feat: improve XLSX to Markdown conversion by processing each sheet via CSV intermediary

Author: ralyodio

src/services/xlsx-to-markdown-service.js

@@ -4,65 +4,85 @@ import path from 'path';
 import { promisify } from 'util';
 import { v4 as uuidv4 } from 'uuid';
 import os from 'os';
+import * as XLSX from 'xlsx';
 
 const execPromise = promisify(exec);
 
 /**
  * Service for converting XLSX files to Markdown content
- * Note: Pandoc's support for XLSX is limited - it may not handle complex spreadsheets well
  */
 export const xlsxToMarkdownService = {
   /**
-   * Convert an XLSX file to Markdown content using pandoc
+   * Convert an XLSX file to Markdown content by first converting to CSV
+   * and then using pandoc to convert CSV to Markdown
    * @param {Buffer} xlsxBuffer - The XLSX file buffer to convert
    * @returns {Promise<string>} - A string containing the Markdown content
    */
   async convertToMarkdown(xlsxBuffer) {
     try {
-      // Create temporary files for input and output
+      // Create temporary directory for our files
       const tempDir = os.tmpdir();
       const inputId = uuidv4();
-      const outputId = uuidv4();
       const inputPath = path.join(tempDir, `${inputId}.xlsx`);
-      const outputPath = path.join(tempDir, `${outputId}.md`);
 
       // Write XLSX buffer to temporary file
       await fs.promises.writeFile(inputPath, xlsxBuffer);
 
-      // Try to use pandoc to convert XLSX to Markdown
-      // Note: This may fail as pandoc has limited XLSX support
-      // We'll attempt it but provide a fallback error message
-      try {
-        const command = `pandoc -f xlsx -t markdown "${inputPath}" -o "${outputPath}"`;
+      // Read the workbook using xlsx package
+      const workbook = XLSX.readFile(inputPath);
+      
+      // Array to store markdown content from each sheet
+      const markdownParts = [];
+      
+      // Process each sheet in the workbook
+      for (const sheetName of workbook.SheetNames) {
+        console.log(`Processing sheet: ${sheetName}`);
+        
+        // Get the worksheet
+        const worksheet = workbook.Sheets[sheetName];
+        
+        // Convert worksheet to CSV
+        const csvContent = XLSX.utils.sheet_to_csv(worksheet);
+        
+        // Create temporary files for CSV and Markdown
+        const csvId = uuidv4();
+        const mdId = uuidv4();
+        const csvPath = path.join(tempDir, `${csvId}.csv`);
+        const mdPath = path.join(tempDir, `${mdId}.md`);
+        
+        // Write CSV content to file
+        await fs.promises.writeFile(csvPath, csvContent);
+        
+        // Use pandoc to convert CSV to Markdown
+        const command = `pandoc -f csv -t markdown "${csvPath}" -o "${mdPath}"`;
         console.log(`Executing pandoc command: ${command}`);
 
         await execPromise(command);
 
         // Read the generated Markdown file
-        const markdownContent = await fs.promises.readFile(outputPath, 'utf8');
+        const sheetMarkdown = await fs.promises.readFile(mdPath, 'utf8');
 
-        // Clean up temporary files
-        try {
-          await fs.promises.unlink(inputPath);
-          await fs.promises.unlink(outputPath);
-        } catch (cleanupError) {
-          console.warn('Error cleaning up temporary files:', cleanupError);
-        }
-        
-        return markdownContent;
-      } catch (pandocError) {
-        // If pandoc doesn't support XLSX directly, provide a helpful error
-        console.warn('Pandoc XLSX conversion failed, this format may not be fully supported:', pandocError.message);
+        // Add sheet name as header and append to markdown parts
+        markdownParts.push(`# ${sheetName}\n\n${sheetMarkdown}\n\n`);
 
-        // Clean up input file
+        // Clean up temporary CSV and MD files
         try {
-          await fs.promises.unlink(inputPath);
+          await fs.promises.unlink(csvPath);
+          await fs.promises.unlink(mdPath);
         } catch (cleanupError) {
-          console.warn('Error cleaning up input file:', cleanupError);
+          console.warn('Error cleaning up temporary files:', cleanupError);
         }
-        
-        throw new Error('XLSX to Markdown conversion is not fully supported by pandoc. Consider converting the Excel file to CSV or another supported format first.');
       }
+      
+      // Clean up the input XLSX file
+      try {
+        await fs.promises.unlink(inputPath);
+      } catch (cleanupError) {
+        console.warn('Error cleaning up input file:', cleanupError);
+      }
+      
+      // Combine all markdown parts
+      return markdownParts.join('---\n\n');
     } catch (error) {
       console.error('Error converting XLSX to Markdown:', error);