HtmlParser.php

HtmlParser.php

<?php
declare(strict_types=1);

class HtmlParser
{
    private \DOMDocument $dom;
    private array $errors = [];

    public function __construct(string $html)
    {
        $this->dom = new \DOMDocument();
        
        // Configura opções para melhor tratamento de HTML
        $this->dom->preserveWhiteSpace = false;
        $this->dom->strictErrorChecking = false;
        $this->dom->recover = true;
        
        // Suprime erros temporariamente
        $internalErrors = libxml_use_internal_errors(true);
        
        // Adiciona metatag de encoding para evitar problemas com caracteres especiais
        $html = '<meta charset="UTF-8">' . $html;
        
        // Carrega o HTML
        $this->dom->loadHTML($html, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD);
        
        // Captura erros para diagnóstico
        $this->errors = libxml_get_errors();
        libxml_clear_errors();
        libxml_use_internal_errors($internalErrors);
    }

    /**
     * Obtém todos os elementos por tag
     */
    public function getElementsByTagName(string $tagName): array
    {
        $elements = $this->dom->getElementsByTagName($tagName);
        return $this->convertNodeListToArray($elements);
    }

    /**
     * Obtém elementos por classe (suporta múltiplas classes)
     */
    public function getElementsByClassName(string $className): array
    {
        $elements = $this->dom->getElementsByTagName('*');
        $result = [];

        foreach ($elements as $element) {
            if ($element->nodeType === XML_ELEMENT_NODE) {
                $classes = preg_split('/\s+/', trim($element->getAttribute('class')));
                if (in_array($className, $classes, true)) {
                    $result[] = $element;
                }
            }
        }

        return $result;
    }

    /**
     * Obtém elemento por ID
     */
    public function getElementById(string $id): ?\DOMElement
    {
        return $this->dom->getElementById($id) ?: null;
    }

    /**
     * Extrai todos os estilos inline BRUTOS (sem processamento CSS)
     * Retorna a string completa do atributo style
     */
    public function getRawInlineStyles(): array
    {
        $elements = $this->dom->getElementsByTagName('*');
        $styles = [];

        foreach ($elements as $element) {
            if ($element->nodeType === XML_ELEMENT_NODE && $element->hasAttribute('style')) {
                $styles[] = [
                    'element' => $element->tagName,
                    'id' => $element->hasAttribute('id') ? $element->getAttribute('id') : null,
                    'class' => $element->hasAttribute('class') ? $element->getAttribute('class') : null,
                    'selector' => $this->buildCssSelector($element),
                    'rawStyle' => $element->getAttribute('style'),
                    'node' => $element
                ];
            }
        }

        return $styles;
    }

    /**
     * Extrai todos os estilos de tags <style> BRUTOS (sem processamento CSS)
     */
    public function getRawStyleTags(): array
    {
        $styles = [];
        foreach ($this->dom->getElementsByTagName('style') as $style) {
            $media = $style->hasAttribute('media') ? $style->getAttribute('media') : 'all';
            $styles[] = [
                'content' => $style->nodeValue,
                'media' => $media,
                'type' => $style->hasAttribute('type') ? $style->getAttribute('type') : 'text/css',
                'node' => $style
            ];
        }
        return $styles;
    }

    /**
     * Constrói um seletor CSS representativo para o elemento
     */
    private function buildCssSelector(\DOMElement $element): string
    {
        $selector = $element->tagName;
        
        if ($element->hasAttribute('id')) {
            $selector .= '#' . $element->getAttribute('id');
        }
        
        if ($element->hasAttribute('class')) {
            $classes = preg_split('/\s+/', trim($element->getAttribute('class')));
            foreach ($classes as $class) {
                $selector .= '.' . $class;
            }
        }
        
        return $selector;
    }

    /**
     * Converte NodeList para array
     */
    private function convertNodeListToArray(\DOMNodeList $nodeList): array
    {
        $result = [];
        foreach ($nodeList as $node) {
            $result[] = $node;
        }
        return $result;
    }

    /**
     * Obtém o DOMDocument subjacente
     */
    public function getDom(): \DOMDocument
    {
        return $this->dom;
    }

    /**
     * Obtém erros de parsing
     */
    public function getErrors(): array
    {
        return $this->errors;
    }
}