This is a project inspired by a blog post by Apple about JSON parsing. The idea is to compare popular JSON mappers and native parsing against each other using Xcode's performance test (as part of the XCTest suite). The exception to this is Big Nerd Ranch's Freddy (in which they wrote their own parser).

see NormalPerfTest.swift

func testNormalPerformance() {
    self.measure {
        do {
            guard let json = try JSONSerialization.jsonObject(with: self.data!) as? MyJSON,
                let items = json["items"] as? [MyJSON] else {
                    XCTFail("bad JSON")
                    return
            }

            _ = try items.map{
                try Repo(myJson: $0)
            }
        } catch {
            XCTFail("bad json")
        }
    }
}

The idea is to parse data into a JSON object and then pass that object to the initializer of each mapper. Each JSON mapping package/technique has its own test class (because of some naming overlaps between imports). The data used to generate the JSON is a computed variable in an extension on XCTestCase. Doing this allows the tests to isolate the steps of creating an object after a network call has been made.

Using the measure block allows me to run the test ten times and create a baseline with a standard deviation. These numbers then allow us, as developers, to compare the packages and choose which makes the most sense for our projects.

I encourage you to pull this project and check out the results (I was a bit surprised)! In order to do so, be sure you have Cocoapods installed -- as that's how I imported the parsing packages. Simply pull the code and:

$ cd ~/path/to/source
$ pod install

• SwiftyJSON
• Freddy
• Gloss
• Genome
• JSONCodable
• JSONSerialization

If I've missed a package you'd like to be tested either create an issue, pull request, or send me a tweet @mdiasdev.