Argo

The Greek word for swift and the ship used by Jason, son of Aeson, of the Argonauts. Aeson is the JSON parsing library in Haskell that inspired Argo, much like Aeson inspired his son Jason.

NOTE: For Swift 1.1 support, use the versions tagged 0.3.x, which you can read about in the releases.

Installation

Git Submodules

I guess you could do it this way if that’s your thing.

Add this repo as a submodule, and add the project file to your workspace. You can then link against Argo.framework for your application target.

You’ll also need to add Runes and Box to your project the same way.

Usage tl;dr:

import Argo
import Runes

struct User {
  let id: Int
  let name: String
  let email: String?
  let role: Role
  let companyName: String
  let friends: [User]
}

extension User: Decodable {
  static func create(id: Int)(name: String)(email: String?)(role: Role)(companyName: String)(friends: [User]) -> User {
    return User(id: id, name: name, email: email, role: role, companyName: companyName, friends: friends)
  }

  static func decode(j: JSON) -> Decoded<User> {
    return User.create
      <^> j <| "id"
      <*> j <| "name"
      <*> j <|? "email" // Use ? for parsing optional values
      <*> j <| "role" // Custom types that also conform to Decodable just work
      <*> j <| ["company", "name"] // Parse nested objects
      <*> j <|| "friends" // parse arrays of objects
  }
}

// Wherever you receive JSON data:

let json: AnyObject? = NSJSONSerialization.JSONObjectWithData(data, options: NSJSONReadingOptions(0), error: nil)

if let j: AnyObject = json {
  let user: User? = decode(j)
}

Ideology

Argo’s core concept is that in order to maintain type safety, you should only be able to successfully decode an object if all parameters are satisfied properly. So if you have a model that looks like this:

struct User {
  let id: Int
  let name: String
}

but the JSON you receive from the server looks like this:

{
  "user": {
    "id": "this isn't a number",
    "name": "Gob Bluth"
  }
}

then ideally, JSON parsing would fail, and you’d get an error state instead of a User object. In Argo, if JSON parsing succeeds you’ll recieve the User object and you can be sure that it is full and valid. If it fails, you will instead be given the reason why the User couldn’t be constructed.

If you’re interested in learning more about the concepts and ideology that went into building Argo, we recommend reading the series of articles that were written alongside its development:

Functional Concepts

Argo really wants to be used with patterns borrowed from functional programming such as map and apply. We feel that these patterns greatly reduce the pain felt in trying to use JSON (an inherently loosely typed format) with Swift (a strictly typed language). It also gives us a way to succinctly maintain the core concept described above, and short circuit the decoding process if any part of it fails.

Additionally, we feel that the use of operators for these functions greatly improves the readability of the code we’re suggesting. Using named functions would lead to nested functions and a confusing number of parenthesis.

If you aren’t familiar with how these functions work (or just aren’t comfortable with using operators), that’s totally OK. It’s possible to use the library without using them, although it might be a little more painful.

If you’re looking to learn more about these functions, we would recommend reading the following articles:

And check out this talk:

Usage

The first thing you need to do when you receive JSON data is convert it from NSData to an AnyObject using the built-in NSJSONSerialization API. Once you have the AnyObject, you can call the global decode function to get back the decoded model.

var error: NSError?
let json: AnyObject? = NSJSONSerialization.JSONObjectWithData(responseData, options: NSJSONReadingOptions(0), error: &error)

if let j: AnyObject = json {
  let user: User? = decode(j) // ignore error info or
  let decodedUser: Decoded<User> = decode(j) // preserve error info
} else {
  // handle error
}

Note that you probably want to use an error pointer to track errors from NSJSONSerialization.

The JSON enum exists to help with some of the type inference, and also wraps up some of the casting that you’ll need to do to transform the JSON into native types.

Argo 1.0 introduces a new type: Decoded<T>. This is now the type returned from the decode function that you implement as part of the Decodable protocol. This new type allows you to preserve information about why a decoding failed. You can choose to either ignore the Decoded type and just get back the optional value or keep the Decoded type and use it to debug decoding errors. When you decode an AnyObject into a model using the global decode function, you can specify whether you want an Optional model or a Decoded model by specifying the return type as seen in the code block above.

Next, you need to make sure that models that you wish to decode from JSON conform to the Decodable protocol:

public protocol Decodable {
  typealias DecodedType = Self
  class func decode(JSON) -> Decoded<DecodedType>
}

You will need to implement the decode function to perform any kinds of transformations you need to transform your model from a JSON value. A simple implementation for an enum value might look like:

enum RoleType: String {
  case Admin = "Admin"
  case User = "User"
}

extension RoleType: Decodable {
  static func decode(j: JSON) -> Decoded<RoleType> {
    switch j {
    case let .String(s): return .fromOptional(RoleType(rawValue: s))
    default: return .TypeMismatch("\(j) is not a String") // Provide an Error message for a string type mismatch
    }
  }
}

The real power of Argo can be seen when decoding actual model objects. To illustrate this, we will decode the simple User object that we used earlier.

Create your model normally:

struct User {
  let id: Int
  let name: String
}

You will also want to create a curried creation function. This is needed because of a deficiency in Swift that doesn’t allow us to pass init functions around like other functions.

extension User {
  static func create(id: Int)(name: String) -> User {
    return User(id: id, name: name)
  }
}

Using this curried syntax will allow us to partially apply the function over the course of the decoding process. If you’d like to learn more about currying, we recommend the following articles:

The last thing to do will be to conform to Decodable and implement the required decode function. We will implement this function by using map (<^>) and apply (<*>) to conditionally pass the required parameters to the curried creation function. The common pattern will look like:

return Model.create <^> paramOne <*> paramTwo <*> paramThree

and so on. If any of those parameters are an error, the entire creation process will fail, and the function will return the first error. If all of the parameters are successful, the value will be unwrapped and passed to the create function.

In order to help with the decoding process, Argo introduces two new operators for parsing a value out of the JSON:

  • <| will attempt to parse a single value from the JSON
  • <|| will attempt to parse an array of values from the JSON

The usage of these operators is the same regardless:

  • json <| "key" is analogous to json["key"]
  • json <| ["key", "nested"] is analogous to json["key"]["nested"]

Both operators will attempt to parse the value from the JSON and will also attempt to cast the value to the expected type. If it can’t find a value, the function will return a Decoded.MissingKey(message: String) error. If the value it finds is the wrong type, the function will return a Decoded.TypeMismatch(message: String) error.

There are also Optional versions of these operators:

  • <|? will attempt to parse an optional value from the JSON
  • <||? will attempt to parse an optional array of values from the JSON

Usage is the same as the non-optionals. The difference is that if these fail parsing, the parsing continues. This is useful for including parameters that truly are optional values. For example, if your system doesn’t require someone to supply an email address, you could have an optional property: let email: String? and parse it with json <|? "email".

So to implement our decode function, we can use the JSON parsing operator in conjunction with map and apply:

extension User: Decodable {
  static func decode(j: JSON) -> Decoded<User> {
    return User.create
      <^> j <| "id"
      <*> j <| "name"
  }
}

For comparison, this same function without Argo would look like so:

extension User {
  static func decode(j: NSDictionary) -> User? {
    if let id = j["id"] as Int,
       let name = j["name"] as String
    {
      return User(id: id, name: name)
    }

    return .None
  }
}

You could see how this would get much worse with a more complex model.

You can decode custom types the same way, as long as the type also conforms to Decodable.

For more examples on how to use Argo, please check out the tests.

Contributing

See the CONTRIBUTING document. Thank you, contributors!

Need Help?

We offer 1-on-1 coaching. We can help you with functional programming in Swift, get started writing unit tests, and convert from Objective-C to Swift. Get in touch.

License

Argo is Copyright © 2015 thoughtbot, inc. It is free software, and may be redistributed under the terms specified in the LICENSE file.

About

thoughtbot

Argo is maintained and funded by thoughtbot, inc. The names and logos for thoughtbot are trademarks of thoughtbot, inc.

We love open source software! See our other projects or look at our product case studies and hire us to help build your iOS app.