Networking with Swift

Introduction

We all know why dependencies are bad, and writing software dependant on other software is not necessarily good. It probably reduces complexity for the time being, and helps you write code faster, but I wouldn’t say it’s exactly the way things are to be done, from my experience.

Writing software independent of 3rd party libraries, helps build robust software that won’t go out of date with time. I’ve used networking libraries in the past, some really excellent ones (cough Alamofire cough) to abstract and make things easier while building the networking layer. But of late, I’m trying to write code with zero to no-dependencies. Especially, when working with the iOS Platform. The iOS platform comes baked in with most, if not all, things that you need to build out your app.

Networking is something most apps need to deal with these days, the apps on your phone have to communicate with the internet one way or the other, process requests/responses, and display information. When that’s being done all over the place in your code, it’s nightmare.

Recently, I came across a Swift talk by Florian Kugler and Chris Eidhof of objc.io where they talk about writing the Networking Layer for an app, and how best to do that. In this post, I intend to explain their process with some fundamental Swift concepts, and how to keep concerns separate and why it’s good.

The Architecture: Model, Resource, Webservice

Architecting a networking layer can be a task, a difficult one at that, especially without direction. Once you know how to do something, doing something isn’t hard. Let’s see one way to architect your networking layer that you can consider as a direction to write your own.

There are three concerns that need to be addressed when dealing with Networking:

1. Send a request to a server for a resource
2. Get a response back
3. Parse responses into native swift types

We’ll see how to address each of this with Swift.

• Send a request to a server for a resource
• Get a response back

Usually when you’re sending a request over to a server, it’s for a resource of some kind. This could be a post, an article, a movie, or anything. Let’s call this a resource from now.

We could handle this by having a class do that for us, say a Webservice class that could load a resource, that returns a response.

• Parse responses into native swift types

Usually the responses we get back from server these days is in JSON. So, the task at hand is to parse JSON into native swift types. There are many ways to do this, but let’s see how to handle it so its nicer and cleaner and in a way you don’t have to deal with it each and every time you need to request for a response.

Let’s assume we have the following milestones:

Milestone One: Get Binary Data

Milestone Two: Data —> JSON —> Native Types

Milestone Three: Clean up

Webservice — Class that handles requests

We need a way to send requests to a server for a resource, don’t we? We’ll call that class Webservice. This class could have a method load that could be responsible to load a resource. We’ll see how to build this out in a while.

Resource

We also need a way to tell the Webservice class to load a resource, we can accomplish that by having a Resource object that encapsulates the request URL and a way to parse the response.

Model — Native Swift Objects with Value Types

Finally, you need native types of the data that you receive from the server to be able to use it across the app. We’ll use structs for each model to address that.

Code

Example:

Let’s assume we have json data being served from a server that looks like this:

Copy 
1[
2  {
3    "title": "Pulp Fiction",
4    "actors": ["Samuel L Jackson", "John Travolta"]
5  },
6  {
7    "title": "Ocean's Eleven",
8    "actors": ["George Clooney", "Brad Pitt", "Matt Damon"]
9  },
10  {
11    "title": "Goodfellas",
12    "actors": ["Robert De Niro", "Ray Liotta"]
13  },
14  {
15    "title": "I Am Sam",
16    "actors": ["Sean Penn"]
17  }
18]

For the sake of this example, we’ll put this data in a file called movies.json, and try to build a networking layer in Swift that parses this data into native types.

Resource

The Resource is a struct that encapsulates the url and a way to parse, a parse function, that helps parse the data from the URL. For example, In our case we could have a MoviesResource that could retrieve movies.

We could model Resource like —>

Copy 
1struct Resource<A> {
2  let url: URL
3  let parse: (Data) -> A?
4}

• URL: A url property that denotes the URL of the resource
• parse function: A function that takes some Data and returns an object of the associated resource.

Let’s write an initialiser for a Resource in an extension like —>

Copy 
1extension Resource {
2  init?(url: URL, parse: @escaping (Any) -> A?) {
3    self.url = url
4    self.parse = parse
5  }
6}

Here, along with the URL, we’re just passing a closure to the parse parameter and assigning it to the resource’s parse property.

Now, a movies resource could be initialised like —>

Copy 
1let url = URL(string: "http://localhost:8080/movies.json")!
2let moviesResource = Resource<Data>(url: url) { data in
3  return data
4}

Webservice

Now that we’ve seen how we could create a Resource type for our data, we need a way for us to fetch the data from the server.

Luckily the URLSession class offers a way for us to do this easily without having to use any third party libraries, we’ll wrap this up in a function within a Webservice class so we could use it.

We could write this like —>

Copy 
1final class Webservice {
2  func load<T>(resource: Resource<T>, completion: @escaping (Any) -> ()) {
3    URLSession.shared.dataTask(with: resource.url) { (data, _ \_) in
4      completion(data)
5    }.resume()
6  }
7}

We’re simply using a shared URLSession object to call the following method

Copy 
1URLSession.shared.dataTask(with: URL, completionHandler: (Data?, URLResponse?, Error?) -> Void)

with the resource’s URL parameter and call the completion with data, this will just call the closure with whatever data it retrieves at this point.

We could now use the Webservice by creating an instance of it like —>

Copy 
1let sharedWS = Webservice()
2sharedWS.load(resource: moviesResource) { (result) in
3  print(result)
4}

This should return whatever the size of the data it retrieved was, when you run it a playground, you get something like this —>

Copy 
1Optional(324 bytes)

Now, this means we have a service trying to communicate with the server and returning response, but the response is in binary and not useful for us.

🏁 We have now reached Milestone One.

We’ll see how to get meaningful data and parse it to native types so we can use it.

Models

Looking at the JSON data, we figure that we need a Movie type to represent the movies, and possibly an Actor type with a name property that can hold their names.

We could model this with native types like —>

Copy 
1// Actor
2struct Actor {
3  let name: String
4}
5    
6// Movie
7struct Movie {
8  let title: String
9  let actors: [Actor]
10}

We now need a way to model JSON data with native types, we could use a type alias for a dictionary like —>

Copy 
1typealias JSONDictionary = [String: AnyObject]

We can now write an initialiser for a Movie like this —>

Copy 
1extension Movie {
2  init?(dictionary: JSONDictionary) {
3    guard let title = dictionary["title"] as? String,
4      let actors = dictionary["actors"] as? [String] else { return nil }
5    self.title = title
6    self.actors = actors.flatMap(Actor.init)
7  }
8}

What’s happening here is pretty straightforward —>

The guard statement pulls out our fields title and actors from the dictionary, which is of type JSONDictionary and assigns it to the respective properties. We also flatMap the actors string and pass each string to the Actor’s init to instantiate actor objects within the Movie.

Now that we have a way to create Movie objects, let’s rewrite our moviesResource to return them —>

Copy 
1let moviesResource = Resource<[Movie]>(url: url) { data in
2  let json = try? JSONSerialization.jsonObject(with: data, options: [])
3  guard let dictionaries = json as? [JSONDictionary] else { return nil }
4  return dictionaries.flatMap(Movie.init)
5}

We do the following here:

• Instantiate moviesResource as a Resource that returns array of Movie objects
• Use JSONSerialization class to convert our data (Data) object into json JSON
• Typecast json into an array of JSONDictioanry objects
• flatMap over the dictionaries array to instantiate a Movie with each block of json

We now have a process setup that can parse the data into our required objects, but we still haven’t updated our load method to make use of the parse method that we’ve modified above.

This can look like —>

Copy 
1final class Webservice {
2  func load<T>(resource: Resource<T>, completion: @escaping (Any) -> ()) {
3    URLSession.shared.dataTask(with: resource.url) { (data, \_, \_) in
4      completion(data.flatMap(resource.parse))
5    }.resume()
6  }
7}

We modify our load method for a resource type T inside which we simply try to flatMap over the stream of data that gets loaded as Data, thanks to URSession, and pass the resource’s parse method to it so the parse functionality we’ve written is useful, and send it to the completion handler.

Now, let’s see what data we’re able to load with our web service —>

Copy 
1sharedWS.load(resource: moviesResource) { (result) in
2     print(result)
3    }

If you run this in a playground, you should see something like this —>

Copy 
1Optional(
2    [__lldb_expr_55.Movie(
3        title: "Pulp Fiction",
4        actors: [
5            __lldb_expr_55.Actor(name: "Samuel L Jackson"), __lldb_expr_55.Actor(name: "John Travolta")]),
6            __lldb_expr_55.Movie(title: "Ocean\'s Eleven", actors: [__lldb_expr_55.Actor(name: "George Clooney"),
7            __lldb_expr_55.Actor(name: "Brad Pitt"), __lldb_expr_55.Actor(name: "Matt Damon")]),
8            __lldb_expr_55.Movie(title: "Goodfellas", actors: [__lldb_expr_55.Actor(name: "Robert Di Niro")]),
9            __lldb_expr_55.Movie(title: "I Am Sam", actors: [__lldb_expr_55.Actor(name: "Sean Penn")
10            ])
11      ])

Which means good news. You’re now able to get native objects from the json now.

We could refactor the load method to give us something nicer —>

Copy 
1sharedWS.load(resource: moviesResource) { (result) in
2  guard let movies = result as? [Movie] else { return }
3  for movie in movies {
4    print("\nMovie: \(movie.title)")
5    for actor in movie.actors {
6      print("Actor: \(actor.name)")
7    }
8  }
9}

You’ll see this prettified output —>

Copy 
1Movie: Pulp Fiction
2Actor: Samuel L Jackson
3Actor: John Travolta
4     
5Movie: Ocean's Eleven
6Actor: George Clooney
7Actor: Brad Pitt
8Actor: Matt Damon
9     
10Movie: Goodfellas
11Actor: Robert Di Niro
12     
13Movie: I Am Sam
14Actor: Sean Penn

🏁 We have now reached Milestone Two.

Refactor

Now that we’ve seen how to get to what we need, we could get to cleaning up the code a bit, and refactoring it to make it much nicer.

We could start by moving our moviesResource into the Movie struct to make it more concise using an all property —>

Copy 
1struct Movie {
2  let title: String
3  let actors: [Actor]
4     
5  static let all = Resource<[Movie]>(url: url) { data in
6    let json = try? JSONSerialization.jsonObject(with: data, options: [])
7    guard let dictionaries = json as? [JSONDictionary] else { return nil }
8    return dictionaries.flatMap(Movie.init)
9  }
10}

We could also simplify this further by abstracting away the JSON decoding part to the Resource type by creating an extension —>

Copy 
1extension Resource {
2  init(url: URL, parseJSON: @escaping (Any) -> A?) {
3    self.url = url
4    self.parse = { data in
5      let json = try? JSONSerialization.jsonObject(with: data, options: [])
6      return json.flatMap(parseJSON)
7    }
8  }
9}

Here, we convert data object into JSON and flatMap over it to recursively parse the closure to parse the json data, now our Resource could become simpler —>

Copy 
1struct Movie {
2  let title: String
3  let actors: [Actor]
4     
5  static let all = Resource<[Movie]>(url: url) { json in
6    guard let dictionaries = json as? [JSONDictionary] else { return nil }
7    return dictionaries.flatMap(Movie.init)
8  }
9}

🏁 We have now reached our final milestone, Milestone Three.

Now, we have code that’s cleaner, concise and keeps concerns separate. This pattern can be followed with whatever Resource that you want to work with.

You can download the files we worked with incase you want to try it our yourself:

• movies.json
• Neworking.playground

So, that was about how to write a Networking layer without any third-party libraries. Feel free to leave any feedback or questions!