Choice

A module for unobtrusively saving and restoring selections in contenteditable elements.

Installation

$ npm install choice

How it works

Consider the layout of a simple rich text editor; there is a document which contains blocks of text (paragraphs, headings, etc.), and, in these blocks, text with various styles (bold, italic, etc.). In HTML, this might look like:

<!-- The article represent the editor "document" -->
<article contenteditable="true">
 
    <!-- These are the blocks of text -->
    <p>Some <strong>rich</strong> text.</p>
    <h2>This is a heading</h2>
    <p><em>More</em> text. Ladida.</p>
</article>

Choice represents the endpoints of a selection as an integer pair [childIndex, textIndex], where childIndex is the index, relative to the “document”, of the block which contains the endpoint, and textIndex is the number of characters in the block before the endpoint. If the selection spanned the word “This” in the example above, the selection would be represented as:

{
    start: [1, 0],
    end: [1, 4]
}

API

new Choice( element[, getChildren ] )

Creates an instance of Choice. element is the root contenteditable element that represent the “document” of the editor.

Although working with child indices may work for simple use cases, like the example above, the shortcomings of that method quickly become evident when your editor produces more complex markup. Consider:

<article contenteditable="true">
    <section>
        <p>Some <strong>rich</strong> text.</p>
        <p>I can edit it.</p>
    </section>
    <section>
        <hr contenteditable="false">
        <h2>This is a new section</h2>
        <p>More text. Ladida.</p>
    </section>
</article>

In this case, using child indices wouldn’t work; the “blocks” of text aren’t direct children of the “document”. For these situations, Choice takes a second parameter, getChildren, a function that return an array containing the relevant “blocks” of text. For the above example, getChildren might look like:

function getChildren() {
    var article = document.querySelector('article'),
        children = article.querySelectorAll('p, h2')
 
    return Array.prototype.slice.call(children)
}

If a getChildren function is not given, Choice defaults to using the root element’s child nodes.

Choice#getSelection( )

Returns an an instance of Choice.Selection. This has two properties, start and end, which contain the two integer pairs representing the start and end points of the selection. See below for more information on Choice.Selection.

If the user’s selection is not contained within the root element, getSelection returns null.

Choice#restore( savedSelection )

Sets the user’s selection to match that represented by the given instance of Choice.Selection.

---

Selection

Choice#getSelection() returns an instance of Choice.Selection.

Selection#start

An array containing the integer pair corresponding to the start of the selection.

Selection#end

An array containing the integer pair corresponding to the end of the selection.

Selection#absoluteStart

In a right-to-left selection, Selection#start is after Selection#end. Selection#absoluteStart is a getter that returns the endpoint that occurs first, visually, in the document.

Selection#absoluteEnd

As above, but returns the last endpoint.

Selection#isCollapsed

Getter that returns a boolean indicating whether the endpoints of the selection are identical.

Selection#isBackwards

Getter that returns a boolean indicating whether the selection represents a selection in the direction opposite to the text direction.

Selection#clone( )

Returns a new selection identical to the selection this method was called on.

Selection#equals( other )

Returns a boolean indicating whether or not other and the selection this method was called on represent the same selection.

Selection.equals( first, second )

This static method is similar to Selection#equals, but can be used to determine the equality of two null selections.

Motivation

Choice’s main purpose is to unobtrusively save and restore the selection in contenteditable regions. This can help “standardize” certain behaviours when working with the contenteditable API. Consider the following markup; it’s faily typical in rich text editing.

<p>Some <b><i>fancy</i></b> text</p>

Note that when the caret is placed at the end of ‘fancy’, there are actually three ways it can be placed (| represent the caret):

<p>Some <b><i>fancy</i></b>| text</p>
<!-- OR -->
<p>Some <b><i>fancy</i>|</b> text</p>
<!-- OR -->
<p>Some <b><i>fancy|</i></b> text</p>

You’ll often have no control over which of these situations occur, which can lead to very inconsistent behaviour; when the caret is at the end of ‘fancy’, will text a user enters be bold, italic and bold, or plain?

Saving and restoring the selection with Choice, however, normalizes this behaviour. In general, a restored selection tries to “lean left.” A couple examples:

<!-- This: -->
<p>Some <b>bold</b>| text</p>
<!-- becomes this: -->
<p>Some <b>bold|</b> text</p>
 
<!-- and this: -->
<p>Some <b>|bold</b> text</p>
<!-- becomes this: -->
<p>Some |<b>bold</b> text</p>

There are some exceptions to this rule (usually involving line breaks, <br>s), but they should be intuitive.

Note: Choice does not normalize the selection every time it changes; rather, it gives you a mechanism to do it yourself.

A more interesting use case, perhaps, is to manipulate elements without losing a user’s selection. Consider the following situation, in which a user might press a button to make text bold or to change the paragraph to a heading:

<article contenteditable="true">
    <p>I want to make |this block| a heading.</p>
    <p>This stuff is not as important.</p>
</article>

Naively swapping the paragraph for a heading would (probably, depending on the browser) lose the selection. To turn the paragraph into a heading without losing the selection, save the selection using Choice#getSelection, swap out the paragraph, and restore the selection using Choice#restore.

Caveats

There are some restriction to saving and restoring the selection. Anything that would mess up the integer pairs representing the endpoints of the selection will result in a poorly restored selection. This includes, but may not be limited to, inserting/removing “blocks” or inserting/removing text. If you plan on doing those things, you should update the saved selection manually to account for your changes.

Browser support

Choice works best in browsers implement the native Selection#extend method (all browsers except Internet Explorer). In Internet Explorer, Choice identically except that restored selections do not have any directionality. See the following for an example of what this means:

<!-- Starting with a collapsed selection: -->
<p>12|34</p>
 
<!-- Shift+Right; the selection is now left-to-right. -->
<p>12|3|4</p>
 
<!-- Save and restore the selection. -->
 
<!--
In browsers that support Selection#extend, the restored selection “knows” that
it was left-to-right; pressing Shift-Left collapses the selection back to its
origin:
-->
<p>12|34</p>
 
<!--
In Internet Explorer, because the selection has no directionality, the
selection *always* gets extended:
-->
<p>1|23|4</p>

This problem will only manifest itself when saving and restoring the selection excessively often; be mindful of this issue, and try to save and restore the selection only when necessary.

License

MIT