java: reading large file with charset

Question

My file is 14GB and I would like to read line by line and will be export to excel file.

As the file include different language, such as Chinese and English,
I tried to use FileInputStream with UTF-16 for reading data,
but result in java.lang.OutOfMemoryError: Java heap space
I have tried to increase the heap space but problem still exist
How should I change my file reading code?

createExcel();     //open a excel file
try {

    //success but cannot read and output for different language
    //br = new BufferedReader(
    //        new FileReader("C:\\Users\\brian_000\\Desktop\\appdatafile.json"));


    //result in java.lang.OutOfMemoryError: Java heap space
    br = new BufferedReader(new InputStreamReader(
            new FileInputStream("C:\\Users\\brian_000\\Desktop\\appdatafile.json"), 
            "UTF-16"));

} catch (FileNotFoundException e) {
    // TODO Auto-generated catch block
    e.printStackTrace();
} catch (UnsupportedEncodingException e) {
    // TODO Auto-generated catch block
    e.printStackTrace();
} 

System.out.println("cann be print");


String line;
int i=0;
try {
    while ((line = br.readLine()) != null) {
        // process the line.
        try{
            System.out.println("cannot be print");
            //some statement for storing the data in variables.



                   //a function for writing the variable into excel
writeToExcel(platform,kind,title,shareUrl,contentRating,userRatingCount,averageUserRating
                            ,marketLanguage,pricing
                            ,majorVersionNumber,releaseDate,downloadsCount);


            }
            catch(com.google.gson.JsonSyntaxException exception){
                System.out.println("error");
            }



            // trying to get the first 1000rows
            i++;

            if(i==1000){
                br.close();

                break;
            }
        }
    } catch (IOException e) {
        // TODO Auto-generated catch block
        e.printStackTrace();
    }


    closeExcel();




public static void writeToExcel(String platform,String kind,String title,String shareUrl,String contentRating,String userRatingCount,String averageUserRating
            ,String marketLanguage,String pricing,String majorVersionNumber,String releaseDate,String downloadsCount){

        currentRow++;
        System.out.println(currentRow);

        if(currentRow>1000000){
            currentsheet++;
            sheet = workbook.createSheet("apps"+currentsheet, 0);
            createFristRow();
            currentRow=1;
        }



        try {

                //character id
                Label label = new Label(0, currentRow, String.valueOf(currentRow), cellFormat);
                sheet.addCell(label);

                //12 of statements for write the data to excel
                label = new Label(1, currentRow, platform, cellFormat);
                sheet.addCell(label);




            } catch (WriteException e) {
                e.printStackTrace();
            }

Answer 1

Excel, UTF-16

As mentioned, the problem is likely caused by the Excel document construction. Try whether UTF-8 yields a lesser size; for instance Chinese HTML still is better compressed with UTF-8 rather than UTF-16 because of the many ASCII chars.

Object creation java

You can share common small Strings. Useful for String.valueOf(row) and such. Cache only strings with a small length. I assume the cellFormat to be fixed.

DIY with xlsx

Excel builds a costly DOM. If CSV text (with a Unicode BOM marker) is no options (you could give it the extension .xls to be opened by Excel), try generating an xslx. Create an example workbook in xslx. This is a zip format you can process in java easiest with a zip filesystem. For Excel there is a content XML and a shared XML, sharing cell values with an index from content to shared strings. Then no overflow happens as you write buffer-wise. Or use a JDBC driver for Excel. (No recent experience on my side, maybe JDBC/ODBC.)

Best

Excel is hard to use with that much data. Consider more effort using a database, or write every N rows in a proper Excel file. Maybe you can later import them with java in one document. (I doubt it.)