Importing html with nested lists causes invalid doc structure

[Coding-Kiwi]

See my comment #2 (comment)

If a <li> tag contains a <p> and a sublist <ul>, the ListItem Node wraps the content with an additional <p> tag.

The resulting json doc can be loaded into tiptap but as soon as you edit a list item, an Invalid content for node paragraph is thrown because there is now a <p> tag containing a <p> and <ul> tag which is not valid, because a <p> tag can only contain phrasing content.

[hivokas]

I'm seeing this issue as well.

> (new \Tiptap\Editor(['content' => '<p>List:</p><ul><li><p>bullet</p><ol><li><p>one</p></li></ol></li></ul>']))->getJSON()
= "{"type":"doc","content":[{"type":"paragraph","content":[{"type":"text","text":"List:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","content":[{"type":"paragraph","content":[{"type":"text","text":"bullet"}]},{"type":"orderedList","content":[{"type":"listItem","content":[{"type":"paragraph","content":[{"type":"text","text":"one"}]}]}]}]}]}]}]}"

> (new \Tiptap\Editor(['content' => '<p>List:</p><ul><li><p>bullet</p><ol><li><p>one</p></li></ol></li></ul>']))->getHTML()
= "<p>List:</p><ul><li><p><p>bullet</p><ol><li><p>one</p></li></ol></p></li></ul>"

Note how <p>bullet</p> becomes <p><p>bullet</p>.

[hivokas]

Similarly to @Coding-Kiwi, I've also noticed this issue while trying to migrate from HTML to JSON TipTap doc.

[hivokas]

@timoisik if you can take a look at this issue when you have time, it would be great.

[lukasleitsch]

I have the same issue with the sanitize method. Example:

$string = "<p>An example</p><ul><li><p>list item a</p><ul><li><p>list item b</p></li></ul></li></ul>";

$sanitizedHtml = (new Tiptap\Editor)->sanitize($string);

// output: <p>An example</p><ul><li><p><p>list item a</p><ul><li><p>list item b</p></li></ul></p></li></ul>

[lukasleitsch]

The workaround from @Coding-Kiwi in #2 (comment) solves the issue for me at this moment:

class TipTapListItem extends \Tiptap\Nodes\ListItem
{
    public static function wrapper($DOMNode)
    {
        return null;
    }
}

$string = "<p>An example</p><ul><li><p>list item a</p><ul><li><p>list item b</p></li></ul></li></ul>";

$sanitizedHtml = (new Tiptap\Editor([
        'extensions' => [
            new Tiptap\Extensions\StarterKit([
                'listItem' => false,
            ]),
            new TipTapListItem(),
        ],
    ]))->sanitize($string)

[timoisik]

Hi, thank you for reporting this. As mentioned in #2 I would love to get rid of the wrapper function as well.

The workaround you are using here is like turn off the wrapper anyways, right? The issue here seems to be not just with the wrapper function but how the whole HTML is parsed in the DOMParser.php.

I would not plan to add support for schema-based normalization on the PHP side, because that would require essentially re-implementing a large part of ProseMirror’s document model. Instead, I would drop the wrapper function entirely and expect consumers of this library to send the correct Tiptap JSON up front.

I’m open to feedback or ideas.