In all iOS classes that use Firebase you will have code like this,
private func clearObservations() {
// your method for clearing observations, probably something like
blah blah. removeAllObservers()
}
In view controllers, it's essential that you call this in viewWillDisappear (or viewDidDisappear)
override func viewWillDisappear(_ animated: Bool) {
super.viewWillDisappear(animated)
clearObservations()
}
That's fine.
Assume that you have created an observation in a UITableViewCell.
What is the best place in a cell to "clear observations" ?
Note that prepareForReuse is useless, try it.
The only approach we've found is
override func willMove(toSuperview newSuperview: UIView?) {
if newSuperview == nil {
clearObservations()
}
super.willMove(toSuperview: newSuperview)
}
Seems flakey/bizarre though.
What's the deal on this?
Update
Note while "XY Answers" are interesting and informative, if anyone knows the answer to the question that would be great also!
Preface
This was an attempt to answer the question but the question was misunderstood. I'll leave it here as it does have some relevance regarding observers, handles and tableView cell interaction.
While you can go through those gyrations, it's not really needed in most use cases.
For example, if you add and observer to a node, there wouldn't necessarily be a someRef? variable hanging around. So here we are watching the Posts node for new posts
let postsRef = self.ref.child("Posts")
postsRef.observe(.childAdded, with: { snapshot in
print(snapshot) //add the post to the dataSource and reloadTableview/cell
})
Here's another example of watching for any posts that are changed by uid_2
let postsRef = self.ref.child("Posts")
let queryRef = postsRef.queryOrdered(byChild: "poster_id").queryEqual(toValue: "uid_2")
queryRef.observe(.childChanged) { (snapshot) in
print(snapshot) //change the post in the dataSource and reloadTableview/cell
}
No class vars are needed for this functionality and nothing needs be nil'd. The point here being that you do not have to have class vars to get observing functionality and you do not need to keep a handle for every observer (keep reading)
In view controllers, it's essential that you call this
(someRef?.removeAllObservers()) in viewWillDisappear (or Did)..
will use Firebase in the cells of tables.
To clarify; I wouldn't want to put Firebase observers in the cells of tables. The observers should be in whichever viewController controls the tableView that has cells. Cells should pull data from the dataSource array (which is backed by Firebase)
There are some circumstances where you may want to remove all observers, again no need to have a class var or a need to nil a var.
let postsRef = self.ref.child("Posts")
postsRef.removeAllObservers()
There are times when a specific observer needs to be removed (in the case where a node has observers on it's child nodes for example), and in those cases, we store a handle to that observer as say, a class var (keeping them in an array is a tidy way to do it)
class ViewController: UIViewController {
var myPostHandle : DatabaseHandle?
func addObserver() {
let postsRef = self.ref.child("Posts")
self.myPostHandle = postsRef.observe(.childAdded, with: { snapshot in
print(snapshot)
})
func stopObserving() {
if self.myPostHandle != nil {
let postsRef = self.ref.child("Posts")
postsRef.removeObserver(withHandle: self.myPostHandle) //remove only the .childAdded observer
}
}
}
Again though, once the observer is removed, the handle would go out of scope once the class closes.
Tableviews that contain cells are backed by a dataSource and that dataSource get's it's data from firebase. When something is added, changed or removed from Firebase, your app is notified and the array is updated and then the cell refreshed. No need for an observer in the cell itself.
There's no need to add dozens of observers (in the cells) - add one central observer and let it keep the array current. Refresh tableView only when something changes.
EDIT
To Address a comment regarding the use of removeAllObservers: code is worth 1000 words:
Create a new Firebase project with two button actions. Here's the code for button0 which adds an observer to a node:
func button0() {
let testRef = self.ref.child("test_node")
testRef.observe( .value) { snapshot in
print(snapshot)
}
}
when this button0 is clicked, from there on, any adds, changes, or deletes to the test node will print it's contents to the log.
func button1() {
let testRef = self.ref.child("test_node")
testRef.removeAllObservers()
}
This will remove all observers for the node specified. Once clicked, no events will print to the console.
Try it!
It is not right to clear observations in cell and therefore there is not a best place to do it in cell, because, firstly, this approach contradicts MVC pattern. Views only responsible for displaying content and they should only contain code that describes how they must be draw. And in the view controller you give the content for showing by views. Usually content has provided by your model. So controller connects views and model. In your case, when you place clearObservations() in cell class, you also have someRef as a class property, so you have a model in your view class and this is incorrect.
Secondly, if you try to clear observations in table cell you definitely make logic of showing some content in table in wrong way. Cell only show data that has to be generated by some object that conforms to UITableViewDataSource protocol and implements protocol methods. For instance, in cellForRow method you generate cell and setup it with some content from array. This array is generated from model (Firebase service). Your view controller may be this data source object. You have to include array property to controller class and someRef, than you fill array and reload table data. If controller's view disappeared you clear observations, but you do it only inside view controller (in viewWillDisappear()).
Overall, all manipulations with someRef you should do in view controller and therefore "clear observations" also inside controller.
Related
How can I stop a diffable data source scrolling the view to the top after applying the snapshot. I currently have this...
fileprivate func configureDataSource() {
self.datasource = UICollectionViewDiffableDataSource<Section, PostDetail>(collectionView: self.collectionView) {
(collectionView: UICollectionView, indexPath: IndexPath, userComment: PostDetail) -> UICollectionViewCell? in
guard let cell = collectionView.dequeueReusableCell(withReuseIdentifier: PostDetailCell.reuseIdentifier, for: indexPath) as? PostDetailCell else { fatalError("Cannot create cell")}
cell.user = self.user
cell.postDetail = userComment
cell.likeCommentDelegate = self
return cell
}
var snapshot = NSDiffableDataSourceSnapshot<Section, PostDetail>()
snapshot.appendSections([.main])
snapshot.appendItems(self.userComments)
self.datasource.apply(snapshot, animatingDifferences: true)
}
fileprivate func applySnapshot() {
//let contentOffset = self.collectionView.contentOffset
var snapshot = NSDiffableDataSourceSnapshot<Section, PostDetail>()
snapshot.appendSections([.main])
snapshot.appendItems(self.userComments)
self.datasource.apply(snapshot, animatingDifferences: false)
//self.collectionView.contentOffset = contentOffset
}
store the offset, then reapply it. Sometimes it works perfectly and sometimes the view jumps. Is there a better way of doing this?
The source of this problem is probably your Item identifier type - the UserComment.
Diffable data source uses the hash of your item identifier type to detect if it is a new instance or an old one which is represented currently.
If you implement Hashable protocol manually, and you use a UUID which is generated whenever a new instance of the type is initialized, this misguides the Diffable data source and tells it this is a new instance of item identifier. So the previous ones must be deleted and the new ones should be represented. This causes the table or collection view to scroll after applying snapshot.
To solve that replace the uuid with one of the properties of the type that you know is unique or more generally use a technique to generate the same hash value for identical instances.
So to summarize, the general idea is to pass instances of the item identifiers with the same hash values to the snapshot to tell the Diffable data source that these items are not new and there is no need to delete previous ones and insert these ones. In this case you will not encounter unnecessary scrolls.
Starting from iOS 15
dataSource.applySnapshotUsingReloadData(snapshot, completion: nil)
Resets the UI to reflect the state of the data in the snapshot without computing a diff or animating the changes
First up: in most cases #Amirrezas answer will be the correct reason for the problem. In my case it was not the item, but the section identifier that caused the problem. That was Hashable and Identifiable with correct values, but it was a class, and therefore the hash functions were never called. Took me a while to spot that problem. Changing to a struct (and therefore adopting some things ;) ) helped in my case.
For reference here's a link to the topic on the Apple-Dev forums: https://developer.apple.com/forums/thread/657499
Hope my answer helps somebody :)
You'd think that any of these methods would work:
https://developer.apple.com/documentation/uikit/uicollectionviewdelegate/1618007-collectionview
https://developer.apple.com/documentation/uikit/uicollectionviewlayout/1617724-targetcontentoffset
But (in my case) they did not. You might get more mileage out of them, I am doing some crazy stuff with a custom UICollectionViewCompositionalLayout
What I did get to work is manually setting the offset in my custom layout class:
override func finalizeCollectionViewUpdates() {
if let offset = collectionView?.contentOffset {
collectionView?.contentOffset = targetContentOffset(forProposedContentOffset: offset)
}
super.finalizeCollectionViewUpdates()
}
where I have targetContentOffset also overridden and defined (I tried that first, didn't work, figured it was cleanest to just use that here. I suspect if you define targetContentOffset on the delegate without overriding it in the layout the above will also work, but you already need a custom layout to get this far so it's all the same.)
I have two Views:
UITableViewController (View A)
UIViewController (View B)
I was wondering, if it's possible to load and setup the table from View B and then segue to View A, when the loading is done. I need this, since the Table View loads Data from Core Data and that takes some time; I would then show a Loading Animation or something. I have a function called loadData() in View A, which fetches all Elements from Core Data and then calls tableView.reloadData().
Does anyone know, how I could implement this? Or should I somehow show the loading View directly from View A with a SubView or something?
Remember to not think about the specifics but instead, think generally:
You want to move from one VC to another and you have some data that needs to be fetched asynchronically. Let's assume you can't know how long it will take.
My suggestion is to contain all data fetching related to a VC inside that VC itself (or services/facades related to it). So basically you should present the UITableViewController and then have it fetch the data while showing skeleton-cells/spinner/etc.
You want to have separation of concerns which means you don't want your ViewController to handle data related to another view controller.
Think about the following use-case: if you have code to fetch data in the previous VC, before presenting the TVC, what happens when you need to re-fetch the data or refresh something? You will have to duplicate the code in both the VC and the TVC.
That's why it's suggested to keep data fetching inside the view controller that needs it.
If, for some reason, you still want to have your answer for this specific question:
You can have the initial VC create the TVC, but not present it yet, call its methods to fetch the data, and have it send a callback (closure/delegate/etc) when it's done fetching. When the fetching is done, present the TVC.
Here is a quick example:
class MyTableVC: UITableViewController {
private var myData: [Int] = []
public func fetchData(completion: () -> Void) {
//Fetch data asyncly
myData = [1, 2 ,3]
completion()
}
}
class MyVC: ViewController {
private func loadTableVC() {
let tableVC = MyTableVC()
tableVC.fetchData { [weak self] in
self?.present(tableVC, animated: true, completion: nil)
}
}
}
Again, I wouldn't use this due to having tight coupling between the 2 view controllers, but it's always up to you to decide how to design your code.
I have a View-Hierarchy like this:
UIViewController (SingleEventViewController)
UIScrollView (EventScrollView)
UIView (contentView)
3xUITableView (SurePeopleTV, MaybePeopleTV, NopePeopleTV (all inherited from the same UITableView)), & all other UI-Elements
The SingleEventViewController stores one Event (passed within the initializer). (All Events are stored in Core-Data).
The three UITableViews are there for displaying the users which are participating (or not or maybe) at the Event. My question is, what are the possibilities to fill the tableViews with the data and what would you recommend in which situation.
Currently I have a property parentVC: SingleEventViewController in all Subviews and get the data like this:
override func loadUsers() {
//class SurePeopleTV
guard let parentController = parentVC else { return }
users = (parentController.thisEvent.eventSureParticipants?.allObjects as! [User])
finishedLoading = true
super.loadUsers()
}
.
func applyDefaultValues() {
//class EventScrollView
guard let parent = parentVC else { return }
titleLabel.text = parent.eventName
}
I'm new to programming but I got a feeling that I should not create a parentVC reference in all of my classes.
An object should not (ideally) know about its parent - if it does they are "tightly coupled". If you change the object's parent, your code may break. In your case, your parent object must have a thisEvent property.
You want your objects to be "loosely coupled", so the object doesn't know about a specific parent object.
In Swift, the usual ways to pass information "back up the chain" is to use the delegate design pattern… https://developer.apple.com/documentation/swift/cocoa_design_patterns or to use closures.
See also https://www.andrewcbancroft.com/2015/04/08/how-delegation-works-a-swift-developer-guide/ for info on delegation
First of all, if you create a reference to the parent ViewController make sure it is weak, otherwise you can run into memory management issues.
Edit: As Ashley Mills said, delegates the way to handle this
The recommended way to pass data between ViewControllers is using something like this
Every time a segue is performed from the view controller this function is in this function is called. This code first checks what identifier the segue has, and if it is the one that you want, you can access a reference to the next view controller and pass data to it.
I searched a lot through Google and SO, so please forgive me, if this question has already been answered!
The problem:
I have a UICollectionView with n UICollectionViewCells. Each cell contains a UIView from a XIB file. The Views are used for data entry, so all cells have a unique reuseIdentifier. Each View has also a unique restorationIdentifier. Everything works in normal usage, but not when it comes to state restoration:
The first 3 or 4 cells are getting restored properly because they are visible on the screen on startup, but the remaining cells, which are not visble, are not getting restored.
Current solution:
So I've discovered so far that a View is only restored if it's added to userinterface at startup.
My current working solution is to set the height of all cells to 1 in the process of restoring. Now every cell is loaded and all views are restored.
When applicationFinishedRestoringState() is called, I reload the CollectionView with the correct height.
Now my question is: I'm not happy with this solution, is there a more clean way to achieve restoring of all the UIViews?
I think you are getting a bit confused between your data model and your views. When first initialised, your table view is constructed from a data model, pulling in stored values in order to populate whatever is in each cell. However, your user does not interact directly with the data model, but with the view on the screen. If the user changes something in the table view, you need to signal that change back up to the view controller so that it can record the change to the data model. This means in turn that if the view needs to be recreated the view controller has the information it needs to rebuild whatever was in the table when your app entered the background.
I have put together a simple gitHub repository here: https://github.com/mpj-chandler/StateManagementDemo
This comprises a CustomTableViewController class which manages a standard UITableView populated with CustomTableViewCells. The custom cells contain three switch buttons, allowing the state of each cell to be represented by an array of Boolean values.
I created a delegate protocol for the cells such that if any of the switches is tripped, a signal is sent back to the view controller:
protocol CustomTableViewCellDelegate {
func stateDidChange(sender: CustomTableViewCell) -> Void
}
// Code in CustomTableViewCell.swift:
#objc fileprivate func switched(sender: UISwitch) -> Void {
guard let index : Int = switches.index(of: sender) else { return }
state[index] = sender.isOn
}
// The cell's state is an observed parameter with the following didSet method:
fileprivate var state : [Bool] = Array(repeating: false, count: 3) {
didSet {
if state != oldValue, let _ = delegate {
delegate!.stateDidChange(sender: self)
}
}
}
CustomTableViewController is registered to the CustomTableViewCellDelegate protocol, so that it can record the change in the model as follows:
// Code in CustomTableViewController.swift
//# MARK:- CustomTableViewCellDelegate methods
internal func stateDidChange(sender: CustomTableViewCell) -> Void {
guard let indexPath : IndexPath = tableView.indexPath(for: sender) else { return }
guard indexPath.row < model.count else { print("Error in \(#function) - cell index larger than model size!") ; return }
print("CHANGING MODEL ROW [\(indexPath.row)] TO: \(sender.getState())")
model[indexPath.row] = sender.getState()
}
You can see here that the function is set up to output model changes to the console.
If you run the project in simulator and exit to the home screen and go back again you will see the state of the tableView cells is preserved, because the model reflects the changes that were made before the app entered the background.
Hope that helps.
In my app I fetch live data like this:
//Firebase
var ref: FIRDatabaseReference?
var handle: FIRDatabaseHandle?
override func viewDidLoad() {
ref = FIRDatabase.database().reference()
handle = ref?.child("posts").child(String(itemId)).observe(.childChanged, with: { (snapShot) in
if let item = snapShot.value as? String {
print(item)
}
})
.....
Now reading the firebase docs I see this:
Observers don't automatically stop syncing data when you leave a ViewController. If an observer isn't properly removed, it continues to sync data to local memory.
So I added this function that gets fired when I exit the VC:
#IBAction func backButtonDidTouch(_ sender: AnyObject) {
if let handle = handle {
ref?.removeObserver(withHandle: handle)
}
showNavBar = true
_ = navigationController?.popViewController(animated: true)
}
But I can also call removeAllObservers() insetad of removeObserver() and the docs also says:
Calling removeObserverWithHandle or removeAllObservers on a listener does not automatically remove listeners registered on its child nodes; you must also keep track of those references or handles to remove them.
So looking at my code am I doing it right? I dont want to keep data syncing between my app and firebase when I exit my VC
You seem to be calling an observer on the specific post, but you are removing the observer from the parent reference. As the documentation states, removing a listener from a reference does not clear the observers from the children, hence I believe you have not removed the observer as you intended.
I have run into this issue myself. Particularly when I log a user out, for a brief moment, the presenting view controller is trying to read from firebase and crashes. What I have done is define a set of type DatabaseReference among the singleton I am using. And where I call
ref.observe(.value) {(snapshot) in
singleton.refsUsed.insert(snapshot.ref)
...
}
Then upon logging out, before I dismiss the current view controller, I am iterating over all the items in the reference set and removing all observers.