Friday, 7 February 2014

Convert Excel file to CSV file using Java Code in Talend ETL - Why we need to convert Excel file using java code instead components do it ???

Hi Guys,

Aim of this post :
Converting excel file(.xlsx and/or .xls) to CSV file(.csv) using Java Code in Talend Job.

Why to convert Excel to CSV using java code in Talend ETL when you can directly input ExcelInput to CSVoutput?

B'z some times when you dump data taking from
2007 Excel work book to CSV file  OR
2007 Excel work book to Table output (Any database table : Postgres or MySQL or SQL)

the data in the few columns in the work book will fall into infinite loop and causes to give "Stack Over flow error "

What I found with Excel Work book 2007 to cause this "Stack over flow " Exception
Categories of fields in Excel 
- Few fields could be General
- Few  fields could be Date
- Few fields could be Formulas
- Few fields could be Number

Though we Tick the check box "Read excel2007 file format(xlsx)" some times we may not able to dump the all rows of data to the target .

In my case I have an Excel work book of 2007 with 2000 rows and 10+ fields where few fields like date1, date2 having date type in the excel work book and ETL was not allowed me to dump all the data to the target instead giving me "Stack Over Flow" Exception..

I thought it of Memory problem but it was not and then converted Excel 2007 work book to CSV file and dumped the data to Target.

I've directly mapped ExcelComponent Input to CSV component Output but it also given me the same " Stack Over Flow" Exception and

the above situation leads me to convert Excel Workbook to CSV file.... I've tried to

Software setup for this example
1) TOS_DI-r96646-V5.1.3
2) 2007 Excel Work book with data

Code Execution in Eclipse (Testing Externally)
1) Eclipse version :
     Version: Juno Service Release 2
     Build id: 20130225-0426
2) Code compiled/Executed on JDK 1.7 in eclipse

I've taken code from this site

You can download the same code with Eclipse project here(.zipfile)
The code uses library files of Apache POI - the Java API for Microsoft Document
You need to download the all the jar files of POI and need to add them to your eclipse project
Download jar files from here

ETL job design and code implementation of the same

1) Drag and drop "tLibraryLoad" component from Palette to Design area.
2) Add all the required jar files of Apache POI - the Java API for Microsoft Document
3) How to add multiple jar files - You can learn from here(Click this link)
4) Drag and drop "tJava" component.
5) Connect "tLibrayLoad" component to "tJava" component.
    a) Right click on "tLibraryLoad'
    b) Click on Trigger then drop On sub job ok to "tJava" component.

6) Write below java code in "tJava" component Code area ,which will convert the input Excel work book(.xlsx or .xls) to CSV file(.csv).

class CSVUtils {
     * Convert Excel spreadsheet to CSV.
   * Works for <strong>Excel 97-2003 Workbook (<em>.xls)</strong> and <strong>Excel Workbook (</em>.xlsx)</strong>.
  * Does not work for the <strong>XML Spreadsheet 2003 (*.xml)</strong> format produced by BIRT.
     * @param fileName
  * @throws InvalidFormatException
   * @throws IOException
public void convertExcelToCSV(String fileName) throws InvalidFormatException, IOException {

        BufferedWriter output = new BufferedWriter(new FileWriter(fileName.substring(0, fileName.lastIndexOf(".")) + ".csv"));

        InputStream is = new FileInputStream(fileName);

       Workbook wb = WorkbookFactory.create(is);

     Sheet sheet = wb.getSheetAt(0);

       // hopefully the first row is a header and has a full compliment of
        // cells, else you'll have to pass in a max (yuck)
        int maxColumns = sheet.getRow(0).getLastCellNum();

        for (Row row : sheet) {

           // row.getFirstCellNum() and row.getLastCellNum() don't return the
            // first and last index when the first or last column is blank
         int minCol = 0; // row.getFirstCellNum()
           int maxCol = maxColumns; // row.getLastCellNum()
           for (int i = minCol; i < maxCol; i++) {

                Cell cell = row.getCell(i);
                String buf = "";
             if (i > 0) {
                    buf = "|";

             if (cell == null) {
               } else {

                  String v = null;

                  switch (cell.getCellType()) {
                  case Cell.CELL_TYPE_STRING:
                        v = cell.getRichStringCellValue().getString();
                 case Cell.CELL_TYPE_NUMERIC:
                       if (DateUtil.isCellDateFormatted(cell)) {
                          // SimpleDateFormat datetemp = new SimpleDateFormat("yyyy-MM-dd");
                           //v = datetemp.format(cell.getDateCellValue());
                          v = cell.getDateCellValue().toString();
                        } else {
                           //v = String.valueOf(cell.getNumericCellValue());
                            v = new HSSFDataFormatter().formatCellValue( cell );
                 case Cell.CELL_TYPE_BOOLEAN:
                       v = String.valueOf(cell.getBooleanCellValue());
                 case Cell.CELL_TYPE_FORMULA:
                       v = cell.getCellFormula();
                      switch(cell.getCachedFormulaResultType()) {
                       case Cell.CELL_TYPE_NUMERIC:
                           v =   String.valueOf(cell.getNumericCellValue());
                           //v = new HSSFDataFormatter().formatCellValue( cell );
                       case Cell.CELL_TYPE_STRING:
                           v =   String.valueOf(cell.getRichStringCellValue());
                     //v = String.valueOf(cell.getDateCellValue());

                 if (v != null) {
                       buf = buf + toCSV(v);


     * Escape the given value for output to a CSV file.
    * Assumes the value does not have a double quote wrapper.
  * @return
    public String toCSV(String value) {
        String v = null;
       boolean doWrap = true;
       if (value != null) {
           v = value;
           if (v.contains("\"")) {
             v = v.replace("\"", "\"\""); // escape embedded double quotes
               doWrap = true;
           if (v.contains(",") || v.contains("\n")) {
             doWrap = true;
           if (doWrap) {
              v = "\"" + v + "\""; // wrap with double quotes to hide the comma
       return v;


CSVUtils csv = new CSVUtils();

7) You need to import all the packages to run the above code.
You can do that by Clicking " Advanced settings" of  tJava Component and writing below code at "Import" area.

//import java.util.List;

import org.apache.poi.openxml4j.exceptions.InvalidFormatException;

import java.text.SimpleDateFormat;

import org.apache.poi.hssf.usermodel.HSSFDataFormatter;

8) Save and Run the job
9) Find the converted file in the same location where you are referring your Excel Work book.

Note that when you are taking code from Eclipse console to talend you need to skip few code lines like static key word from the methods.