I am receiving up to four push notifications for each event I am subscribed to. I have gone through everything related to my CloudKit subscriptions and notification registry and I am convinced this is an Apple problem. I have instead turned my attention toward correctly processing the notifications no matter how many I receive. Here is a simplified version of what I am doing:
func recievePrivatePush(_ pushInfo: [String:NSObject], completion: #escaping ()->Void) {
let notification = CKNotification(fromRemoteNotificationDictionary: pushInfo)
let alertBody = notification.alertBody
if let queryNotification = notification as? CKQueryNotification {
let recordID = queryNotification.recordID
guard let body = queryNotification.alertBody else {
return
}
if recordID != nil {
switch body {
case "Notification Type":
let id = queryNotification.recordID
switch queryNotification.queryNotificationReason {
case .recordCreated:
DataCoordinatorInterface.sharedInstance.fetchDataItem(id!.recordName, completion: {
//
})
break
default:
break
}
}
}
}
}
The fetching code looks something like this:
func fetchDataItem(_ id: String, completion: #escaping ()-> Void) {
if entityExistsInCoreData(id) {return}
let db = CKContainer.default().privateCloudDatabase
let recordID = CKRecordID(recordName: id)
db.fetch(withRecordID: recordID) { (record, error) in
if let topic = record {
//Here I create and save the object to core data.
}
completion()
}
}
All of my code works, the problem I am having is that when I receive multiple notifications, multiple fetch requests are started before the first core data entity is created, resulting in redundant core data objects.
What I would like to do is find a way to add the fetch requests to a serial queue so they are processed one at a time. I can put my request calls in a serial queue, but the callbacks always run asynchronously, so multiple fetch requests are still make before the first data object is persisted.
I have tried using semaphores and dispatch groups with a pattern that looks like this:
let semaphore = DispatchSemaphore(value: 1)
func recievePrivatePush(_ pushInfo: [String:NSObject], completion: #escaping ()->Void) {
_ = semaphore.wait(timeout: .distantFuture)
let notification = CKNotification(fromRemoteNotificationDictionary: pushInfo)
let alertBody = notification.alertBody
if let queryNotification = notification as? CKQueryNotification {
let recordID = queryNotification.recordID
guard let body = queryNotification.alertBody else {
return
}
if recordID != nil {
switch body {
case "Notification Type":
let id = queryNotification.recordID
switch queryNotification.queryNotificationReason {
case .recordCreated:
DataCoordinatorInterface.sharedInstance.fetchDataItem(id!.recordName, completion: {
semaphore.signal()
})
break
default:
break
}
}
}
}
}
Once the above function is called for the second time, and semaphore.wait is called, the execution of the first network request pauses, resulting in a frozen app.
Again, what I would like to accomplish it adding the asynchronous network requests to a queue so that they are made only one at a time i.e. the first network call is completed before the second request is started.
Carl,
Perhaps you'll find your solutions with dispatch groups, a few key expressions to look into.
let group = DispatchGroup()
group.enter()
... code ...
group.leave
group.wait()
I use them to limit the number of http requests I send out in a batch, to wait for the response. Perhaps you could use them together with the suggestion in my comment. Watch this video too, dispatch groups in here, I think more.
https://developer.apple.com/videos/play/wwdc2016/720/
These simple classes helped me solve the problem.
class PushQueue {
internal var pushArray: Array<String> = [String]()
internal let pushQueue = DispatchQueue(label: "com.example.pushNotifications")
public func addPush(_ push: Push) {
pushQueue.sync {
if pushArray.contains(push.id) {
return
} else {
pushArray.append(push.id)
processNotification(push: push)
}
}
}
internal func processNotification(push: Push) {
PushInterface.sharedInstance.recievePrivatePush(push.userInfo as! [String: NSObject])
}
}
class CKPush: Equatable {
init(userInfo: [AnyHashable: Any]) {
let ck = userInfo["ck"] as? NSDictionary
let id = ck?["nid"] as? String
self.id = id!
self.userInfo = userInfo
}
var id: String
var userInfo: [AnyHashable:Any]
public static func ==(lhs: CKPush, rhs: CKPush) -> Bool {
return lhs.id == rhs.id ? true : false
}
}
Please ignore the sloppy force unwraps. They need to be cleaned up.
Related
I apologize if this question is simple or the problem is obvious as I am still a beginner in programming.
I am looping over an array and trying to make an async Firestore call. I am using a DispatchGroup in order to wait for all iterations to complete before calling the completion.
However, the Firestore function is not even getting called. I tested with print statements and the result is the loop iterations over the array have gone through with an enter into the DispatchGroup each time and the wait is stuck.
func getUserGlobalPlays(username: String, fixtureIDs: [Int], completion: #escaping (Result<[UserPlays]?, Error>) -> Void) {
let chunkedArray = fixtureIDs.chunked(into: 10)
var plays: [UserPlays] = []
let group = DispatchGroup()
chunkedArray.forEach { ids in
group.enter()
print("entered")
DispatchQueue.global().async { [weak self] in
self?.db.collection("Users").document("\(username)").collection("userPlays").whereField("fixtureID", in: ids).getDocuments { snapshot, error in
guard let snapshot = snapshot, error == nil else {
completion(.failure(error!))
return
}
for document in snapshot.documents {
let fixtureDoc = document.data()
let fixtureIDx = fixtureDoc["fixtureID"] as! Int
let choice = fixtureDoc["userChoice"] as! Int
plays.append(UserPlays(fixtureID: fixtureIDx, userChoice: choice))
}
group.leave()
print("leaving")
}
}
}
group.wait()
print(plays.count)
completion(.success(plays))
}
There are a few things going on with your code I think you should fix. You were dangerously force-unwrapping document data which you should never do. You were spinning up a bunch of Dispatch queues to make the database calls in the background, which is unnecessary and potentially problematic. The database call itself is insignificant and doesn't need to be done in the background. The snapshot return, however, can be done in the background (which this code doesn't do, so you can add that if you wish). And I don't know how you want to handle errors here. If one document gets back an error, your code sends back an error. Is that how you want to handle it?
func getUserGlobalPlays(username: String,
fixtureIDs: [Int],
completion: #escaping (_result: Result<[UserPlays]?, Error>) -> Void) {
let chunkedArray = fixtureIDs.chunked(into: 10)
var plays: [UserPlays] = []
let group = DispatchGroup()
chunkedArray.forEach { id in
group.enter()
db.collection("Users").document("\(username)").collection("userPlays").whereField("fixtureID", in: id).getDocuments { snapshot, error in
if let snapshot = snapshot {
for doc in snapshot.documents {
if let fixtureIDx = doc.get("fixtureIDx") as? Int,
let choice = doc.get("choice") as? Int {
plays.append(UserPlays(fixtureID: fixtureIDx, userChoice: choice))
}
}
} else if let error = error {
print(error)
// There was an error getting this one document. Do you want to terminate
// the entire function and pass back an error (through the completion
// handler)? Or do you want to keep going and parse whatever data you can
// parse?
}
group.leave()
}
}
// This is the completion handler of the Dispatch Group.
group.notify(queue: .main) {
completion(.success(plays))
}
}
I am developing a chatting application where I could receive a number of messages at a time which leads to app freezing. Following is my socket receiver:
func receiveNewDirectMessages() {
self.socket?.on(EventListnerKeys.message.rawValue, callback: { (arrAckData, ack) in
print_debug(arrAckData)
guard let dictMsg = arrAckData.first as? JSONDictionary else { return }
guard let data = dictMsg[ApiKey.data] as? JSONDictionary else { return }
guard let chatData = data[ApiKey.data] as? JSONDictionary else { return }
guard let messageId = chatData[ApiKey._id] as? String , let chatId = chatData[ApiKey.chatId] as? String else { return }
if MessageModel.getMessageModel(msgId: messageId) != nil { return }
let isChatScreen = self.isChatScreen
let localMsgId = "\(arc4random())\(Date().timeIntervalSince1970)"
if let senderInfo = data[ApiKey.senderInfo] as? JSONDictionary, let userId = senderInfo[ApiKey.userId] as? String, userId != User.getUserId() {
_ = AppUser.writeAppUserModelWith(userData: senderInfo)
}
let msgModel = MessageModel.saveMessageData(msgData: chatData, localMsgId: localMsgId, msgStatus: 2, seenByMe: false)
let chatModel = ChatModel.saveInboxData(localChatId: msgModel.localChatId, inboxData: chatData)
if isChatScreen {
self.emitMessageStatus(msgId: messageId, chatId: chatId, socketService: .messageStatus, status: .delivered)
self.emitMessageStatus(msgId: messageId, chatId: chatId, socketService: .messageStatus, status: .seen)
} else {
ChatModel.updateUnreadCount(localChatId: chatModel.localChatId, incrementBy: 1)
self.emitMessageStatus(msgId: messageId, chatId: chatId, socketService: .messageStatus, status: .delivered)
}
TabController.shared.updateChatBadgeCount()
})
}
What's happening above:
1. I am receiving all the undelivered messages ONE-By-ONE in this socket listener.
2. Fetching the message data
3. Saving the received sender's info to Realm DB
4. Saving the message model to realm DB
5. SAVING/UPDATING Chat Thread in realm DB
6. Emitting acknowledgement for the received message
7. Update Chat badge count on tab bar
Below is my emitter for acknowledging the message delivery.
func emitMessageStatus(msgId: String, chatId: String, socketService: SocketService, status: MessageStatusAction) {
// Create Message data packet to be sent to socket server
var msgDataPacket = [String: Any]()
msgDataPacket[ApiKey.type] = socketService.type
msgDataPacket[ApiKey.actionType] = socketService.listenerType
msgDataPacket[ApiKey.data] = [
ApiKey.messageId: msgId,
ApiKey.chatId: chatId,
ApiKey.userId: User.getUserId(),
ApiKey.statusAction: status.rawValue
]
// send the messsage data packet to socket server & wait for the acknowledgement
self.emit(with: EventListnerKeys.socketService.rawValue, msgDataPacket) { (arrAckData) in
print_debug(arrAckData)
guard let dictMsg = arrAckData.first as? JSONDictionary else { return }
if let msgData = dictMsg[ApiKey.data] as? [String: Any] {
// Update delivered Seen Status here
if let msgId = msgData[ApiKey.messageId] as? String, let actionType = msgData[ApiKey.statusAction] as? String, let msgStatusAction = MessageStatusAction(rawValue: actionType) {
switch msgStatusAction {
case .delivered:
if let deliveredTo = msgData[ApiKey.deliveredTo] as? [[String: Any]] {
_ = MessageModel.updateMsgDelivery(msgId: msgId, deliveredTo: deliveredTo)
}
case .seen:
if let seenBy = msgData[ApiKey.seenBy] as? [[String: Any]] {
_ = MessageModel.updateMsgSeen(msgId: msgId, seenBy: seenBy)
}
case .pin:
MessageModel.clearPinnedMessages(chatId: chatId)
if let pinTime = msgData[ApiKey.pinTime] as? Double {
MessageModel.updatePinnedStatus(msgId: msgId, isPinned: true, pinTime: pinTime)
}
case .unPin:
if let pinTime = msgData[ApiKey.pinTime] as? Double {
MessageModel.updatePinnedStatus(msgId: msgId, isPinned: false, pinTime: pinTime)
}
case .delete:
MessageModel.deleteMessage(msgId: msgId)
case .ackMsgStatus, .like, .unlike:
break
}
}
}
}
}
What's happening above:
Encapsulating all the related information to acknowledge the event
Update realm DB after acknowledgement delivery
Now, I'm not able to defies a perfect threading policy here. What to write in background thread and what should I write in Main thread. I tried doing it however but that leades to random crashes or packet lossses.
Can anyone please lead me forward on this topic. I will be highly grateful.
Try to use background thread for data processing/ non-UI processing.
Reduce number of updating UI times
Instead of processing 1 by 1 message, using debounce - like. You can store new messages, then update UI with n new messages. So instead of updating UI/saving data to db 100 times for 100 messages, you can do it 1 time for 100 messages. More detail: with every new message, add it to an array. Call debouncer. The Debouncer will delay a function call, and every time it's getting called it will delay the preceding call until the delay time is over. So after e.g 200ms, if there're no new message, the update func will be call (the callback func that debounce's handling). Then update ui/db with n new stored messages.
You can group message by time, like group by 1 hour. And then update with delay between each time group. You can do it when the debouncer is called -> group messages by time -> update db/ui by each group. You can use setTimeout, like update group 1, 100ms later update group 2, so the ui won't be freezed
I have created a function getFriends that reads a User's friendlist from firestore and puts each friend in a LocalUser object (which is my custom user class) in order to display the friendlist in a tableview. I need the DispatchSemaphore.wait() because I need the for loop to iterate only when the completion handler inside the for loop is called.
When loading the view, the app freezes. I know that the problem is that semaphore.wait() is called in the main thread. However, from reading DispatchQueue-tutorials I still don't understand how to fix this in my case.
Also: do you see any easier ways to implement what I want to do?
This is my call to the function in viewDidLoad():
self.getFriends() { (friends) in
self.foundFriends = friends
self.friendsTable.reloadData()
}
And the function getFriends:
let semaphore = DispatchSemaphore(value: 0)
func getFriends(completion: #escaping ([LocalUser]) -> ()) {
var friendsUID = [String : Any]()
database.collection("friends").document(self.uid).getDocument { (docSnapshot, error) in
if error != nil {
print("Error:", error!)
return
}
friendsUID = (docSnapshot?.data())!
var friends = [LocalUser]()
let friendsIdents = Array(friendsUID.keys)
for (idx,userID) in friendsIdents.enumerated() {
self.getUser(withUID: userID, completion: { (usr) in
var tempUser: LocalUser
tempUser = usr
friends.append(tempUser)
self.semaphore.signal()
})
if idx == friendsIdents.endIndex-1 {
print("friends at for loop completion:", friends.count)
completion(friends)
}
self.semaphore.wait()
}
}
}
friendsUID is a dict with each friend's uid as a key and true as the value. Since I only need the keys, I store them in the array friendsIdents. Function getUser searches the passed uid in firestore and creates the corresponding LocalUser (usr). This finally gets appended in friends array.
You should almost never have a semaphore.wait() on the main thread. Unless you expect to wait for < 10ms.
Instead, consider dispatching your friends list processing to a background thread. The background thread can perform the dispatch to your database/api and wait() without blocking the main thread.
Just make sure to use DispatchQueue.main.async {} from that thread if you need to trigger any UI work.
let semaphore = DispatchSemaphore(value: 0)
func getFriends(completion: #escaping ([LocalUser]) -> ()) {
var friendsUID = [String : Any]()
database.collection("friends").document(self.uid).getDocument { (docSnapshot, error) in
if error != nil {
print("Error:", error!)
return
}
DispatchQueue.global(qos: .userInitiated).async {
friendsUID = (docSnapshot?.data())!
var friends = [LocalUser]()
let friendsIdents = Array(friendsUID.keys)
for (idx,userID) in friendsIdents.enumerated() {
self.getUser(withUID: userID, completion: { (usr) in
var tempUser: LocalUser
tempUser = usr
friends.append(tempUser)
self.semaphore.signal()
})
if idx == friendsIdents.endIndex-1 {
print("friends at for loop completion:", friends.count)
completion(friends)
}
self.semaphore.wait()
}
// Insert here a DispatchQueue.main.async {} if you need something to happen
// on the main queue after you are done processing all entries
}
}
I have the following case. The root controller is UITabViewController. There is a ProfileViewController, in it I make an observer that users started to be friends (and then the screen functions change). ProfileViewController can be opened with 4 tabs out of 5, and so the current user can open the screen with the same user in four places. In previous versions, when ProfileViewController opened in one place, I deleted the observer in deinit and did the deletion just by ref.removeAllObservers(), now when the user case is such, I started using handle and delete observer in viewDidDisappear. I would like to demonstrate the code to find out whether it can be improved and whether I'm doing it right in this situation.
I call this function in viewWillAppear
fileprivate func firObserve(_ isObserve: Bool) {
guard let _user = user else { return }
FIRFriendsDatabaseManager.shared.observeSpecificUserFriendshipStart(observer: self, isObserve: isObserve, userID: _user.id, success: { [weak self] (friendModel) in
}) { (error) in
}
}
This is in the FIRFriendsDatabaseManager
fileprivate var observeSpecificUserFriendshipStartDict = [AnyHashable : UInt]()
func observeSpecificUserFriendshipStart(observer: Any, isObserve: Bool, userID: String, success: ((_ friendModel: FriendModel) -> Void)?, fail: ((_ error: Error) -> Void)?) {
let realmManager = RealmManager()
guard let currentUserID = realmManager.getCurrentUser()?.id else { return }
DispatchQueue.global(qos: .background).async {
let specificUserFriendRef = Database.database().reference().child(MainGateways.friends.description).child(currentUserID).child(SubGateways.userFriends.description).queryOrdered(byChild: "friendID").queryEqual(toValue: userID)
if !isObserve {
guard let observerHashable = observer as? AnyHashable else { return }
if let handle = self.observeSpecificUserFriendshipStartDict[observerHashable] {
self.observeSpecificUserFriendshipStartDict[observerHashable] = nil
specificUserFriendRef.removeObserver(withHandle: handle)
debugPrint("removed handle", handle)
}
return
}
var handle: UInt = 0
handle = specificUserFriendRef.observe(.childAdded, with: { (snapshot) in
if snapshot.value is NSNull {
return
}
guard let dict = snapshot.value as? [String : Any] else { return }
guard let friendModel = Mapper<FriendModel>().map(JSON: dict) else { return }
if friendModel.friendID == userID {
success?(friendModel)
}
}, withCancel: { (error) in
fail?(error)
})
guard let observerHashable = observer as? AnyHashable else { return }
self.observeSpecificUserFriendshipStartDict[observerHashable] = handle
}
}
Concerning your implementation of maintaining a reference to each viewController, I would consider moving the logic to an extension of the viewController itself.
And if you'd like to avoid calling ref.removeAllObservers() like you were previously, and assuming that there is just one of these listeners per viewController. I'd make the listener ref a variable on the view controller.
This way everything is contained to just the viewController. It also is potentially a good candidate for creating a protocol if other types of viewControllers will be doing similar types of management of listeners.
I have an application using RealmSwift on iOS that stores a small amount of data. The data is updated via CloudKit push notifications every ~5minutes or so.
It works except my Realm file continuously grows forever until the application no longer has enough memory to launch.
I managed to work around it slightly by using the "writeCopy" function to compact the realm at launch - this mostly works if the app is stopped semi-frequently so that has a chance to run, but if this does not happen and the app continues to update the data from push notifications in the background for a couple of days, the database will end up too large and compacting it takes ages, or it just crashes trying.
I tried reading the "file size" section in the FAQ's but I don't think I fully understand the rules that cause the size to grow. It mentions keeping "old realms" open - does that mean instances of Realm e.g. "let realm = try! Realm()" or does it really mean any open object?
If its the latter then that is annoying as I use notifications on result sets while in the background to determine when to update a complication on a companion watch app.
I really like realm and I want to stick with it but I feel like my somewhat unusual use case (long running app that frequently updates in the background) might mean that I cannot use it.
Edit:
Background updates are happening as a result of the 'didReceiveRemoteNotification' app delegate callback, like this:
func application(_ application: UIApplication, didReceiveRemoteNotification userInfo: [AnyHashable : Any], fetchCompletionHandler completionHandler: #escaping (UIBackgroundFetchResult) -> Void) {
let dict = userInfo as! [String: NSObject]
let notification = CKNotification(fromRemoteNotificationDictionary: dict)
if notification.notificationType == .query {
let notification = notification as! CKQueryNotification
if let recordID = notification.recordID {
let requestOperation = CKFetchRecordsOperation(recordIDs: [recordID])
requestOperation.fetchRecordsCompletionBlock = { records, error in
if let error = error {
os_log("Error fetching record: %#", log: Logger.default, type: .error, String(describing: error))
} else {
autoreleasepool {
let realm = try! Realm()
try! realm.write {
records?.forEach { (key, record) in
switch record.recordType {
case "Freeway":
let dict = RealmFreeway.dictionary(from: record)
realm.create(RealmFreeway.self, value: dict, update: true)
case "Segment":
let dict = RealmSegment.dictionary(from: record)
realm.create(RealmSegment.self, value: dict, update: true)
default:
os_log("Unknown record type: %#", log: Logger.default, type: .error, record.recordType)
}
}
}
}
}
}
CKContainer.default().publicCloudDatabase.add(requestOperation)
}
}
completionHandler(.newData)
}
And the watch complication is updated via RxSwift like so:
Observable.combineLatest(watchViewModel.freewayName.asObservable(), watchViewModel.startName.asObservable(), watchViewModel.endName.asObservable(), watchViewModel.travelTime.asObservable(), watchViewModel.color.asObservable(), watchViewModel.direction.asObservable()) {
freewayName, startName, endName, travelTime, color, direction -> [String:Any] in
guard let freewayName = freewayName,
let startName = startName,
let endName = endName,
let direction = direction else {
return [String:Any]()
}
let watchData:[String:Any] = [
"freewayName": freewayName,
"startName": startName,
"endName": endName,
"travelTime": travelTime,
"color": color.htmlRGBColor,
"direction": direction,
"transfers": session.remainingComplicationUserInfoTransfers
]
return watchData
}
.filter { $0.keys.count > 0 }
.throttle(2.0, scheduler: MainScheduler.instance)
.subscribe( onNext: { watchData in
let MIN_DURATION = 24.0 * 60.0 * 60.0 / 50.0 // 50 guaranteed updates per day...
var timeSinceLastUpdate = MIN_DURATION
let now = Date()
if let lastUpdated = self.lastComplicationUpdate {
timeSinceLastUpdate = now.timeIntervalSince(lastUpdated)
}
// Should we use a complication update or an application context update?
let complicationUpdate = timeSinceLastUpdate >= MIN_DURATION
// Send the data via the appropriate method.
if complicationUpdate {
session.transferCurrentComplicationUserInfo(watchData)
} else {
try? session.updateApplicationContext(watchData)
}
self.lastComplicationUpdate = now
})
.addDisposableTo(bag)