Computed Properties and Watchers

Computed Properties

In-template expressions are very convenient, but they are meant for simple operations. Putting too much logic in your templates can make them bloated and hard to maintain. For example:

<div id="example">
  {{ message.split('').reverse().join('') }}
</div>

At this point, the template is no longer simple and declarative. You have to look at it for a second before realizing that it displays message in reverse. The problem is made worse when you want to include the reversed message in your template more than once.

That’s why for any complex logic, you should use a computed property.

Basic Example

<div id="example">
  <p>Original message: "{{ message }}"</p>
  <p>Computed reversed message: "{{ reversedMessage }}"</p>
</div>

var vm = new Kdu({
  el: '#example',
  data: {
    message: 'Hello'
  },
  computed: {
    // a computed getter
    reversedMessage: function () {
      // `this` points to the vm instance
      return this.message.split('').reverse().join('')
    }
  }
})

Result:

Here we have declared a computed property reversedMessage. The function we provided will be used as the getter function for the property vm.reversedMessage:

console.log(vm.reversedMessage) // => 'olleH'
vm.message = 'Goodbye'
console.log(vm.reversedMessage) // => 'eybdooG'

You can open the console and play with the example vm yourself. The value of vm.reversedMessage is always dependent on the value of vm.message.

You can data-bind to computed properties in templates just like a normal property. Kdu is aware that vm.reversedMessage depends on vm.message, so it will update any bindings that depend on vm.reversedMessage when vm.message changes. And the best part is that we’ve created this dependency relationship declaratively: the computed getter function has no side effects, which makes it easier to test and understand.

Computed Caching vs Methods

You may have noticed we can achieve the same result by invoking a method in the expression:

<p>Reversed message: "{{ reverseMessage() }}"</p>

// in component
methods: {
  reverseMessage: function () {
    return this.message.split('').reverse().join('')
  }
}

Instead of a computed property, we can define the same function as a method. For the end result, the two approaches are indeed exactly the same. However, the difference is that computed properties are cached based on their reactive dependencies. A computed property will only re-evaluate when some of its reactive dependencies have changed. This means as long as message has not changed, multiple access to the reversedMessage computed property will immediately return the previously computed result without having to run the function again.

This also means the following computed property will never update, because Date.now() is not a reactive dependency:

computed: {
  now: function () {
    return Date.now()
  }
}

In comparison, a method invocation will always run the function whenever a re-render happens.

Why do we need caching? Imagine we have an expensive computed property A, which requires looping through a huge Array and doing a lot of computations. Then we may have other computed properties that in turn depend on A. Without caching, we would be executing A’s getter many more times than necessary! In cases where you do not want caching, use a method instead.

Computed vs Watched Property

Kdu does provide a more generic way to observe and react to data changes on a Kdu instance: watch properties. When you have some data that needs to change based on some other data, it is tempting to overuse watch - especially if you are coming from an AngularJS background. However, it is often a better idea to use a computed property rather than an imperative watch callback. Consider this example:

<div id="demo">{{ fullName }}</div>

var vm = new Kdu({
  el: '#demo',
  data: {
    firstName: 'Foo',
    lastName: 'Bar',
    fullName: 'Foo Bar'
  },
  watch: {
    firstName: function (val) {
      this.fullName = val + ' ' + this.lastName
    },
    lastName: function (val) {
      this.fullName = this.firstName + ' ' + val
    }
  }
})

The above code is imperative and repetitive. Compare it with a computed property version:

var vm = new Kdu({
  el: '#demo',
  data: {
    firstName: 'Foo',
    lastName: 'Bar'
  },
  computed: {
    fullName: function () {
      return this.firstName + ' ' + this.lastName
    }
  }
})

Much better, isn’t it?

Computed Setter

Computed properties are by default getter-only, but you can also provide a setter when you need it:

// ...
computed: {
  fullName: {
    // getter
    get: function () {
      return this.firstName + ' ' + this.lastName
    },
    // setter
    set: function (newValue) {
      var names = newValue.split(' ')
      this.firstName = names[0]
      this.lastName = names[names.length - 1]
    }
  }
}
// ...

Now when you run vm.fullName = 'John Doe', the setter will be invoked and vm.firstName and vm.lastName will be updated accordingly.

Watchers

While computed properties are more appropriate in most cases, there are times when a custom watcher is necessary. That’s why Kdu provides a more generic way to react to data changes through the watch option. This is most useful when you want to perform asynchronous or expensive operations in response to changing data.

For example:

<div id="watch-example">
  <p>
    Ask a yes/no question:
    <input k-model="question">
  </p>
  <p>{{ answer }}</p>
</div>

<!-- Since there is already a rich ecosystem of ajax libraries    -->
<!-- and collections of general-purpose utility methods, Kdu core -->
<!-- is able to remain small by not reinventing them. This also   -->
<!-- gives you the freedom to use what you're familiar with.      -->
<script src="https://cdn.jsdelivr.net/npm/axios@0.12.0/dist/axios.min.js"></script>
<script src="https://cdn.jsdelivr.net/npm/lodash@4.13.1/lodash.min.js"></script>
<script>
var watchExampleVM = new Kdu({
  el: '#watch-example',
  data: {
    question: '',
    answer: 'I cannot give you an answer until you ask a question!'
  },
  watch: {
    // whenever question changes, this function will run
    question: function (newQuestion, oldQuestion) {
      this.answer = 'Waiting for you to stop typing...'
      this.debouncedGetAnswer()
    }
  },
  created: function () {
    // _.debounce is a function provided by lodash to limit how
    // often a particularly expensive operation can be run.
    // In this case, we want to limit how often we access
    // yesno.wtf/api, waiting until the user has completely
    // finished typing before making the ajax request. To learn
    // more about the _.debounce function (and its cousin
    // _.throttle), visit: https://lodash.com/docs#debounce
    this.debouncedGetAnswer = _.debounce(this.getAnswer, 500)
  },
  methods: {
    getAnswer: function () {
      if (this.question.indexOf('?') === -1) {
        this.answer = 'Questions usually contain a question mark. ;-)'
        return
      }
      this.answer = 'Thinking...'
      var vm = this
      axios.get('https://yesno.wtf/api')
        .then(function (response) {
          vm.answer = _.capitalize(response.data.answer)
        })
        .catch(function (error) {
          vm.answer = 'Error! Could not reach the API. ' + error
        })
    }
  }
})
</script>

Result:

In this case, using the watch option allows us to perform an asynchronous operation (accessing an API), limit how often we perform that operation, and set intermediary states until we get a final answer. None of that would be possible with a computed property.

In addition to the watch option, you can also use the imperative vm.$watch API.