laralog/app/Blog/Support/SablogHtmlNormalizer.php

<?php

declare(strict_types=1);

namespace App\Blog\Support;

use App\Blog\Services\PostContentRenderer;
use DOMDocument;
use DOMElement;
use DOMNode;
use DOMXPath;
use League\HTMLToMarkdown\HtmlConverter;
use Throwable;

/**
 * Cleans leftover SaBlog-X HTML so imported posts render without literal &nbsp;
 * and without SyntaxHighlighter markup stretching the page.
 */
class SablogHtmlNormalizer
{
    /**
     * @return array<string, string>
     */
    private function languageAliases(): array
    {
        return [
            'js' => 'javascript',
            'javascript' => 'javascript',
            'xml' => 'xml',
            'html' => 'xml',
            'xhtml' => 'xml',
            'css' => 'css',
            'php' => 'php',
            'sql' => 'sql',
            'java' => 'java',
            'py' => 'python',
            'python' => 'python',
            'c#' => 'csharp',
            'cs' => 'csharp',
            'bash' => 'bash',
            'shell' => 'bash',
            'sh' => 'bash',
            'text' => 'text',
        ];
    }

    public function normalize(string $content): string
    {
        if ($content === '') {
            return '';
        }

        $content = $this->decodeNbsp($content);

        if (! preg_match('/<\s*(div|p|ol|ul|li|pre|code|span|br|table)\b/i', $content)) {
            return $this->recoverFlattenedShell($content);
        }

        try {
            $content = $this->transformHtml($content);
        } catch (Throwable) {
            // Keep the decoded text when HTML parsing fails.
        }

        return $this->recoverFlattenedShell($content);
    }

    public function toMarkdown(string $html, HtmlConverter $converter, PostContentRenderer $renderer): string
    {
        return $converter->convert($renderer->toMarkdownSafeTokens($this->normalize($html)));
    }

    public function plainText(string $content): string
    {
        $text = $this->decodeNbsp(strip_tags($content));
        $text = html_entity_decode($text, ENT_QUOTES | ENT_HTML5, 'UTF-8');
        $text = str_replace("\xC2\xA0", ' ', $text);
        $text = preg_replace('/\s+/u', ' ', $text) ?? $text;

        return trim($text);
    }

    public function decodeNbsp(string $content): string
    {
        $content = str_replace(
            ['&amp;nbsp;', '&nbsp;', '&#160;', '&#xA0;', '&#xa0;', "\xC2\xA0"],
            ' ',
            $content
        );

        return str_replace('&amp;nbsp;', ' ', $content);
    }

    private function transformHtml(string $html): string
    {
        $document = $this->loadDocument($html);
        $xpath = new DOMXPath($document);

        $this->convertHighlighterBlocks($document, $xpath);
        $this->convertHighlighterLists($document, $xpath);
        $this->convertStackedCodeDivs($document);
        $this->removeEmptyBlocks($xpath);

        return $this->innerHtml($document);
    }

    private function convertHighlighterBlocks(DOMDocument $document, DOMXPath $xpath): void
    {
        $nodes = [];
        foreach ($xpath->query('//*[contains(concat(" ", normalize-space(@class), " "), " codeText ")]') ?: [] as $node) {
            if ($node instanceof DOMElement) {
                $nodes[] = $node;
            }
        }

        foreach ($nodes as $node) {
            if ($node->parentNode === null) {
                continue;
            }

            $this->replaceWithPre($document, $xpath, $node);
        }
    }

    private function convertHighlighterLists(DOMDocument $document, DOMXPath $xpath): void
    {
        $lists = [];
        foreach ($xpath->query('//ol[contains(@class, "dp-")]') ?: [] as $node) {
            if (! $node instanceof DOMElement || $node->parentNode === null) {
                continue;
            }

            if ($this->closestTag($node, 'pre') !== null) {
                continue;
            }

            $lists[] = $node;
        }

        foreach ($lists as $list) {
            $this->replaceWithPre($document, $xpath, $list);
        }
    }

    private function convertStackedCodeDivs(DOMDocument $document): void
    {
        $xpath = new DOMXPath($document);
        $parents = [];
        foreach ($xpath->query('//*') ?: [] as $node) {
            if ($node instanceof DOMElement) {
                $parents[] = $node;
            }
        }

        foreach ($parents as $parent) {
            if ($parent->parentNode === null) {
                continue;
            }

            if (in_array(strtolower($parent->nodeName), ['pre', 'code'], true)) {
                continue;
            }

            $run = [];
            $flush = function () use (&$run, $document): void {
                if (count($run) < 2) {
                    $run = [];

                    return;
                }

                $lines = array_map(fn (DOMElement $div): string => $this->plainText($this->innerHtmlOf($div)), $run);
                if (! $this->looksLikeCodeLines($lines)) {
                    $run = [];

                    return;
                }

                $first = $run[0];
                $pre = $this->makePre($document, implode("\n", $lines), $this->detectLanguage('', implode("\n", $lines)));
                $first->parentNode?->insertBefore($pre, $first);
                foreach ($run as $div) {
                    $div->parentNode?->removeChild($div);
                }
                $run = [];
            };

            foreach (iterator_to_array($parent->childNodes) as $child) {
                if ($child instanceof DOMNode && $child->nodeType === XML_TEXT_NODE && trim((string) $child->textContent) === '') {
                    continue;
                }

                if ($child instanceof DOMElement && strtolower($child->nodeName) === 'div' && $this->isSimpleLineDiv($child)) {
                    $run[] = $child;
                    continue;
                }

                $flush();
            }

            $flush();
        }
    }

    private function removeEmptyBlocks(DOMXPath $xpath): void
    {
        foreach ($xpath->query('//p|//div') ?: [] as $node) {
            if (! $node instanceof DOMElement || $node->parentNode === null) {
                continue;
            }

            if ($node->getElementsByTagName('*')->length > 0) {
                continue;
            }

            $text = trim(str_replace("\xC2\xA0", ' ', $node->textContent));
            if ($text === '') {
                $node->parentNode->removeChild($node);
            }
        }
    }

    private function replaceWithPre(DOMDocument $document, DOMXPath $xpath, DOMElement $node): void
    {
        $hint = '';
        if (preg_match('/dp-([a-z0-9#+]+)/i', $node->getAttribute('class'), $match)) {
            $hint = $match[1];
        }

        $lines = [];
        foreach ($xpath->query('.//li', $node) ?: [] as $item) {
            $lines[] = $this->plainText($document->saveHTML($item) ?: '');
        }

        if ($lines === []) {
            foreach ($xpath->query('.//*[contains(@class, "codeHead")]', $node) ?: [] as $head) {
                $head->parentNode?->removeChild($head);
            }
            $lines = preg_split('/\R/u', $this->plainText($node->textContent)) ?: [];
        }

        $code = implode("\n", $lines);
        $pre = $this->makePre($document, $code, $this->detectLanguage($hint, $code));
        $node->parentNode?->replaceChild($pre, $node);
    }

    private function makePre(DOMDocument $document, string $code, string $language): DOMElement
    {
        $pre = $document->createElement('pre');
        $codeNode = $document->createElement('code');
        $codeNode->setAttribute('class', 'language-'.$language);
        $codeNode->appendChild($document->createTextNode($code));
        $pre->appendChild($codeNode);

        return $pre;
    }

    /**
     * @param  list<string>  $lines
     */
    private function looksLikeCodeLines(array $lines): bool
    {
        $codeish = 0;
        foreach ($lines as $line) {
            if ($line === '') {
                continue;
            }

            if (preg_match('/\\\\$/', $line)
                || preg_match('/^\s*(docker|npm|npx|pnpm|yarn|git|curl|wget|sudo|ssh|kubectl|composer|php|python|pip|node)\b/i', $line)
                || preg_match('/^\s+--[\w-]+/', $line)
                || preg_match('/^\s*\|/', $line)
                || preg_match('/[{};]$/', $line)
            ) {
                $codeish++;
            }
        }

        return $codeish >= 2;
    }

    private function isSimpleLineDiv(DOMElement $div): bool
    {
        foreach ($div->childNodes as $child) {
            if ($child instanceof DOMElement && ! in_array(strtolower($child->nodeName), ['br', 'span', 'code', 'b', 'i', 'em', 'strong'], true)) {
                return false;
            }
        }

        $text = $this->plainText($div->textContent);

        return $text !== '';
    }

    private function detectLanguage(string $hint, string $code): string
    {
        if (preg_match('/^\s*(docker|kubectl|npm|npx|pnpm|yarn|composer|curl|wget|sudo|ssh)\b/m', $code)) {
            return 'bash';
        }

        $aliases = $this->languageAliases();
        $key = strtolower(trim($hint));

        return $aliases[$key] ?? ($key !== '' ? $key : 'text');
    }

    /**
     * Recover a docker/shell command that HtmlConverter flattened into one line.
     */
    private function recoverFlattenedShell(string $content): string
    {
        if (str_contains($content, '<pre')) {
            return $content;
        }

        return preg_replace_callback(
            '/(?:docker|sudo docker)\s+run\b(?:[^<\n]*\\\\)+\s*[^<\n]+/i',
            function (array $match): string {
                $code = trim($this->plainText($match[0]));
                $code = preg_replace('/\\\\\s+/', "\\\n", $code) ?? $code;

                return '<pre><code class="language-bash">'.e($code).'</code></pre>';
            },
            $content
        ) ?? $content;
    }

    private function closestTag(DOMNode $node, string $tag): ?DOMElement
    {
        $current = $node->parentNode;
        while ($current instanceof DOMElement) {
            if (strtolower($current->nodeName) === $tag) {
                return $current;
            }
            $current = $current->parentNode;
        }

        return null;
    }

    private function loadDocument(string $html): DOMDocument
    {
        $document = new DOMDocument('1.0', 'UTF-8');
        $previous = libxml_use_internal_errors(true);
        $wrapped = '<div id="ll-root">'.$html.'</div>';
        $document->loadHTML('<?xml encoding="UTF-8">'.$wrapped, LIBXML_NOERROR | LIBXML_HTML_NODEFDTD);
        libxml_clear_errors();
        libxml_use_internal_errors($previous);

        return $document;
    }

    private function innerHtml(DOMDocument $document): string
    {
        $root = $document->getElementById('ll-root');
        if (! $root instanceof DOMElement) {
            $root = (new DOMXPath($document))->query('//*[@id="ll-root"]')->item(0);
        }

        if (! $root instanceof DOMElement) {
            return '';
        }

        return $this->innerHtmlOf($root);
    }

    private function innerHtmlOf(DOMElement $element): string
    {
        $html = '';
        foreach ($element->childNodes as $child) {
            $html .= $element->ownerDocument?->saveHTML($child) ?: '';
        }

        return $html;
    }
}