Node#isEmpty considers nbsp-only node as non-empty

[aduth]

Do you want to request a feature or report a bug?

Bug. (Or verification of intent).

What is the current behavior?

When a new block is created, it is padded with an   non-breaking space character. The behavior of Node#isEmpty disregards whitespace, but not specifically the non-breaking space character.

If the current behavior is a bug, please provide the steps to reproduce and if possible a minimal demo of the problem via fiddle.tinymce.com or similar.

Demo: http://fiddle.tinymce.com/Pugaab

Enter some content and press Enter.

More directly, it could be reproduced by a test case on creating an empty text node.

const documentNode = Node.create( '#document-fragment' );
const textNode = Node.create( '#text' );
textNode.value = String.fromCharCode( 160 );
documentNode.append( textNode );
console.log( documentNode.isEmpty() );
// false

Possible solution: If changing the regular expression to use \s instead of , isEmpty correctly returns true.

Before:

tinymce/src/core/main/ts/api/html/Node.ts

Line 14 in 7dfd810

const whiteSpaceRegExp = /^[ \t\r\n]*$/;

After:

const whiteSpaceRegExp = /^[\s\t\r\n]*$/;

What is the expected behavior?

The logged result should show "Is empty: true" since it is for the newly-created empty paragraph.

Which versions of TinyMCE, and which browser / OS are affected by this issue? Did this work in previous versions of TinyMCE?

TinyMCE 4.7.13
Unaware of previous working state.

[spocke]

An element is only considered empty if there is no caret positions or just white space. However nbsp is considered contents since it's a padded paragraph. Not exactly sure why you need this but you can disable the padding behavior of paragraphs by setting for example valid_elements: '*[*]' in tiny then all empty paragraphs will be keep as is.

We need a really good reason to break the isEmpty functionality and consider padded block elements as empty. So would be nice if we could get a use case where this is needed.

[aduth]

Here's a more direct example, where a brand new (empty) initialization of the editor reports as being non-empty:

http://fiddle.tinymce.com/1ygaab/1

To me, it seems like the semantics of being empty should trump whether there's a padded node there. If I initialized the editor with no content, should it not be considered empty?

I'll dig to see if there was a case in Gutenberg we encountered this. We decided against the tree format, so it may not be a pressing need at the moment, but still seems like an unexpected behavior.

[aduth]

This was originally discovered in WordPress/gutenberg#7583, which was a closed experiment of using format: 'tree' (considered failed due to non-performant behavior of tree in repeated calls), where a new block's tree node was reported as non-empty.

[Afraithe]

Hey Aduth, we are looking into this.

Trying to understand what/how you need this.

Wouldn't adding something like editor.isEmpty() not work better in this case?

As you said, your not using the format: tree functionality, so how does this manifest for you?

[aduth]

As far as prioritization, this isn't actively affecting us in Gutenberg. I do think it's a legitimate issue to consider for anyone who might want to use the tree nodes result. We've / I've explored it on a few occasions, and it seems like it would be a more compatible option, but based on previous findings we're not exploring it further at this time.