Custom error is returned from custom validator and/or decode() function

[haemi]

I have a custom validator like this:

public class ErrorMessageValidator: HTTPValidator {

    public func validate(response: HTTPResponse, forRequest request: HTTPRequest) -> HTTPResponseValidatorResult {
        if [HTTPStatusCode.ResponseType.success, .redirection].contains(response.statusCode.responseType) {
            return .nextValidator
        }

        guard let data = response.data else {
            return .failChain(HTTPError(.invalidResponse))
        }

        let jsonData = JSON(data)

        guard let message = jsonData.dictionary?["message"]?.string else {
            return .nextValidator
        }

        return .failChain(MyError(title: message))
    }
}

And I get to the last line as expected, so a MyError is returned, wrapped into the failChain.

But in my calling code:

let request = myServerCall(myParameters).httpRequest()
let response = try await request.fetch(MyModel.self)

an error is thrown (expected) but it is an HTTPError with an empty error property, so my MyError isn't available here

[malcommac]

May you post the structure of MyError so I can replicate the error? Thanks!

[haemi]

it's very simple struct:

public struct MyError: Error {
    public let title: String

    public init(title: String) {
        self.title = title
    }
}

[malcommac]

Okay I've fixed the dispatch of custom errors. Moreover I decided to dispatch original response of the request and attach additional error information based upon the library activity (max attempts, validation errors etc.).
May you take a look to see if it works for you?
I've also added two dedicated unit tests.

[haemi]

it works when I use

let response = try await request.fetch()
// now, response.error?.error is MyError

but when I use

let myModel = try await request.fetch(MyModel.self)

the error isn't thrown if the body can be encoded. I would have expected that somehow I get to the error then as well, but there's no way as myModel is of type MyModel. Currently it's only possible to do it in multiple steps:

let response = try await request.fetch()
// check if response.error?.error is set, if not:
let myModel = try response.decode(MyModel.self)

[malcommac]

You are absolutely right, the fetch error should be passed outside the decode() function.
However the error must be an HTTPError so you would read it as:

do {
  _ = try await req.fetch(client: newClient, DummyDecodeStruct.self)
} catch {
  let customUnderlyingError = (error as? HTTPError)?.error as? MyError
  print(customUnderlyingError)
}

[malcommac]

ok I've fixed the code and added the necessary unit test