Contribute to this guideReport an issue

guideCore editor architecture

The @ckeditor/ckeditor5-core package is relatively simple. It comes with just a handful of classes. The ones you need to know are presented below.

# Editor classes

The Editor class represents the base of the editor. It is an entry point of the application, gluing all other components. It provides a couple of properties that you need to know:

  • config – The configuration object.
  • plugins and commands – The collection of loaded plugins and commands.
  • model – The entry point to the editor’s data model.
  • data – The data controller. It controls how data is retrieved from the document and set inside it.
  • editing – The editing controller. It controls how the model is rendered to the user for editing.
  • keystrokes – The keystroke handler. It allows to bind keystrokes to actions.

Besides that, the editor exposes a few of methods:

  • create() – The static create() method. Editor constructors are protected and you should create editors using this static method. It allows the initialization process to be asynchronous.
  • destroy() – Destroys the editor.
  • execute() – Executes the given command.
  • setData() and getData() – A way to retrieve data from the editor and set data in the editor. The data format is controlled by the data controller’s data processor and it does not need to be a string (it can be e.g. JSON if you implement such a data processor). See, for example, how to produce Markdown output.

For the full list of methods check the API docs of the specific editor class you use. Specific editor implementations may provide additional methods.

The Editor class is a base to implement your own editors. CKEditor 5 Framework comes with a few editor types (for example, classic, inline and balloon) but you can freely implement editors which work and look completely different. The only requirement is that you implement the Editor interface.

# Plugins

Plugins are a way to introduce editor features. In CKEditor 5 even typing is a plugin. What is more, the Typing plugin requires Input and Delete plugins which are responsible for handling, methods of inserting text and deleting content, respectively. At the same time, a couple of other plugins need to customize Backspace behavior in certain cases, which is handled by themselves. This leaves the base plugins free of any non-generic knowledge.

Another important aspect of how existing CKEditor 5 plugins are implemented is the split into engine and UI parts. For example, the BoldEditing plugin introduces schema definition, mechanisms rendering <strong> tags, commands to apply and remove bold from text, while the Bold plugin adds the UI of the feature (i.e. a button). This feature split is meant to allow for greater reuse (one can take the engine part and implement their own UI for a feature) as well as for running CKEditor 5 on the server side. At the same time, the feature split is not perfect yet and will be improved.

The tl;dr of this is that:

  • Every feature is implemented or at least enabled by a plugin.
  • Plugins are highly granular.
  • Plugins know everything about the editor.
  • Plugins should know as little about other plugins as possible.

These are the rules based on which the official plugins were implemented. When implementing your own plugins, if you do not plan to publish them, you can reduce this list to the first point.

After this lengthy introduction (which is aimed at making it easier for you to digest the existing plugins), the plugin API can be explained.

All plugins need to implement the PluginInterface. The easiest way to do so is by inheriting from the Plugin class. The plugin initialization code should be located in the init() method (which can return a promise). If some piece of code needs to be executed after other plugins are initialized, you can put it in the afterInit() method. The dependencies between plugins are implemented using the static requires property.

import MyDependency from 'some/other/plugin';

class MyPlugin extends Plugin {
    static get requires() {
        return [ MyDependency ];
    }

    init() {
        // Initialize your plugin here.

        this.editor; // The editor instance which loaded this plugin.
    }
}

You can see how to implement a simple plugin in the Quick start guide.

# Commands

A command is a combination of an action (a callback) and a state (a set of properties). For instance, the bold command applies or removes bold attribute from the selected text. If the text in which the selection is placed has bold applied already, the value of the command is true, false otherwise. If the bold command can be executed on the current selection, it is enabled. If not (because, for example, bold is not allowed in this place), it is disabled.

All commands need to inherit from the Command class. Commands need to be added to the editor’s command collection so they can be executed by using the Editor#execute() method.

Take this example:

class MyCommand extends Command {
    execute( message ) {
        console.log( message );
    }
}

class MyPlugin extends Plugin {
    init() {
        const editor = this.editor;

        editor.commands.add( 'myCommand', new MyCommand( editor ) );
    }
}

Calling editor.execute( 'myCommand', 'Foo!' ) will log Foo! to the console.

To see how state management of a typical command like bold is implemented, have a look at some pieces of the AttributeCommand class on which bold is based.

First thing to notice is the refresh() method:

refresh() {
    const doc = this.editor.document;

    this.value = doc.selection.hasAttribute( this.attributeKey );
    this.isEnabled = doc.schema.checkAttributeInSelection(
        doc.selection, this.attributeKey
    );
}

This method is automatically called (by the command itself) when any changes are applied to the model. This means that the command automatically refreshes its own state when anything changes in the editor.

The important thing about commands is that every change in their state as well as calling the execute() method fires an event (e.g. change:value or execute).

These events make it possible to control the command from the outside. For instance, if you want to disable specific commands when some condition is true (for example, according to your application’s logic, they should be temporarily disabled) and there is no other, cleaner way to do that, you can block the command manually:

const command = editor.commands.get( 'someCommand' );

command.on( 'change:isEnabled', forceDisable, { priority: 'lowest' } );
command.isEnabled = false;

function forceDisabled() {
    this.isEnabled = false;
}

The command will now be disabled as long as you will not off this listener, regardless of how many times someCommand.refresh() is called.

# Event system and observables

CKEditor 5 has an event-based architecture so you can find EmitterMixin and ObservableMixin mixed all over the place. Both mechanisms allow for decoupling the code and make it extensible.

Most of the classes which were already mentioned are either emitters or observables (observable is an emitter too). Emitter can emit (fire events) as well as listen to them.

class MyPlugin extends Plugin {
    init() {
        // Make MyPlugin listen to someCommand#execute.
        this.listenTo( someCommand, 'execute', () => {
            console.log( 'someCommand was executed' );
        } );

        // Make MyPlugin listen to someOtherCommand#execute and block it.
        // You listen with high priority to block the event before
        // someOtherCommand's execute() method is called.
        this.listenTo( someOtherCommand, 'execute', ( evt ) => {
            evt.stop();
        }, { priority: 'high' } );
    }

    // Inherited from Plugin:
    destroy() {
        // Removes all listeners added with this.listenTo();
        this.stopListening();
    }
}

The second listener to 'execute' shows one of the very common practices in CKEditor 5 code. Basically, the default action of 'execute' (which is calling the execute() method) is registered as a listener to that event with a default priority. Thanks to that, by listening to the event using 'low' or 'high' priorities you can execute some code before or after execute() is really called. If you stop the event, then the execute() method will not be called at all. In this particular case, the Command#execute() method was decorated with the event using the ObservableMixin#decorate() function:

import ObservableMixin from '@ckeditor/ckeditor5-utils/src/observablemixin';
import mix from '@ckeditor/ckeditor5-utils/src/mix';

class Command {
    constructor() {
        this.decorate( 'execute' );
    }

    // Will now fire the #execute event automatically.
    execute() {}
}

// Mix ObservableMixin into Command.
mix( Command, ObservableMixin );

Besides decorating methods with events, observables allow to observe their chosen properties. For instance, the Command class makes its #value and #isEnabled observable by calling set():

class Command {
    constructor() {
        this.set( 'value', undefined );
        this.set( 'isEnabled', undefined );
    }
}

mix( Command, ObservableMixin );

const command = new Command();

command.on( 'change:value', ( evt, propertyName, newValue, oldValue ) => {
    console.log(
        `${ propertyName } has changed from ${ oldValue } to ${ newValue }`
    );
} )

command.value = true; // -> 'value has changed from undefined to true'

Observable properties are marked in API documentation strings with the @observable keyword but we do not mark them in API documentation (yet).

Observables have one more feature which is widely used by the editor (especially in the UI library) — the ability to bind the value of one object’s property to the value of some other property or properties (of one or more objects). This, of course, can also be processed by callbacks.

Assuming that target and source are observables and that used properties are observable:

target.bind( 'foo' ).to( source );

source.foo = 1;
target.foo; // -> 1

// Or:
target.bind( 'foo' ).to( source, 'bar' );

source.bar = 1;
target.foo; // -> 1

You can find more about bindings in the UI library architecture guide.

Once you have learned how to create plugins and commands you can read how to implement real editing features in the Editing engine guide.