I'm trying to migrate my project from Alamofire 4.9 to 5.3 and I'm having a hard time with error handling. I would like to use Decodable as much as possible, but my API endpoints return one JSON structure when everything goes well, and a different JSON structure when there is an error, the same for all errors across all endpoints. The corresponding Codable in my code is ApiError.
I would like to create a custom response serializer that can give me a Result<T, ApiError> instead of the default Result<T, AFError>. I found this article that seems to explain the general process but the code in there does not compile.
How can I create such a custom ResponseSerializer?
I ended up making it work with the following ResponseSerializer:
struct APIError: Error, Decodable {
let message: String
let code: String
let args: [String]
}
final class TwoDecodableResponseSerializer<T: Decodable>: ResponseSerializer {
lazy var decoder: JSONDecoder = {
let decoder = JSONDecoder()
decoder.dateDecodingStrategy = .iso8601
return decoder
}()
private lazy var successSerializer = DecodableResponseSerializer<T>(decoder: decoder)
private lazy var errorSerializer = DecodableResponseSerializer<APIError>(decoder: decoder)
public func serialize(request: URLRequest?, response: HTTPURLResponse?, data: Data?, error: Error?) throws -> Result<T, APIError> {
guard error == nil else { return .failure(APIError(message: "Unknown error", code: "unknown", args: [])) }
guard let response = response else { return .failure(APIError(message: "Empty response", code: "empty_response", args: [])) }
do {
if response.statusCode < 200 || response.statusCode >= 300 {
let result = try errorSerializer.serialize(request: request, response: response, data: data, error: nil)
return .failure(result)
} else {
let result = try successSerializer.serialize(request: request, response: response, data: data, error: nil)
return .success(result)
}
} catch(let err) {
return .failure(APIError(message: "Could not serialize body", code: "unserializable_body", args: [String(data: data!, encoding: .utf8)!, err.localizedDescription]))
}
}
}
extension DataRequest {
#discardableResult func responseTwoDecodable<T: Decodable>(queue: DispatchQueue = DispatchQueue.global(qos: .userInitiated), of t: T.Type, completionHandler: #escaping (Result<T, APIError>) -> Void) -> Self {
return response(queue: .main, responseSerializer: TwoDecodableResponseSerializer<T>()) { response in
switch response.result {
case .success(let result):
completionHandler(result)
case .failure(let error):
completionHandler(.failure(APIError(message: "Other error", code: "other", args: [error.localizedDescription])))
}
}
}
}
And with that, I can call my API like so:
AF.request(request).validate().responseTwoDecodable(of: [Item].self) { response in
switch response {
case .success(let items):
completion(.success(items))
case .failure(let error): //error is an APIError
log.error("Error while loading items: \(String(describing: error))")
completion(.failure(.couldNotLoad(underlyingError: error)))
}
}
I simply consider that any status code outside of the 200-299 range corresponds to an error.
ResponseSerializers have a single requirement. Largely you can just copy the existing serializers. For example, if you wanted to parse a CSV (with no response checking):
struct CommaDelimitedSerializer: ResponseSerializer {
func serialize(request: URLRequest?, response: HTTPURLResponse?, data: Data?, error: Error?) throws -> [String] {
// Call the existing StringResponseSerializer to get many behaviors automatically.
let string = try StringResponseSerializer().serialize(request: request,
response: response,
data: data,
error: error)
return Array(string.split(separator: ","))
}
}
You can read more in Alamofire's documentation.
Related
I'm using Alamofire5 and Swift 5.7.x.
There is a problem when I receive an response data from API server.
The response data default is:
{
statusCode: 200
message: "OK"
items: null
}
But the "items" attribute can be anything data type.
null, string, int, array, object..
I was about to solve using Generics.
But I don't know how to handle the null.
It's API common code:
struct Response<T: Codable>: Codable {
var statusCode: Int
var message: String
var items: T?
}
class API {
func request<Parameters: Encodable, T: Decodable>(_ path: String,
method: HTTPMethod = .get,
params: Parameters? = nil,
completion: #escaping (Result<Response<T?>, NetworkError>) -> Void) {
//some codes..
AF.request("\(path)",
method: method,
parameters: params,
encoder: JSONParameterEncoder.prettyPrinted
)
.validate(statusCode: 200..<400)
.validate(contentType: ["application/json"])
.responseData { response in
switch response.result {
case .success(let data):
guard let decodedData = try? JSONDecoder().decode(Response<T?>.self, from: data) else { return }
print(decodedData)
completion(.success(decodedData as Response<T?>))
case.failure(let error):
// some codes..
}
}
}
}
It's caller:
API.shared.request("/users/device", method: .post, params: reqParam) { (response: Result<Response<This is the problem..!!>, NetworkError>) in
switch response {
case .success(let data):
print("userDevice updated")
debugPrint(data)
case .failure(let error):
// some codes..
}
}
How can I pass nil on the caller?
Considering you have following response for all your API requests
{
statusCode: 200
message: "OK"
items: null
}
And your Codable struct as following
struct Response<T: Codable>: Codable {
var statusCode: Int
var message: String
var items: T? // Notice this is declared as Optional
}
As your items variable is declared as optional it will work as expected with little change.
You can declare an Empty Struct which can be used when you are expecting null value for items key, like following.
struct EmptyResponse: Codable {
}
You can use EmptyResponse struct as following:
API.shared.request("/users/device", method: .post, params: reqParam) { (response: Result<Response<EmptyResponse>, NetworkError>) in
switch response {
case .success(let data):
debugPrint(data)
if let items = data.items {
// Use items here
} else {
// Items is nil
}
case .failure(let error):
// some codes..
}
}
Given an API that for invalid requests, along with 400-range HTTP status code the server returns a JSON payload that includes a readable message. As an example, the server could return { "message": "Not Found" } with a 404 status code for deleted or non-existent content.
Without using publishers, the code would read,
struct APIErrorResponse: Decodable, Error {
let message: String
}
func request(request: URLRequest) async throws -> Post {
let (data, response) = try await URLSession.shared.data(for: request)
let statusCode = (response as! HTTPURLResponse).statusCode
if 400..<500 ~= statusCode {
throw try JSONDecoder().decode(APIErrorResponse.self, from: data)
}
return try JSONDecoder().decode(Post.self, from: data)
}
Can this be expressed succinctly using only functional code?
In other words, how can the following pattern be adapted to decode a different type based on the HTTPURLResponse.statusCode property, to return as an error, or more generally, how can the response property be handled separately from data attribute?
URLSession.shared.dataTaskPublisher(for: request)
.map(\.data)
.decode(type: Post.self, decoder: JSONDecoder())
.eraseToAnyPublisher()
you could try something like this approach:
func request(request: URLRequest) -> AnyPublisher<Post, any Error> {
URLSession.shared.dataTaskPublisher(for: request)
.tryMap { (output) -> Data in
let statusCode = (output.response as! HTTPURLResponse).statusCode
if 400..<500 ~= statusCode {
throw try JSONDecoder().decode(APIErrorResponse.self, from: output.data)
}
return output.data
}
.decode(type: Post.self, decoder: JSONDecoder())
.eraseToAnyPublisher()
}
I use a helper method for this:
extension Publisher where Output == (data: Data, response: HTTPURLResponse) {
func decode<Success, Failure>(
success: Success.Type,
failure: Failure.Type,
decoder: JSONDecoder
) -> AnyPublisher<Success, Error> where Success: Decodable, Failure: DecodableError {
tryMap { data, httpResponse -> Success in
guard httpResponse.statusCode < 500 else {
throw MyCustomError.serverUnavailable(status: httpResponse.statusCode)
}
guard httpResponse.statusCode < 400 else {
let error = try decoder.decode(failure, from: data)
throw error
}
let success = try decoder.decode(success, from: data)
return success
}
.eraseToAnyPublisher()
}
}
typealias DecodableError = Decodable & Error
which allows me to simplify the call sites like so:
URLSession.shared.dataTaskPublisher(for: request)
.decode(success: Post.self, failure: MyCustomError.self, decoder: JSONDecoder())
.eraseToAnyPublisher()
workingdogsupport has provided a good literal translation (+1). And LuLuGaGa has illustrated a nice compositional style (+1).
I might expand upon the latter, though, and recommend pattern matching on the various status codes, e.g. 2xx codes for decoding success, 4xx for graceful web service errors, and a more general .badServerResponse (and includes the diagnostic information so that the developer working on the call point has a chance to figure out what went wrong) for anything else.
E.g., you might have an general extension (which doesn’t use any types particular to the app):
extension Publisher where Output == (data: Data, response: URLResponse) {
func decode<Success: Decodable, Failure: Decodable & Error>(
success: Success.Type = Success.self,
failure: Failure.Type = Failure.self,
decoder: JSONDecoder = JSONDecoder()
) -> AnyPublisher<Success, Error> {
tryMap { data, response -> Success in
switch (response as! HTTPURLResponse).statusCode {
case 200..<300: return try decoder.decode(Success.self, from: data)
case 400..<500: throw try decoder.decode(Failure.self, from: data)
default: throw URLError(.badServerResponse, userInfo: ["data": data, "response": response])
}
}
.eraseToAnyPublisher()
}
}
Or, because I hate force-unwrapping:
extension Publisher where Output == (data: Data, response: URLResponse) {
func decode<Success: Decodable, Failure: Decodable & Error>(
success: Success.Type = Success.self,
failure: Failure.Type = Failure.self,
decoder: JSONDecoder = JSONDecoder()
) -> AnyPublisher<Success, Error> {
tryMap { data, response -> Success in
guard let response = response as? HTTPURLResponse else {
throw URLError(.badServerResponse, userInfo: ["data": data, "response": response])
}
switch response.statusCode {
case 200..<300: return try decoder.decode(Success.self, from: data)
case 400..<500: throw try decoder.decode(Failure.self, from: data)
default: throw URLError(.badServerResponse, userInfo: ["data": data, "response": response])
}
}
.eraseToAnyPublisher()
}
}
Regardless, I might then have an extension for this app that decodes your particular web service’s specific error struct:
extension Publisher where Output == (data: Data, response: URLResponse) {
func decode<Success: Decodable>(
success: Success.Type = Success.self,
decoder: JSONDecoder = JSONDecoder()
) -> AnyPublisher<Success, Error> {
decode(success: success, failure: APIErrorResponse.self, decoder: decoder)
}
}
Then the app code can avail itself of the above (and infer the success type):
func postsPublisher(for request: URLRequest) -> AnyPublisher<Post, Error> {
URLSession.shared.dataTaskPublisher(for: request)
.decode()
.eraseToAnyPublisher()
}
Anyway, that results in a succinct call-point, with a reusable extension.
I'm fairly new to Combine declarative API. I'm trying to implement a generic network layer for a SwiftUI application. For all requests that receive data I understand how to structure the data flow.
My problem is that I have some HTTP POST requests that returns no data. Only a HTTP 200 on success. I can't figure out how to create a publisher that will handle a decoding that can fail since there could be not data in the body of the response. Here's what I tried:
func postResource<Resource: Codable>(_ resource: Resource, to endpoint: Endpoint) -> AnyPublisher<Resource?, NetworkError> {
return Just(resource)
.subscribe(on: queue)
.encode(encoder: JSONEncoder())
.mapError { error -> NetworkError in
return NetworkError.encoding(error)
}
.map { data -> URLRequest in
return endpoint.makeRequest(with: data)
}
.tryMap { request -> Resource? in
self.session.dataTaskPublisher(for: request)
.tryMap { data, response -> Data in
guard let httpUrlResponse = response as? HTTPURLResponse else { throw NetworkError.unknown }
guard (200 ... 299).contains(httpUrlResponse.statusCode) else { throw NetworkError.error(for: httpUrlResponse.statusCode) }
return data
}
.tryMap { data -> Resource? in
return try? JSONDecoder().decode(Resource.self, from: data)
}
}
.mapError({ error -> NetworkError in
switch error {
case is Swift.DecodingError:
return NetworkError.decoding(error)
case let urlError as URLError:
return .urlError(urlError)
case let error as NetworkError:
return error
default:
return .unknown
}
})
.eraseToAnyPublisher()
}
The compiler is complaining with the following error on tryMap row:
Declared closure result 'Publishers.TryMap<URLSession.DataTaskPublisher, Resource?>' is incompatible with contextual type 'Resource?'
Anyone has an idea?
Thanks!
enum NetworkError: Error {
case encoding(Error)
case error(for: Int)
case decoding(Error)
case urlError(URLError)
case unknown
}
func postResource<Resource: Codable>(_ resource: Resource, to endpoint: Endpoint) -> AnyPublisher<Resource?, NetworkError> {
Just(resource)
.subscribe(on: queue)
.encode(encoder: JSONEncoder())
.mapError { error -> NetworkError in
NetworkError.encoding(error)
}
.map { data -> URLRequest in
endpoint.makeRequest(with: data)
}
.flatMap { request in // the key thing is here you should you use flatMap instead of map
URLSession.shared.dataTaskPublisher(for: request)
.tryMap { data, response -> Data in
guard let httpUrlResponse = response as? HTTPURLResponse else { throw NetworkError.unknown }
guard 200 ... 299 ~= httpUrlResponse.statusCode else { throw NetworkError.error(for: httpUrlResponse.statusCode) }
return data
}
.tryMap { data -> Resource? in
try? JSONDecoder().decode(Resource.self, from: data)
}
}
.mapError({ error -> NetworkError in
switch error {
case is Swift.DecodingError:
return NetworkError.decoding(error)
case let urlError as URLError:
return .urlError(urlError)
case let error as NetworkError:
return error
default:
return .unknown
}
})
.receive(on: DispatchQueue.main)
.eraseToAnyPublisher()
}
I have an app where I used RxSwift for my networking by extending ObservableType this works well but the issue I am having now is when I make an API request and there is an error, I am unable to show the particular error message sent from the server. Now how can I get the particular error response sent from the server
extension ObservableType {
func convert<T: EVObject>(to observableType: T.Type) -> Observable<T> where E: DataRequest {
return self.flatMap({(request) -> Observable<T> in
let disposable = Disposables.create {
request.cancel()
}
return Observable<T>.create({observer -> Disposable in
request.validate().responseObject { (response: DataResponse<T>) in
switch response.result {
case .success(let value):
if !disposable.isDisposed {
observer.onNext(value)
observer.onCompleted()
}
case .failure(let error):
if !disposable.isDisposed {
observer.onError(NetworkingError(httpResponse: response.response,
networkData: response.data, baseError: error))
observer.onCompleted()
}
}
}
return disposable
})
})
}
}
let networkRetryPredicate: RetryPredicate = { error in
if let err = error as? NetworkingError, let response = err.httpResponse {
let code = response.statusCode
if code >= 400 && code < 600 {
return false
}
}
return true
}
// Use this struct to pass the response and data along with
// the error as alamofire does not do this automatically
public struct NetworkingError: Error {
let httpResponse: HTTPURLResponse?
let networkData: Data?
let baseError: Error
}
response from the server could be
{
"status" : "error",
"message" : " INSUFFICIENT_FUNDS"
}
or
{
"status" : "success",
"data" : " gghfgdgchf"
}
my response is handled like this
class MaxResponse<T: NSObject>: MaxResponseBase, EVGenericsKVC {
var data: T?
public func setGenericValue(_ value: AnyObject!, forUndefinedKey key: String) {
switch key {
case "data":
data = value as? T
default:
print("---> setGenericValue '\(value)' forUndefinedKey '\(key)' should be handled.")
}
}
public func getGenericType() -> NSObject {
return T()
}
}
the error is
return ApiClient.session.rx.request(urlRequest: MaxApiRouter.topupWall(userId: getUser()!.id!, data: body))
.convert(to: MaxResponse<Wall>.self)
In the official Alamofire docs it is mentioned that validate(), without any parameters:
Automatically validates status code within 200..<300 range, and that
the Content-Type header of the response matches the Accept header of
the request, if one is provided.
So if you do not include Alamofire's validate() you are saying that no matter the status code, if the request did get through, you will consider it successful, so that's why it shows nothing in the failure block.
However if you prefer to use it, yes, it will give you an ResponseValidationFailureReason error, but you still have access to the response.data. Try printing it, you should see the expected error response from the server:
if let responseData = response.data {
print(String(data: responseData, encoding: .utf8))
}
For my networking module, I have this protocol that I adopt for accessing different parts of the API:
protocol Router: URLRequestConvertible {
var baseUrl: URL { get }
var route: Route { get }
var method: HTTPMethod { get }
var headers: [String: String]? { get }
var encoding: ParameterEncoding? { get }
var responseResultType: Decodable.Type? { get }
}
I'm adopting this with enums that look like this:
enum TestRouter: Router {
case getTestData(byId: Int)
case updateTestData(byId: Int)
var route: Route {
switch self {
case .getTestData(let id): return Route(path: "/testData/\(id)")
case .updateTestData(let id): return Route(path: "/testDataOtherPath/\(id)")
}
}
var method: HTTPMethod {
switch self {
case .getTestData: return .get
case .updateTestData: return .put
}
}
var headers: [String : String]? {
return [:]
}
var encoding: ParameterEncoding? {
return URLEncoding.default
}
var responseResultType: Decodable.Type? {
switch self {
case .getTestData: return TestData.self
case .updateTestData: return ValidationResponse.self
}
}
}
I want to use Codable for decoding nested Api responses. Every response consists of a token and a result which content is depending on the request route.
For making the request I want to use the type specified in the responseResultType property in the enum above.
struct ApiResponse<Result: Decodable>: Decodable {
let token: String
let result: Result
}
extension Router {
func asURLRequest() throws -> URLRequest {
// Construct URL
var completeUrl = baseUrl.appendingPathComponent(route.path, isDirectory: false)
completeUrl = URL(string: completeUrl.absoluteString.removingPercentEncoding ?? "")!
// Create URL Request...
var urlRequest = URLRequest(url: completeUrl)
// ... with Method
urlRequest.httpMethod = method.rawValue
// Add headers
headers?.forEach { urlRequest.addValue($0.value, forHTTPHeaderField: $0.key) }
// Encode URL Request with the parameters
if encoding != nil {
return try encoding!.encode(urlRequest, with: route.parameters)
} else {
return urlRequest
}
}
func requestAndDecode(completion: #escaping (Result?) -> Void) {
NetworkAdapter.sessionManager.request(urlRequest).validate().responseData { response in
let responseObject = try? JSONDecoder().decode(ApiResponse<self.responseResultType!>, from: response.data!)
completion(responseObject.result)
}
}
}
But in my requestAndDecode method It throws an compiler error (Cannot invoke 'decode' with an argument list of type '(Any.Type, from: Data)'). I can't use ApiResponse<self.responseResultType!> like that.
I could make this function generic and call it like this:
TestRouter.getTestData(byId: 123).requestAndDecode(TestData.self, completion:)
but then I'd have to pass the response type everytime I want to use this endpoint.
What I want to achieve is that the extension function requestAndDecode takes it response type information from itself, the responseResultType property.
Is this possible?
Ignoring the actual error report you have a fundamental problem with requestAndDecode: it is a generic function whose type parameters are determined at the call site which is declared to return a value of type Result yet it attempts to return a value of type self.responseResultType whose value is an unknown type.
If Swift's type system supported this it would require runtime type checking, potential failure, and your code would have to handle that. E.g. you could pass TestData to requestAndDecode while responseResultType might be ValidationResponse...
Change the JSON call to:
JSONDecoder().decode(ApiResponse<Result>.self ...
and the types statically match (even though the actual type that Result is is unknown).
You need to rethink your design. HTH
Create a Generic function with Combine and AlomFire. You can use it for all method(get, post, put, delete)
func fatchData<T: Codable>(requestType: String, url: String, params: [String : Any]?, myType: T.Type, completion: #escaping (Result<T, Error>) -> Void) {
var method = HTTPMethod.get
switch requestType {
case "Get":
method = HTTPMethod.get
case "Post":
method = HTTPMethod.post
print("requestType \(requestType) \(method) ")
case "Put":
method = HTTPMethod.put
default:
method = HTTPMethod.delete
}
print("url \(url) \(method) \(AppConstant.headers) ")
task = AF.request(url, method: method, parameters: params, encoding: JSONEncoding.default, headers: AppConstant.headers)
.publishDecodable(type: myType.self)
.sink(receiveCompletion: { (completion) in
switch completion{
case .finished:
()
case .failure(let error):
// completion(.failure(error))
print("error \(error)")
}
}, receiveValue: {
[weak self ](response) in
print("response \(response)")
switch response.result{
case .success(let model):
completion(.success(model))
print("error success")
case .failure(let error):
completion(.failure(error))
print("error failure \(error.localizedDescription)")
}
}
)
}