Endpoints 2.2.1

Endpoints 2.2.1

LangLanguage SwiftSwift
License MIT
ReleasedLast Release Dec 2022
SPMSupports SPM

Maintained by mariohahn, Peter Weishapl, Thomas Koller, Robin Mayerhofer, Tailored Apps.

Endpoints 2.2.1

  • By
  • Peter Weishapl, Robin Mayerhofer, Dominik Arnhof and Alexander Kauer

CocoaPods Compatible Carthage compatible Swift Package Manager compatible Platform


Endpoints makes it easy to write a type-safe network abstraction layer for any Web-API.

It requires Swift 5, makes heavy use of generics (and generalized existentials) and protocols (and protocol extensions). It also encourages a clean separation of concerns and the use of value types (i.e. structs).



Here's how to load a random image from Giphy.

// A client is responsible for encoding and parsing all calls for a given Web-API.
let client = AnyClient(baseURL: URL(string: "https://api.giphy.com/v1/")!)

// A call encapsulates the request that is sent to the server and the type that is expected in the response.
let call = AnyCall<DataResponseParser>(Request(.get, "gifs/random", query: ["tag": "cat", "api_key": "dc6zaTOxFJmzC"]))

// A session wraps `URLSession` and allows you to start the request for the call and get the parsed response object (or an error) in a completion block.
let session = Session(with: client)

// enable debug-mode to log network traffic
session.debug = true

// start call
session.start(call: call) { result in
    result.onSuccess { value in
        //value is an object of the type specified in `Call`
    }.onError { error in
        //something went wrong

Response Parsing

A call is supposed to know exactly what response to expect from its request. It delegates the parsing of the response to a ResponseParser.

Some built-in types already adopt the ResponseParser protocol (using protocol extensions), so you can for example turn any response into a JSON array or dictionary:

// Replace `DataResponseParser` with any `ResponseParser` implementation
let call = AnyCall<DictionaryParser<String, Any>>(Request(.get, "gifs/random", query: ["tag": "cat", "api_key": "dc6zaTOxFJmzC"]))
session.start(call: call) { result in
    result.onSuccess { value in
        //value is now a JSON dictionary 🎉

let call = AnyCall<JSONParser<GiphyGif>>(Request(.get, "gifs/random", query: ["tag": "cat", "api_key": "dc6zaTOxFJmzC"]))
session.start(call: call) { result in
    result.onSuccess { value in
        //value is now a `GiphyGif` dictionary 🎉

Provided ResponseParsers

Look up the documentation in the code for further explanations of the types.

  • DataResponseParser
  • DictionaryParser
  • JSONParser
  • NoContentParser
  • StringConvertibleParser
  • StringParser

JSON Codable Integration

Endpoints has a built in JSON Codable support.


The ResponseParser responsible for handling decodable types is the JSONParser.
The JSONParser uses the default JSONDecoder(), however, the JSONParser can be subclassed, and the jsonDecoder can be overwritten with your configured JSONDecoder.

// Decode a type using the default decoder
struct GiphyCall: Call {
    typealias Parser = JSONParser<GiphyGif>

// custom decoder

struct GiphyParser<T>: JSONParser<T> {
    override public var jsonDecoder: JSONDecoder {
        let decoder = JSONDecoder()
        // configure...
        return decoder

struct GiphyCall: Call {
    typealias Parser = GiphyParser<GiphyGif>

Every encodable is able to provide a JSONEncoder() to encode itself via the toJSON() method.

Dedicated Calls

AnyCall is the default implementation of the Call protocol, which you can use as-is. But if you want to make your networking layer really type-safe you'll want to create a dedicated Call type for each operation of your Web-API:

struct GetRandomImage: Call {
    typealias Parser = DictionaryParser<String, Any>
    var tag: String
    var request: URLRequestEncodable {
        return Request(.get, "gifs/random", query: [ "tag": tag, "api_key": "dc6zaTOxFJmzC" ])

// `GetRandomImage` is much safer and easier to use than `AnyCall`
let call = GetRandomImage(tag: "cat")

Dedicated Clients

A client is responsible for handling things that are common for all operations of a given Web-API. Typically this includes appending API tokens or authentication tokens to a request or validating responses and handling errors.

AnyClient is the default implementation of the Client protocol and can be used as-is or as a starting point for your own dedicated client.

You'll usually need to create your own dedicated client that either subclasses AnyClient or delegates the encoding of requests and parsing of responses to an AnyClient instance, as done here:

class GiphyClient: Client {
    private let anyClient = AnyClient(baseURL: URL(string: "https://api.giphy.com/v1/")!)
    var apiKey = "dc6zaTOxFJmzC"
    func encode<C: Call>(call: C) -> URLRequest {
        var request = anyClient.encode(call: call)
        // Append the API key to every request
        request.append(query: ["api_key": apiKey]) 
        return request
    public func parse<C : Call>(sessionTaskResult result: URLSessionTaskResult, for call: C) throws -> C.Parser.OutputType {
        do {
            // Use `AnyClient` to parse the response
            // If this fails, try to read error details from response body
            return try anyClient.parse(sessionTaskResult: result, for: call)
        } catch {
            // See if the backend sent detailed error information
                let response = result.httpResponse,
                let data = result.data,
                let errorDict = try? JSONSerialization.jsonObject(with: data, options: .allowFragments) as? [String: Any],
                let meta = errorDict?["meta"] as? [String: Any],
                let errorCode = meta["error_code"] as? String else {
                // no error info from backend -> rethrow default error
                throw error
            // Propagate error that contains errorCode as reason from backend
            throw StatusCodeError.unacceptable(code: response.statusCode, reason: errorCode)

Dedicated Response Types

You usually want your networking layer to provide a dedicated response type for every supported call. In our example this could look like this:

struct RandomImage: Decodable {
    struct Data: Decodable {
        let url: URL
        private enum CodingKeys: String, CodingKey {
            case url = "image_url"
    let data: Data

struct GetRandomImage: Call {
    typealias Parser = JSONParser<RandomImage>


With all the parts in place, users of your networking layer can now perform type-safe requests and get a type-safe response with a few lines of code:

let client = GiphyClient()
let call = GetRandomImage(tag: "cat")
let session = Session(with: client)

session.start(call: call) { result in
    result.onSuccess { value in
        print("image url: \(value.data.url)")
    }.onError { error in
        print("error: \(error)")


There are multiple ways to make performing a call more convenient. You could write a dedicated GiphyCall that creates the correct Client and Session for your users:

protocol GiphyCall: Call {}

extension GiphyCall {
    func start(completion: @escaping (Result<Parser.OutputType>)->()) {
        let client = GiphyClient()
        let session = Session(with: client)
        session.start(call: self, completion: completion)

When GiphyCall is adopted by GetRandomImage instead of Call, performing a request is much simpler:

GetRandomImage(tag: "cat").start { result in ... }

To make it easer to find supported calls, you could namespace your calls using an extension of your Client:

extension GiphyClient {
    struct GetRandomImage: GiphyCall { ... }

Xcode can now help developers find the right Call instance:

GiphyClient.GetRandomImage(tag: "cat").start { result in ... }



pod "Endpoints"


github "tailoredmedia/Endpoints.git"

Swift Package Manager:

.package(url: "https://github.com/tailoredmedia/Endpoints.git", .upToNextMajor(from: "2.0.0"))


To compile examples you need to open the project in Xcode, the dependencies are added via the Swift Package Manager, Xcode will download them automatically.


  • Swift 5
  • iOS 8
  • tvOS 9
  • macOS 10.11
  • watchOS 2.0