tihao-ai/server/utils/file-parser.ts

/**
 * 简易文件解析工具
 * 支持从上传的 Excel/Word 文件中提取文本内容
 * 注意：完整实现需要 xlsx / mammoth 等库，此处提供基础文本提取
 */

/**
 * 从文件 Buffer 提取可读文本
 * 对于 .docx 文件提取 XML 内的文本节点
 * 对于 .xlsx 文件提取 sharedStrings 中的文本
 */
export function extractTextFromBuffer(buffer: Buffer, fileName: string): string {
  const ext = fileName.substring(fileName.lastIndexOf('.')).toLowerCase();
  console.log(`[FileParser] 文件: ${fileName}, 扩展名: ${ext}, buffer大小: ${buffer.length} bytes`);

  let result: string;
  if (ext === '.docx') {
    result = extractDocxText(buffer);
  } else if (ext === '.xlsx' || ext === '.xls') {
    result = extractXlsxText(buffer);
  } else {
    // Fallback: try to read as plain text
    result = buffer.toString('utf-8').slice(0, 50000);
  }

  console.log(`[FileParser] 提取文本长度: ${result.length} 字符`);
  console.log(`[FileParser] 文本前300字: ${result.substring(0, 300)}`);
  return result;
}

/**
 * 从 .docx (ZIP 内的 XML) 提取纯文本
 * 简易实现：查找 <w:t> 标签内容
 */
function extractDocxText(buffer: Buffer): string {
  // .docx is a ZIP file; for now, do regex extraction on raw buffer
  // A production implementation would use a proper ZIP + XML parser
  const raw = buffer.toString('utf-8');
  const textParts: string[] = [];

  // Extract text between <w:t> tags
  const regex = /<w:t[^>]*>([^<]*)<\/w:t>/g;
  let match: RegExpExecArray | null;
  while ((match = regex.exec(raw)) !== null) {
    if (match[1]) {
      textParts.push(match[1]);
    }
  }

  if (textParts.length > 0) {
    return textParts.join(' ');
  }

  // Fallback: extract any readable text segments
  return extractReadableText(buffer);
}

/**
 * 从 .xlsx (ZIP 内的 sharedStrings.xml) 提取文本
 */
function extractXlsxText(buffer: Buffer): string {
  const raw = buffer.toString('utf-8');
  const textParts: string[] = [];

  // Extract text from <t> tags in shared strings
  const regex = /<t[^>]*>([^<]*)<\/t>/g;
  let match: RegExpExecArray | null;
  while ((match = regex.exec(raw)) !== null) {
    if (match[1] && match[1].trim()) {
      textParts.push(match[1].trim());
    }
  }

  if (textParts.length > 0) {
    return textParts.join(' | ');
  }

  return extractReadableText(buffer);
}

/**
 * 从二进制数据提取可读文本段
 */
function extractReadableText(buffer: Buffer): string {
  const text = buffer.toString('utf-8');
  // Filter to only printable characters and common Chinese
  const readable = text.replace(/[^\u4e00-\u9fff\u0020-\u007e\u00a0-\u00ff\n\r\t]/g, ' ');
  // Collapse whitespace
  return readable.replace(/\s+/g, ' ').trim().slice(0, 30000);
}