jinshen-website/app/utils/parseTable.ts

import type { Element, Node } from 'domhandler';

export interface ParsedTable {
  headers: string[];
  rows: Record<string, string>[];
}

const isElement = (n: Node): n is Element => n.type === 'tag';

export function parseHtmlTable(table: Element): ParsedTable {
  if (table.name !== 'table') {
    throw new Error('parseHtmlTable: node is not <table>');
  }
  // 获取 table 下的所有 tag 子节点
  const getChildren = (node: Node): Element[] =>
    isElement(node) ? node.children.filter(isElement) : [];

  // ---- 找 thead / tbody ----
  const children = getChildren(table);
  const thead = children.find((n) => n.name === 'thead');
  const tbody = children.find((n) => n.name === 'tbody');

  if (!tbody) {
    return { headers: [], rows: [] };
  }

  const bodyRows = getChildren(tbody).filter((n) => n.name === 'tr');

  if (bodyRows.length === 0) {
    return { headers: [], rows: [] };
  }
  // ---- 1. 表头 ----
  let headerCells: Element[] = [];

  if (thead) {
    // 如果 Directus 有 thead（一般不会）
    const headerRow = getChildren(thead).find((n) => n.name === 'tr');
    headerCells = headerRow
      ? getChildren(headerRow).filter((n) => n.name === 'th' || n.name === 'td')
      : [];
  } else {
    // Directus 情况：没有 thead → 用 tbody 第一行作为 header
    headerCells = getChildren(bodyRows[0]).filter(
      (n) => n.name === 'th' || n.name === 'td'
    );
  }

  logger.info(headerCells);

  const headers = headerCells.map((cell, i) => {
    const text = cell.children
      .filter((c) => c.type === 'text')
      .map((t) => t.data.trim())
      .filter(Boolean)
      .join('');
    return text || `列${i + 1}`;
  });

  // ---- 2. 数据行 ----
  // 如果没有 thead，则跳过第一行（它是 header）
  const dataRows = thead ? bodyRows : bodyRows.slice(1);

  const rows = dataRows.map((row) => {
    const cells = getChildren(row).filter(
      (n) => n.name === 'td' || n.name === 'th'
    );

    const record: Record<string, string> = {};

    headers.forEach((header, i) => {
      const cell = cells[i];
      if (!cell) {
        record[header] = '';
      } else {
        const text = cell.children
          .filter((c) => c.type === 'text')
          .map((t) => t.data.trim())
          .filter(Boolean)
          .join('');
        record[header] = text;
      }
    });

    return record;
  });

  logger.info(headers, rows);

  return { headers, rows };
}