Discover ways to load a distant picture into an UIImageView asynchronously utilizing URLSessionDownloadTask and the Mix framework in Swift.
iOS
A easy picture downloader
Downloading a useful resource from an URL looks like a trivial process, however is it actually that straightforward? Properly, it relies upon. If you must obtain and parse a JSON file which is only a few KB, then you’ll be able to go together with the classical means or you need to use the brand new dataTaskPublisher
technique on the URLSession object from the Mix framework.
Dangerous practices ⚠️
There are some fast & soiled approaches that you need to use to get some smaller knowledge from the web. The issue with these strategies is that you must deal loads with threads and queues. Happily utilizing the Dispatch framework helps loads, so you’ll be able to flip your blocking features into non-blocking ones. 🚧
let url = URL(string: "https://jsonplaceholder.typicode.com/todos/1")!
do {
let content material = attempt String(contentsOf: url)
print(content material)
let knowledge = attempt Knowledge(contentsOf: url)
}
catch {
print(error.localizedDescription)
}
DispatchQueue.world().async { [weak self] in
do {
let content material = attempt String(contentsOf: url)
DispatchQueue.important.async {
print(content material)
}
}
catch {
print(error.localizedDescription)
}
}
Apple made an vital notice on their official Knowledge documentation, that you must NOT use these strategies for downloading non-file URLs, however nonetheless persons are instructing / utilizing these dangerous practices, however why? 😥
Do not use this synchronous technique to request network-based URLs.
My recommendation right here: all the time use the URLSession to carry out community associated data-transfers. Creating an information process is easy, it is an asynchronous operation by default, the callback runs on a background thread, so nothing might be blocked by default. Trendy networking APIs are actual good on iOS, in 99% of the circumstances you will not want Alamofire anymore for these form of duties. Say no to dependencies! 🚫
URLSession.shared.dataTask(with: url) { knowledge, response, error in
DispatchQueue.important.async {
}
}.resume()
It is also price to say if you’ll want to use a special HTTP technique (apart from GET), ship particular headers (credentials, settle for insurance policies, and so on.) or present further knowledge within the physique, you’ll want to assemble an URLRequest
object first. You may solely ship these customized requests utilizing the URLSession
APIs.
on Apple platforms you aren’t allowed to make use of the unsecure HTTP protocol anymore. If you wish to attain a URL with out the safe layer (HTTPS) you must disable App Transport Safety.
The issue with knowledge duties
What about massive recordsdata, corresponding to pictures? Let me present you a couple of tutorials earlier than we dive in:
With all due respect, I believe all of those hyperlinks above are actually dangerous examples of loading distant pictures. Certain they do the job, they’re additionally very straightforward to implement, however possibly we must always cowl the entire story… 🤐
For small interactions with distant servers, you need to use the URLSessionDataTask class to obtain response knowledge into reminiscence (versus utilizing the URLSessionDownloadTask class, which shops the information on to the file system). An information process is good for makes use of like calling an online service endpoint.
What’s distinction between URLSessionDataTask vs URLSessionDownloadTask?
If we learn the docs fastidiously, it turns into clear that knowledge process is NOT the fitting candidate for downloading massive property. That class is designed to request solely smaller objects, because the underlying knowledge goes to be loaded into reminiscence. Then again the obtain process saves the content material of the response on the disk (as a substitute of reminiscence) and you’ll obtain an area file URL as a substitute of a Knowledge object. Seems that transferring from knowledge duties to obtain duties could have a HUGE impression in your reminiscence consumption. I’ve some numbers. 📈
I downloaded the following picture file (6000x4000px 💾 13,1MB) utilizing each strategies. I made a model new storyboard primarily based Swift 5.1 challenge. The essential RAM utilization was ~52MB, after I fetched the picture utilizing the URLSessionDataTask
class, the reminiscence utilization jumped to ~82MB. Turning the information process right into a obtain process solely elevated the bottom reminiscence dimension by ~4MB (to a complete ~56MB), which is a major enchancment.
let url = URL(string: "https://pictures.unsplash.com/photo-1554773228-1f38662139db")!
URLSession.shared.dataTask(with: url) { [weak self] knowledge, response, error in
guard let knowledge = knowledge else {
return
}
DispatchQueue.important.async {
self?.imageView.picture = UIImage(knowledge: knowledge)
}
}.resume()
URLSession.shared.downloadTask(with: url) { [weak self] url, response, error in
guard
let cache = FileManager.default.urls(for: .cachesDirectory, in: .userDomainMask).first,
let url = url
else {
return
}
do {
let file = cache.appendingPathComponent("(UUID().uuidString).jpg")
attempt FileManager.default.moveItem(atPath: url.path,
toPath: file.path)
DispatchQueue.important.async {
self?.imageView.picture = UIImage(contentsOfFile: file.path)
}
}
catch {
print(error.localizedDescription)
}
}.resume()
After I rendered the picture utilizing an UIImageView
the reminiscence footprint was ~118MB (whole: ~170MB) for the information process, and ~93MB (whole: ~145MB) for the obtain process. This is a fast abstract:
- Knowledge process: ~30MB
- Knowledge process with rendering: ~118MB
- Obtain process: ~4MB
- Obtain process with rendering: ~93MB
I hope you get my level. Please do not forget that the Basis networking layer comes with 4 forms of session duties. You must all the time use the fitting one that matches the job. We will say that the distinction between URLSessionDataTask vs URLSessionDownloadTask is: quite a lot of reminiscence (on this case about 25MB of RAM).
You should use Kingfisher or SDWebImage to obtain & manipulate distant pictures..
You would possibly say that that is an edge case since many of the pictures (even HD ones) are most a couple of hundred kilobytes. Nonetheless, my takeaway right here is that we are able to do higher, and we must always all the time achieve this if doable. 🤓
Downloading pictures utilizing Mix
WWDC19, Apple introduced the Mix framework, which brings us a couple of new extensions for some Basis objects. Trendy instances require fashionable APIs, proper? If you’re already aware of the brand new SDK that is good, but when you do not know what the heck is that this declarative practical reactive insanity, you must learn my complete tutorial in regards to the Mix framework.
The primary model of Mix shipped with a pleasant dataTaskPublisher
extension technique for the URLSession
class. Wait, the place are the others? No obtain process writer? What ought to we do now? 🤔
Tips on how to write a customized Writer?
SwiftLee has a pleasant tutorial about Mix that may assist you numerous with UIControl occasions. One other nice learn (even higher than the primary one) by Donny Wals is about understanding Publishers and Subscribers. It is a actually well-written article, you must positively examine this one, I extremely advocate it. 🤘🏻
Now let’s begin creating our personal DownloadTaskPublisher
. When you command + click on on the dataTaskPublisher
technique in Xcode, you’ll be able to see the corresponding interface. There may be additionally a DataTaskPublisher
struct, proper under. Primarily based on that template we are able to create our personal extension. There are two variants of the identical knowledge process technique, we’ll replicate this habits. The opposite factor we want is a DownloadTaskPublisher
struct, I am going to present you the Swift code first, then we’ll focus on the implementation particulars.
extension URLSession {
public func downloadTaskPublisher(for url: URL) -> URLSession.DownloadTaskPublisher {
self.downloadTaskPublisher(for: .init(url: url))
}
public func downloadTaskPublisher(for request: URLRequest) -> URLSession.DownloadTaskPublisher {
.init(request: request, session: self)
}
public struct DownloadTaskPublisher: Writer {
public typealias Output = (url: URL, response: URLResponse)
public typealias Failure = URLError
public let request: URLRequest
public let session: URLSession
public init(request: URLRequest, session: URLSession) {
self.request = request
self.session = session
}
public func obtain<S>(subscriber: S) the place S: Subscriber,
DownloadTaskPublisher.Failure == S.Failure,
DownloadTaskPublisher.Output == S.Enter
{
let subscription = DownloadTaskSubscription(subscriber: subscriber, session: self.session, request: self.request)
subscriber.obtain(subscription: subscription)
}
}
}
A Writer can ship an Output or a Failure message to an hooked up subscriber. It’s important to create a brand new typealias for every sort, since they each are generic constraints outlined on the protocol stage. Subsequent, we’ll retailer the session and the request objects for later use. The final a part of the protocol conformance is that you must implement the obtain<S>(subscriber: S)
generic technique. This technique is answerable for attaching a brand new subscriber by a subscription object. Ummm… what? 🤨
A writer/subscriber relationship in Mix is solidified in a 3rd object, the subscription. When a subscriber is created and subscribes to a writer, the writer will create a subscription object and it passes a reference to the subscription to the subscriber. The subscriber will then request quite a lot of values from the subscription with a purpose to start receiving these values.
A Writer
and a Subscriber
is related by a Subscription
. The Writer solely creates the Subscription and passes it to the subscriber. The Subscription comprises the logic that’ll fetch new knowledge for the Subscriber. The Subscriber receives the Subscription, the values and the completion (success or failure).
- The Subscriber subscribes to a Writer
- The Writer creates a Subscription
- The Writer offers this Subscription to the Subscriber
- The Subscriber calls for some values from the Subscription
- The Subscription tries to gather the values (success or failure)
- The Subscription sends the values to the Subscriber primarily based on the demand coverage
- The Subscription sends a Failure completion to the Subscriber if an error occurs
- The Subscription sends completion if no extra values can be found
Tips on how to make a customized Subscription?
Okay, time to create our subscription for our little Mix primarily based downloader, I believe that you’ll perceive the connection between these three objects if we put collectively the ultimate items of the code. 🧩
extension URLSession {
ultimate class DownloadTaskSubscription<SubscriberType: Subscriber>: Subscription the place
SubscriberType.Enter == (url: URL, response: URLResponse),
SubscriberType.Failure == URLError
{
non-public var subscriber: SubscriberType?
non-public weak var session: URLSession!
non-public var request: URLRequest!
non-public var process: URLSessionDownloadTask!
init(subscriber: SubscriberType, session: URLSession, request: URLRequest) {
self.subscriber = subscriber
self.session = session
self.request = request
}
func request(_ demand: Subscribers.Demand) {
guard demand > 0 else {
return
}
self.process = self.session.downloadTask(with: request) { [weak self] url, response, error in
if let error = error as? URLError {
self?.subscriber?.obtain(completion: .failure(error))
return
}
guard let response = response else {
self?.subscriber?.obtain(completion: .failure(URLError(.badServerResponse)))
return
}
guard let url = url else {
self?.subscriber?.obtain(completion: .failure(URLError(.badURL)))
return
}
do {
let cacheDir = FileManager.default.urls(for: .cachesDirectory, in: .userDomainMask).first!
let fileUrl = cacheDir.appendingPathComponent((UUID().uuidString))
attempt FileManager.default.moveItem(atPath: url.path, toPath: fileUrl.path)
_ = self?.subscriber?.obtain((url: fileUrl, response: response))
self?.subscriber?.obtain(completion: .completed)
}
catch {
self?.subscriber?.obtain(completion: .failure(URLError(.cannotCreateFile)))
}
}
self.process.resume()
}
func cancel() {
self.process.cancel()
}
}
}
A Subscriber has an Enter and a Failure sort. A subscriber can solely subscribe to a writer with the identical varieties. The Writer’s Output & Failure varieties must be similar with the Subscription Enter and Failure varieties. This time we won’t go together with an associatedType, however now we have to create a generic worth that has a constraint on these necessities by utilizing a the place clause. The rationale behind that is that we do not know what sort of Subscriber will subscribe to this subscription. It may be both a category A
or B
, who is aware of… 🤷♂️
We have now to go a couple of properties within the init technique, retailer them as occasion variables (watch out with lessons, you must use weak if relevant). Lastly we implement the worth request technique, by respecting the demand coverage. The demand is only a quantity. It tells us what number of values can we ship again to the subscriber at most. In our case we’ll have max 1 worth, so if the demand is bigger than zero, we’re good to go. You may ship messages to the subscriber by calling numerous obtain strategies on it.
It’s important to manually ship the completion occasion with the .completed
or the .failure(T)
worth. Additionally now we have to maneuver the downloaded non permanent file earlier than the completion block returns in any other case we’ll utterly lose it. This time I will merely transfer the file to the applying cache listing. As a free of charge cancellation is a good way to finish battery draining operations. You simply must implement a customized cancel()
technique. In our case, we are able to name the identical technique on the underlying URLSessionDownloadTask
.
That is it. We’re prepared with the customized writer & subscription. Wanna attempt them out?
Tips on how to create a customized Subscriber?
As an instance that there are 4 sorts of subscriptions. You should use the .sink
or the .assign
technique to make a brand new subscription, there’s additionally a factor referred to as Topic
, which will be subscribed for writer occasions or you’ll be able to construct your very personal Subscriber
object. When you select this path you need to use the .subscribe
technique to affiliate the writer and the subscriber. You too can subscribe a topic.
ultimate class DownloadTaskSubscriber: Subscriber {
typealias Enter = (url: URL, response: URLResponse)
typealias Failure = URLError
var subscription: Subscription?
func obtain(subscription: Subscription) {
self.subscription = subscription
self.subscription?.request(.limitless)
}
func obtain(_ enter: Enter) -> Subscribers.Demand {
print("Subscriber worth (enter.url)")
return .limitless
}
func obtain(completion: Subscribers.Completion<Failure>) {
print("Subscriber completion (completion)")
self.subscription?.cancel()
self.subscription = nil
}
}
The subscriber above will merely print out the incoming values. We have now to be extraordinarily cautious with reminiscence administration. The acquired subscription might be saved as a powerful property, however when the writer sends a completion occasion we must always cancel the subscription and take away the reference.
When a worth arrives now we have to return a requirement. In our case it actually would not matter since we’ll solely have 1 incoming worth, however if you would like to restrict your writer, you need to use e.g. .max(1)
as a requirement.
Here’s a fast pattern code for all of the Mix subscriber varieties written in Swift 5.1:
class ViewController: UIViewController {
@IBOutlet weak var imageView: UIImageView!
static let url = URL(string: "https://pictures.unsplash.com/photo-1554773228-1f38662139db")!
static var defaultValue: (url: URL, response: URLResponse) = {
let fallbackUrl = URL(fileURLWithPath: "fallback-image-path")
let fallbackResponse = URLResponse(url: fallbackUrl, mimeType: "foo", expectedContentLength: 1, textEncodingName: "bar")
return (url: fallbackUrl, response: fallbackResponse)
}()
@Printed var worth: (url: URL, response: URLResponse) = ViewController.defaultValue
let topic = PassthroughSubject<(url: URL, response: URLResponse), URLError>()
let subscriber = DownloadTaskSubscriber()
var sinkOperation: AnyCancellable?
var assignOperation: AnyCancellable?
var assignSinkOperation: AnyCancellable?
var subjectOperation: AnyCancellable?
var subjectSinkOperation: AnyCancellable?
override func viewDidLoad() {
tremendous.viewDidLoad()
self.sinkExample()
self.assignExample()
self.subjectExample()
self.subscriberExample()
}
func sinkExample() {
self.sinkOperation = URLSession.shared
.downloadTaskPublisher(for: ViewController.url)
.sink(receiveCompletion: { completion in
print("Sink completion: (completion)")
}) { worth in
print("Sink worth: (worth.url)")
}
}
func assignExample() {
self.assignSinkOperation = self.$worth.sink { worth in
print("Assign worth: (worth.url)")
}
self.assignOperation = URLSession.shared
.downloadTaskPublisher(for: ViewController.url)
.replaceError(with: ViewController.defaultValue)
.assign(to: .worth, on: self)
}
func subjectExample() {
self.subjectSinkOperation = self.topic.sink(receiveCompletion: { completion in
print("Topic completion: (completion)")
}) { worth in
print("Topic worth: (worth.url)")
}
self.subjectOperation = URLSession.shared
.downloadTaskPublisher(for: ViewController.url)
.subscribe(self.topic)
}
func subscriberExample() {
URLSession.shared
.downloadTaskPublisher(for: ViewController.url)
.subscribe(DownloadTaskSubscriber())
}
}
That is very nice. We will obtain a file utilizing our customized Mix primarily based URLSession extension.
Remember to retailer the AnyCancellable pointer in any other case the whole Mix operation might be deallocated means earlier than you can obtain something from the chain / stream.
Placing all the pieces collectively
I promised a working picture downloader, so let me clarify the entire circulation. We have now a customized obtain process writer that’ll save our take away picture file regionally and returns a tuple with the file url and the response. ✅
Subsequent I will merely assume that there was a sound picture behind the url, and the server returned a sound response, so I will map the writer’s output to an UIImage
object. I am additionally going to exchange any form of error with a fallback picture worth. In a real-world utility, you must all the time do some further checkings on the URLResponse
object, however for the sake of simplicity I am going to skip that for now.
The very last thing is to replace our picture view with the returned picture. Since this can be a UI process it ought to occur on the primary thread, so now we have to make use of the obtain(on:)
operation to change context. If you wish to study extra about schedulers within the Mix framework you must learn Vadim Bulavin‘s article. It is a gem. 💎
If you’re not receiving values on sure appleOS variations, that is would possibly as a result of there was a change in Mix round December, 2019. You must examine these hyperlinks: link1, link2
Anyway, here is the ultimate Swift code for a doable picture obtain operation, easy & declarative. 👍
class ViewController: UIViewController {
@IBOutlet weak var imageView: UIImageView!
var operation: AnyCancellable?
override func viewDidLoad() {
tremendous.viewDidLoad()
let url = URL(string: "https://pictures.unsplash.com/photo-1554773228-1f38662139db")!
self.operation = URLSession.shared
.downloadTaskPublisher(for: url)
.map { UIImage(contentsOfFile: $0.url.path)! }
.replaceError(with: UIImage(named: "fallback"))
.obtain(on: DispatchQueue.important)
.assign(to: .picture, on: self.imageView)
}
}
Lastly, we are able to show our picture. Ouch, however wait… there’s nonetheless room for enhancements. What about caching? Plus a 6000x4000px image is kind of enormous for a small show, should not we resize / scale the picture first? What occurs if I need to use the picture in an inventory, should not I cancel the obtain duties when the person scrolls? 😳
Possibly I am going to write about these points in an upcoming tutorial, however I believe that is the purpose the place I ought to finish this text. Be at liberty to mess around with my resolution and please share your concepts & ideas with me on Twitter.