pvs_agent/src/utils/page-analyzer.js

// 确保类定义在全局作用域
export default class PageAnalyzer {
  constructor() {
    this.readability = null;
  }

  /**
   * 分析页面内容
   * @returns {Object} 页面分析结果
   */
  analyzePage(iframe = false) {
    try {
      // 检查Readability是否可用
      if (typeof Readability === "undefined") {
        console.warn(
          "Readability not loaded, falling back to basic extraction"
        );
        return this.fallbackAnalysis(iframe);
      }

      // 创建文档副本以避免修改原始DOM
      const documentClone = document.cloneNode(true);

      // 初始化 Readability
      this.readability = new Readability(documentClone, {
        debug: false,
        charThreshold: 20,
      });

      // 解析页面
      const article = this.readability.parse();

      return {
        title: article.title || document.title,
        url: window.location.href,
        mainContent: article.textContent || article.excerpt || "",
        excerpt: article.excerpt || "",
        siteName: article.siteName || new URL(window.location.href).hostname,
        wordCount: article.length || 0,
      };
    } catch (error) {
      console.warn("Readability failed, using fallback:", error);
      return this.fallbackAnalysis();
    }
  }

  // 基础提取方法作为后备
  fallbackAnalysis(iframe) {
    return {
      title: document.title,
      url: window.location.href,
      mainContent: this.extractMainContent(iframe),
      excerpt: "",
      siteName: new URL(window.location.href).hostname,
      wordCount: 0,
    };
  }

  // 基础的内容提取方法
  extractMainContent(iframe) {
    // 移除脚本、样式等

    //提取页面的表单
    if (iframe) {
      const form = document.querySelector("form");
      if (form) {
        content.textContent = form.outerHTML;
      }
    } else {
      const content = document.body.cloneNode(true);
      cleanPage(content)
      // .trim().replace(/\s+/g, " ")
      // content
      //     .querySelectorAll("script, style, iframe, nav, header, footer,svg")
      //     .forEach((el) => el.remove());

      const regex = /\s+/g;

      // 定义标准 HTML 属性集合 排除id class style href src target
      const standardAttributes = new Set(['alt', 'title', 'type', 'value', 'name', 'placeholder', 'disabled', 'checked', 'selected', 'readonly', 'required', 'maxlength', 'min', 'max', 'step', 'pattern', 'autocomplete', 'autofocus', 'multiple', 'rows', 'cols', 'rel', 'aria-*']);

      // 创建一个临时容器
      const temp = document.createElement('div');
      temp.innerHTML = content.outerHTML;
      // 递归删除空标签
      function removeEmpty(element) {
        const children = Array.from(element.children);
        children.forEach(child => {
          removeEmpty(child); // 递归处理子元素
        });

        // 检查标签是否为空
        if (!element.innerHTML.trim()) {
          element.remove(); // 删除空标签
        }
      }

      // 从临时容器的子元素开始处理
      removeEmpty(temp);

      // 删除 <link> 标签
      const linkTags = temp.querySelectorAll('link');
      linkTags.forEach(link => {
        link.remove();
      });

      // 遍历所有元素
      const elements = temp.querySelectorAll('*');
      elements.forEach(element => {
        // 获取所有属性
        const attributes = Array.from(element.attributes);
        attributes.forEach(attr => {
          // 如果属性不是标准属性，则移除
          if (!standardAttributes.has(attr.name) && !attr.name.startsWith('aria-')) {
            element.removeAttribute(attr.name);
          }
        });
      });

      // 获取处理后的 HTML 字符串
      const cleanedHtml = temp.innerHTML.trim().replace(regex, " ");

      // 销毁临时容器
      temp.remove();

      console.log(cleanedHtml)
      // content.outerHTML.trim().replace(/\s+/g, " ");
      return cleanedHtml;
    }


  }
};

function cleanPage(body) {

  // 移除所有行内样式
  const elementsWithInlineStyle = body.querySelectorAll('[style]');
  elementsWithInlineStyle.forEach(element => {
    element.removeAttribute('style');
  });

  // 移除所有注释节点
  const comments = [];
  const walk = document.createTreeWalker(body, NodeFilter.SHOW_COMMENT, null, false);
  while (walk.nextNode()) {
    comments.push(walk.currentNode);
  }
  comments.forEach(comment => comment.remove());

  // 移除所有SVG图标
  body.querySelectorAll("script, style, iframe, nav, header, footer,svg").forEach(svg => svg.remove());

  // 移除所有图片的src属性
  const images = body.querySelectorAll('img');
  images.forEach(img => {
    img.removeAttribute('src');
  });

  // 可选择移除其他无关内容，比如脚本和广告等
  // const scripts = body.querySelectorAll('script');
  // scripts.forEach(script => script.remove());

  // const iframes = body.querySelectorAll('iframe');
  // iframes.forEach(iframe => iframe.remove());

  const ads = body.querySelectorAll('.ad, .advertisement, .ads');
  ads.forEach(ad => ad.remove());

  console.log('页面清理完成！');
}