Adding A Copy List Command To List-o-Mat

Custom intents are meant to be used only where there is no pre-defined intent, and Siri does actually offer a lot of list and task support in its Lists and Notes Siri Domain.

But, one way to use a list is as a template for a repeated routine or process. To do that we’ll want to copy an existing list and uncheck all of its items. The built-in List intents don’t support this action.

First, we need to add a way to do this manually. Here is a demo of this new behavior in List-o-Mat:

To get this behavior to be invokable by Siri, we’ll “donate an intent,” which means we’ll tell iOS every time you do this. Then, it will eventually learn that in the morning, you like to copy this list and offer it as a shortcut. Users can also look for donated intents and assign phrases manually.

Creating The Custom Intent

The next step is to create the custom intent in Xcode. There is a new file template, so:

1. Choose File → New File and pick “SiriKit Intent Definition File”.
   

   

2. Name the file ListOMatCustomIntents.intentdefinition, and choose to put the file in both the App and Intent Extension targets. This will automatically generate classes into both targets that implement the intent protocols but have your custom behavior implemented.
3. Open the Definition file.
4. Use the + button on the bottom left to add an intent and name it “CopyList”.
5. Set the Category to “Create” and fill in the title and subtitle to describe the intent:
   

   

6. Add a String parameter named “list”.
   

   

7. Add a shortcut type with the list parameter and give it a title named “Copy list”.
   

   

If you look in the Intent plist, you will see that this intent has already been configured for you:

Donating The Intent

When we do a user interaction in our app that we want Siri to know about, we donate it to Siri. Siri keeps track of contextual information, like the time, day of the week, and even location, and if it notices a pattern, it will offer the shortcut to the user.

When we tap the Copy menu, add this code:

@available(iOS 12, *)
func donateCopyListInteraction(listName: String) {
    let copyListInteraction = CopyListIntent()
    copyListInteraction.list = listName
    copyListInteraction.suggestedInvocationPhrase = "Copy (listName)"
    let interaction = INInteraction(intent: copyListInteraction, response: nil)
    interaction.donate { [weak self] (error) in
        self?.show(error: error)
    }
}

This simply creates an object of the auto-generated CopyListIntent class and donates it to Siri. Normally, iOS would collect this info and wait for the appropriate time to show it, but for development, you can open the Settings app, go to the Developer section, and turn on Siri Shortcut debugging settings.

Note: As of this writing, with the first betas, this debug setting only works on devices, and not the simulator. Since the setting is there, I expect it to start working in further betas.

When you do this, your donated shortcut shows up on in Siri Suggestions in Spotlight.

Tapping that will call into your Intent extension because we are allowing background execution. We’ll add support for that next.

Handling The Custom Intent

We already have an Intents extension, and since the custom intent definitions file is already added to the file, it also has the generated intent classes. All we need to do is add a handler.

The first step is to add a new class, named CopyListIntentHandler to the extension. Here is its code:

import Intents

@available(iOS 12, *)
class CopyListIntentHandler: ListOMatIntentsHandler, CopyListIntentHandling {

    func handle(intent: CopyListIntent, completion: @escaping (CopyListIntentResponse) -> Void) {

        // Find the list
        var lists = loadLists()
        guard
            let listName = intent.list?.lowercased(),
            let listIndex = lists.index(where: { $0.name.lowercased() == listName})
        else {
            completion(CopyListIntentResponse(code: .failure, userActivity: nil))
            return
        }

        // Copy the list to the top, and respond with success
        copyList(from: &lists, atIndex: listIndex, toIndex: 0)
        save(lists: lists)
        let response = CopyListIntentResponse(code: .success, userActivity: nil)
        completion(response)
    }

}

Custom intents only have a confirm and handle phase (custom resolution of parameters is not supported). Since the default confirm() returns success, we’ll just implement handle(), which has to look up the list, copy it, and let Siri know if it was successful or not.

You also need to dispatch to this class from the registered intent handler by adding this code:

if #available(iOS 12, *) {
    if intent is CopyListIntent {
        return CopyListIntentHandler()
    }
}

Now you can actually tap that Siri suggestion and it will bring this up:

And tapping the Create button will copy the list. The button says “Create” because of the category we chose in the intent definition file.

Phew, that was a lot. These new Siri shortcuts are the main feature in iOS 12 that has a new large developer surface area to explore. Also, since I happened to have a good (and documented) Siri example to work with, it was reasonable to try to add the new features to it this week.

You can see the update List-o-Mat in GitHub. Until Xcode 10 and iOS 12 are released it’s in its own branch.

The next few days, I’ll mostly be looking at Apple sample code or making much smaller projects.

End Of Day 3: Xcode Playgrounds

The entire previous day was spent in Xcode 10 beta, which didn’t crash once and seemed ready for development. So now I wanted to explore the new Playgrounds features.

The main thing I wanted from playgrounds is to make them more stable and much faster. To make them faster, Apple added a big feature — a REPL mode.

Before Xcode 10, when you were in a Playground that had auto-run on (which is the default), every line of code actually rebuilt the entire file and ran it from the beginning. If you had built up any state, it was lost. But, the real issue was that this was way too slow for iterative development. When I use Playgrounds, I set them to manually run, but even that is slow.

In Xcode 10, manual running is more the norm, but after you run it, you can add more lines at the bottom of the page and continue execution. This means you can explore data and draw views iteratively without constantly rebuilding and starting from scratch.

To get started, I created an iOS playground (File → New → Playground) with the Single View template.

Turn on manual running by bringing down the menu below the Play button (the triangle in the bottom left corner). This puts a vertical strip to the left that shows the current position of the Play head (kind of like breakpoints).

You can tap any line and then tap the play button to its left. This will run the Playground to this point. Then you can go further by tapping lines lower in the Playground. Critically, you can add more lines to the bottom and type Shift + Enter after each one to move the Play head to that point.

Here’s a GIF of me changing the label of a view without needing to restart the Playground. After each line I type, I am pressing Shift + Enter.

Playgrounds also support custom rendering of your types now, and Apple is making a big push for every Swift framework to include a Playground to document it.

End Of Day 4: Create ML

Last year, Apple made a big leap for programming Machine Learning for their devices. There was a new ML model file format and direct support for it in Xcode.

The potential was that there would be a large library of these model files, that there would be tools that would create them, and that many more app developers would be able to incorporate ML into their projects without having to know how to create models.

This hasn’t fully materialized. Apple didn’t add to the repository of models after WWDC, and although there are third-party repositories, they mostly have models that are variations on the image classification demos. ML is used for a lot more than image classification, but a broad selection of examples did not appear.

So, it became clear that any real app would need its developers to train new models. Apple released Turi Create for this purpose, but its far from simple.

At WWDC 2018, Apple did a few things to Core ML:

1. They expanded the Natural Language Processing (NLP) part of Core ML which gives us a new major domain of examples.
2. They added the concept of Transfer Learning to Core ML, which allows you to add training data to an existing model. This means you can take models from the library, and customize them to your own data (for example, have them recognize new objects in images you provide).
3. They released Create ML which is implemented inside of Xcode Playgrounds and lets you drag and drop data for training and generate model extensions (using Transfer Learning).

This is another nice step in democratizing ML. There’s not much code to write here. To extend an image classifier, you just need to gather and label images. Once you have them, you just drag them into Create ML. You can see the demo in this Create ML WWDC video.

End Of The Week: Play With The New AR Demos

ARKit was another big addition last year and it seems even more clear that an AR device is coming.

My ARKit code from last year’s article is still a good way to get started. Most of the new features are about making AR more accurate and faster.

After that, if you have installed a beta, you will definitely want to download the new SwiftShot ARKit demo app. This app takes advantage of the new features of ARKit, especially the multi-player experience. Two or more devices on the same network and in the same place, can communicate with each other and see the same AR experience.

Of course, to play this, you need two or more devices you are willing to put on the iOS 12 beta. I’m waiting for the public beta to do this because I only have one beta-safe device.

The easier AR app to play with is the new Measure app, which allows you to measure the length of real objects you see in AR camera view. There have been third-party apps that do this, but Apple’s is polished and pre-installed with iOS 12.

Links To WWDC Videos And Sample Code

So, I’m looking forward to doing more with Xcode 10 and iOS 12 this summer while we wait for the new phones and whatever devices Apple might release at the end of the summer. In the meantime, iOS developers can enjoy the sun, track our hikes with our new beta Watch OS, and watch these WWDC videos when we get a chance.

You can stream WWDC 2018 videos from the Apple developer portal. There is also this unofficial Mac App for viewing WWDC videos.

Here are the videos referenced in this article:

• WWDC 2018 Keynote
• WWDC 2018 Platforms State of the Union
• Introduction to Siri Shortcuts
• Getting the Most out of Playgrounds in Xcode
• Introducing Create ML, and if you want something more advanced, A Guide to Turi Create

To start playing with Xcode 10 and iOS 12:

• Download the betas (visit on a device to get the beta profile)
• List-o-Mat with Siri Shortcut updates
• Swift Shot (the multi-player ARKit 2 game)

(ra, il)