Gajus/Dindent inserts extra space when an inline HTML tag like <strong> is placed directly against text

[NitinParri]

Scenario:

1. The code in RTE: <p>Lorem ipsum dolor sit amet<strong class="blue">!</strong></p>
2. config.sourceopt.formatHtml is more than 0

Current output: Lorem ipsum dolor sit amet !.
Desired output: Lorem ipsum dolor sit amet! (without extra space before !).

It is a know problem of the Dindent library.
Untill they fix the bug, you can use an XCLASS to remove the extra space:

Create file: /your_extension/Classes/Xclass/Service/CleanHtmlService.php with the following code:

<?php

declare(strict_types=1);

namespace YourVendor\YourExtension\Xclass\Service;

use HTML\Sourceopt\Service\CleanHtmlService as SourceoptCleanHtmlService;

class CleanHtmlService extends SourceoptCleanHtmlService
{
    private const BOUNDARY_MARKER = "\u{E000}";

    private const INLINE_TAGS = 'a|abbr|acronym|b|bdo|big|cite|code|dfn|em|i|kbd|samp|small|span|strong|sub|sup|tt|var';

    public function clean(string $html, array $config = [], string $doctype = ''): string
    {
        $formatHtmlEnabled = isset($config['formatHtml']) && (bool) $config['formatHtml'];

        if ($formatHtmlEnabled) {
            $html = $this->protectInlineBoundaries($html);
        }

        $html = parent::clean($html, $config, $doctype);

        if ($formatHtmlEnabled) {
            $html = $this->restoreInlineBoundaries($html);
        }

        return $html;
    }

    private function protectInlineBoundaries(string $html): string
    {
        $html = (string) preg_replace(
            '/(?<=[^\s>])(?=<(?:' . self::INLINE_TAGS . ')(?:\s[^>]*)?>)/u',
            self::BOUNDARY_MARKER,
            $html
        );

        return (string) preg_replace(
            '/<\/(?:' . self::INLINE_TAGS . ')>(?=[^\s<])/u',
            '$0' . self::BOUNDARY_MARKER,
            $html
        );
    }

    private function restoreInlineBoundaries(string $html): string
    {
        return (string) preg_replace('/\s*' . self::BOUNDARY_MARKER . '\s*/u', '', $html);
    }
}

Register the XCLASS in /your_extension/ext_localconf.php:

if (class_exists(\HTML\Sourceopt\Service\CleanHtmlService::class)) {
    $GLOBALS['TYPO3_CONF_VARS']['SYS']['Objects'][\HTML\Sourceopt\Service\CleanHtmlService::class] = [
        'className' => \YourVendor\YourExtension\Xclass\Service\CleanHtmlService::class,
    ];
}

Replace YourVendor\YourExtension with your extension and vendor and replace your_extension with your extension folder.

[chrissonntag]

Yep, upstream bug in schleuse/dindent: inline elements get re-inserted with sprintf(' %s%s%s ', …), so it always pads with spaces even when the input had none. It hits the closing side too — <a href="/x">Link</a>, text comes out as </a> , text, i.e. a space before punctuation, which is arguably the worse case.

Tried a patch locally (only re-add the space if there was one): all Dindent fixtures still pass except 9999-test, whose expected output happens to contain the bug. Needs a PR over at schleuse/dindent, they have issues disabled there.

Unrelated, but while I was in there: indent() never resets temporary_replacements_*, so reusing one Indenter for two documents corrupts the second one. sourceopt is fine only because it creates a new Indenter per call.

@xerc Would you be willing and are you able to patch schleuse/dindent respectively? If not, I can give it a shot and create a PR for it.

[xerc]

@chrissonntag yea i saw this prob. with <span> to some days ago