NSPersistentCloudKitContainer keep track of local changes that couldn't be synced doe to the missing schema and automatically reattempt syncing them now that the schema is live? If not, what would be the best approach to ensure this "orphaned" data gets saved to CloudKit retroactively. Thanks in advance for any guidance or suggestions.

I’m trying to build a CRUD app using SwiftData, @Query model and multidatepicker. The data from a multidatepicker is stored or persists in SwiftData as Set = []. My current dilemma is how to use SwiftData and @Query model Predicate to find all records on the current date. I can’t find any SwiftData documentation or examples @Query using Set = []. My CRUD app should retrieve all records for the current date. Unfortunately, I don’t know the correct @Query model syntax for Set = [].

I'm trying to build a custom FetchRequest that I can use outside a View. I've built the following ObservableFetchRequest class based on this article: https://augmentedcode.io/2023/04/03/nsfetchedresultscontroller-wrapper-for-swiftui-view-models @Observable @MainActor class ObservableFetchRequest&lt;Result: Storable&gt;: NSObject, @preconcurrency NSFetchedResultsControllerDelegate { private let controller: NSFetchedResultsController&lt;Result.E&gt; private var results: [Result] = [] init(context: NSManagedObjectContext = .default, predicate: NSPredicate? = Result.E.defaultPredicate(), sortDescriptors: [NSSortDescriptor] = Result.E.sortDescripors) { guard let request = Result.E.fetchRequest() as? NSFetchRequest&lt;Result.E&gt; else { fatalError("Failed to create fetch request for \(Result.self)") } request.predicate = predicate request.sortDescriptors = sortDescriptors controller = NSFetchedResultsController(fetchRequest: request, managedObjectContext: context, sectionNameKeyPath: nil, cacheName: nil) super.init() controller.delegate = self fetch() } private func fetch() { do { try controller.performFetch() refresh() } catch { fatalError("Failed to fetch results for \(Result.self)") } } private func refresh() { results = controller.fetchedObjects?.map { Result($0) } ?? [] } var predicate: NSPredicate? { get { controller.fetchRequest.predicate } set { controller.fetchRequest.predicate = newValue fetch() } } var sortDescriptors: [NSSortDescriptor] { get { controller.fetchRequest.sortDescriptors ?? [] } set { controller.fetchRequest.sortDescriptors = newValue.isEmpty ? nil : newValue fetch() } } internal func controllerDidChangeContent(_ controller: NSFetchedResultsController&lt;any NSFetchRequestResult&gt;) { refresh() } } Till this point, everything works fine. Then, I conformed my class to RandomAccessCollection, so I could use in a ForEach loop without having to access the results property. extension ObservableFetchRequest: @preconcurrency RandomAccessCollection, @preconcurrency MutableCollection { subscript(position: Index) -&gt; Result { get { results[position] } set { results[position] = newValue } } public var endIndex: Index { results.endIndex } public var indices: Indices { results.indices } public var startIndex: Index { results.startIndex } public func distance(from start: Index, to end: Index) -&gt; Int { results.distance(from: start, to: end) } public func index(_ i: Index, offsetBy distance: Int) -&gt; Index { results.index(i, offsetBy: distance) } public func index(_ i: Index, offsetBy distance: Int, limitedBy limit: Index) -&gt; Index? { results.index(i, offsetBy: distance, limitedBy: limit) } public func index(after i: Index) -&gt; Index { results.index(after: i) } public func index(before i: Index) -&gt; Index { results.index(before: i) } public typealias Element = Result public typealias Index = Int } The issue is, when I update the ObservableFetchRequest predicate while searching, it causes a Index out of range error in the Collection subscript because the ForEach loop (or a List loop) access a old version of the array when the item property is optional. List(request, selection: $selection) { item in VStack(alignment: .leading) { Text(item.content) if let information = item.information { // here's the issue, if I leave this out, everything works Text(information) .font(.callout) .foregroundStyle(.secondary) } } .tag(item.id) .contextMenu { if Item.self is Client.Type { Button("Editar") { openWindow(ClientView(client: item as! Client), id: item.id!) } } } } Is it some RandomAccessCollection issue or a SwiftUI bug?