slake-school-service/app/Services/Crawl/NewsContentHtml.php

<?php

namespace App\Services\Crawl;

/**
 * 资讯正文 HTML：提取正文区域、补全图片地址、懒加载属性转换。
 */
class NewsContentHtml
{
    public static function extractBody(string $html): ?string
    {
        $candidates = [];

        $patterns = [
            // 上海交大等：Article_content + Article-source
            '#<div[^>]+class=["\'][^"\']*Article_content[^"\']*["\'][^>]*>(.*?)</div>\s*<div[^>]+class=["\'][^"\']*Article-source#is',
            '#<div[^>]+class=["\'][^"\']*Article_content[^"\']*["\'][^>]*>(.*?)</div>#is',
            '#<div[^>]+id=["\']ivs_content["\'][^>]*>(.*?)</div>#is',
            '#<div[^>]+id=["\'](?:vsb_content|v_news_content|news_content|zoom)["\'][^>]*>(.*?)</div>#is',
            '#<div[^>]+class=["\'][^"\']*v_news_content[^"\']*["\'][^>]*>(.*?)</div>#is',
            '#<article[^>]*>(.*?)</article>#is',
            '#<div[^>]+id=["\'](?:news-content|article-body|article-content|main-content|post-content)["\'][^>]*>(.*?)</div>#is',
            '#<div[^>]+class=["\'][^"\']*(?:article-body|article-content|news-content|post-content|entry-content|content-detail|detail-content)[^"\']*["\'][^>]*>(.*?)</div>#is',
        ];

        foreach ($patterns as $pattern) {
            if (preg_match($pattern, $html, $m)) {
                $body = trim($m[1]);
                $len = mb_strlen(strip_tags($body));
                if ($len >= 30) {
                    $candidates[$len] = $body;
                }
            }
        }

        if ($candidates === []) {
            return null;
        }

        ksort($candidates);

        return (string) array_pop($candidates);
    }

    public static function normalize(?string $html, ?string $pageUrl = null): ?string
    {
        if ($html === null || trim($html) === '') {
            return null;
        }

        $html = html_entity_decode($html, ENT_QUOTES | ENT_HTML5, 'UTF-8');
        $origin = self::pageOrigin($pageUrl);

        $html = preg_replace_callback(
            '/<img\b([^>]*)>/iu',
            static function (array $m) use ($origin, $pageUrl): string {
                $attrs = $m[1];

                foreach (['data-src', 'data-original', 'data-url', 'original-src'] as $attr) {
                    if (preg_match('/\s'.preg_quote($attr, '/').'=(["\'])([^"\']+)\1/i', $attrs, $data)) {
                        $real = $data[2];
                        if (preg_match('/\ssrc=(["\'])([^"\']*)\1/i', $attrs)) {
                            $attrs = preg_replace('/\ssrc=(["\'])[^"\']*\1/i', ' src="'.$real.'"', $attrs);
                        } else {
                            $attrs .= ' src="'.$real.'"';
                        }
                    }
                }

                if (preg_match('/\ssrc=(["\'])([^"\']*)\1/i', $attrs, $srcMatch)) {
                    $resolved = self::resolveAssetUrl($srcMatch[2], $pageUrl, $origin);
                    $attrs = preg_replace(
                        '/\ssrc=(["\'])[^"\']*\1/i',
                        ' src="'.$resolved.'"',
                        $attrs
                    ) ?? $attrs;
                }

                if (! preg_match('/\sreferrerpolicy=/i', $attrs)) {
                    $attrs .= ' referrerpolicy="no-referrer"';
                }

                return '<img'.$attrs.'>';
            },
            $html
        ) ?? $html;

        return trim($html);
    }

    public static function resolveAssetUrl(string $url, ?string $pageUrl, ?string $origin = null): string
    {
        $url = trim(html_entity_decode($url));
        if ($url === '') {
            return $url;
        }

        if (preg_match('#^https?://#i', $url)) {
            return $url;
        }

        $origin ??= self::pageOrigin($pageUrl);
        if ($origin === null) {
            return $url;
        }

        if (str_starts_with($url, '//')) {
            $scheme = parse_url($pageUrl ?? $origin, PHP_URL_SCHEME) ?: 'https';

            return $scheme.':'.$url;
        }

        if (str_starts_with($url, '/')) {
            return $origin.$url;
        }

        if ($pageUrl) {
            $absolute = HtmlCrawlSupport::absoluteUrl($url, $pageUrl);
            if ($absolute) {
                return $absolute;
            }
        }

        return rtrim($origin, '/').'/'.ltrim($url, '/');
    }

    protected static function pageOrigin(?string $pageUrl): ?string
    {
        if (! $pageUrl) {
            return null;
        }
        $parts = parse_url($pageUrl);
        if (! is_array($parts) || empty($parts['scheme']) || empty($parts['host'])) {
            return null;
        }

        $origin = $parts['scheme'].'://'.$parts['host'];
        if (! empty($parts['port'])) {
            $origin .= ':'.$parts['port'];
        }

        return $origin;
    }
}