Please refactor your unwrapping optionals

[AF-cgi]

I use MapboxDirections 0.15.1. After response my crashed with an unwrapping optional issue in MBRouteStep line 600.
let drivingSide = DrivingSide(description: json["driving_side"] as! String) ?? .right
I resolve the issue by:
let drivingSide = DrivingSide(description: json["driving_side"] as? String ?? "") ?? .right
I can imaging that they are many other optional issues in the project.

[bsudekum]

Thanks for the bug report @AFcgi, this sounds reasonable to me.

[bsudekum]

Actually, after thinking about this more, driving_side is always present on a RouteStep when making a request to Mapbox Directions API. @AFcgi are you hitting our API or have you stood up your own server?

[AF-cgi]

Hi @bsudekum,
thanks for your reply. We use an own server. The driving_side isn't present on our RouteStep. Also we use SwiftLint. One of these rules says do not unwrapped optionals in your code. So I think it's a better way to unwrap all your optionals.

[1ec5]

Related: mapbox/MapboxGeocoder.swift#106.

We should replace most of the force-unwrapping and assertions with exceptions, reserving assertions for things that are entirely within the control of the client-side library (and not due to a malformed response or an error in the application code).

[bsudekum]

@1ec5 asserting would not really help @AFcgi out here, correct?

[1ec5]

To clarify:

• When dealing with input from the developer or from the Directions API, throw exceptions.
• Otherwise, when dealing with data internal to the library, assert or force-unwrap (preferably the former).

In general, this means that initializers would be failable and non-initializers would have more assertions.

[1ec5]

In general, this means that initializers would be failable and non-initializers would have more assertions.

Migrating to JSONDecoder entails making the initializers failable. #221 still force-unwraps some things, but it should be much easier to throw instead now.

/cc @frederoni

[1ec5]

#382 adopts Codable and eliminates nearly all assertions and upfront optional unwrapping.

[1ec5]

#382 also added a preconditions to the coding methods that are enforced by force-unwrapping optionals. #388 makes the preconditions gentler by throwing errors instead, but that change should probably go in a separate PR.