Java - Decimal Format.parse to return double value

Question

I want to be able to convert a string to a Double given a number of decimal places in a format string. So "###,##0.000" should give me a Double to 3 decimal places.

Edit - added more info to what happens

The user enters the value in the UI - which is input into a String. The rule is this value is limited to 3 decimal places. The underlying code stores the value in the database which is then used in a calculation. Therefore the trailing decimal places will cause the calculations to be out slightly to what would be expected.

I have the following code:

try {
        // output current locale we are running under (this happens to be "nl_BE")
        System.out.println( "Current Locale is " + Locale.getDefault().toString() );

        // number in Central European Format with a format string specified in UK format
        String numberCE = "1,234567"; // 1.234567
        String formatUK = "###,##0.000";

        // do the format
        DecimalFormat formatterUK = new DecimalFormat( formatUK );
        Double valCEWithUKFormat = formatterUK.parse( numberCE ).doubleValue();

        // I want the number to DPs in the format string!!!
        System.out.println( "CE Value     " + numberCE + " in UK format (" + formatUK + ") is "
        + valCEWithUKFormat );

    } catch( ParseException ex ) {
        System.out.println("Cannot parse number");
    }
}

The DecimalFormat seems to ignore the format string and gives me the complete string as a Double of 1.234567.

Can DecimalFormat be forced to use the format string when parsing? Am I missing something?

Cheers,

Andez

Answer 1

Sure you can. Try running this:

String in = "1,234567";
System.out.println(NumberFormat.getNumberFormat(new Locale("fr", "FR")).parse(in));
System.out.println(NumberFormat.getNumberFormat(new Locale("en", "GB")).parse(in));

Clearly they result in different output, the first reading 1.234567 and the second 1234567. Maybe there's something wrong with your pattern? Anyway the last line there would be the preferred way of getting the UK standard format.

Answer 2

The restriction of decimal places in DecimalFormat is really meant for use in the format() method and doesn't have much effect in the parse() method.

In order to get what you want you need this:

    try {
        // output current locale we are running under (this happens to be "nl_BE")
        System.out.println("Current Locale is " + Locale.getDefault().toString());

        // number in Central European Format with a format string specified in UK format
        String numberCE = "1,234567"; // 1.234567
        String formatUK = "###,##0.000";

        // do the format
        DecimalFormat formatterUK = new DecimalFormat(formatUK);
        Double valCEWithUKFormat = formatterUK.parse(numberCE).doubleValue();

        // first convert to UK format string
        String numberUK = formatterUK.format(valCEWithUKFormat);
        // now parse that string to a double
        valCEWithUKFormat = formatterUK.parse(numberUK).doubleValue();

        // I want the number to DPs in the format string!!!
        System.out.println("CE Value     " + numberCE + " in UK format (" + formatUK + ") is " + valCEWithUKFormat);

    } catch (ParseException ex) {
        System.out.println("Cannot parse number");
    }

You first need to get the number as a UK format string and then parse that number, using the UK formatter. That will get you the result you're looking for. NB, this will round the number to 3 decimal places, not truncate.

By the way, I'm slightly surprised that your UK formatter is able to parse the CE format number. You really should be parsing the original number with a CE format parser.

Answer 3

Taking on board what you said I have modified my code slightly to cover different locales. The key was taking a value string in a localised format to a Double that is rounded based on the format string.

The format string is always a UK based format with the decimal seperators specified as "." and thousand seperators specified as ",".

I am using the DecimalFormat to initially parse the localised format based on a specified locale. This gives a Double equivalent of the string correctly. I then use a BigDecimal to handle the rounding. I can get the number of decimal places from the DecimalFormat instance and call setScale on the BigDecimal to perform the rounding.

The initial code structure has been modified to allow you to see what happens under different locale circumstances thanks @RD01 for noting importance of other locales.

I now have code as follows:

private void runTests3() {
    // output current locale we are running under
    System.out.println( "Current Locale is " + Locale.getDefault().toString() );

    // number in Central European Format with a format string specified in UK format
    String numbersInEuropeanFormatString[] = new String[] { "1.000,234567", "1,2345678", "1.222.333,234567" };
    String formatUK = "###,##0.0000";

    // output numbers using the german locale
    System.out.println("Output numbers using the German locale\n");
    for(String num : numbersInEuropeanFormatString ) {
        formatNumberAsDouble(num, formatUK, Locale.GERMAN);
    }

    // output numbers using the UK locale.  
    // this should return unexpected results as the number is in European format
    System.out.println("Output numbers using the UK locale\n");
    for(String num : numbersInEuropeanFormatString ) {
        formatNumberAsDouble(num, formatUK, Locale.UK);
    }

    // output numbers using new DecimalFormat( formatUK ) - no locale specified
    System.out.println("\n\nOutput numbers using new DecimalFormat( " + formatUK + " )\n");
    for(String num : numbersInEuropeanFormatString ) {
        formatNumberAsDouble( num, formatUK, null);
    }
}

private void formatNumberAsDouble(String value, String format, Locale locale) {


    NumberFormat formatter;
    int decimalPlaces;

    // create the formatter based on the specified locale
    if( locale != null ) {
         formatter = NumberFormat.getNumberInstance(locale);
         // creating the above number format does not take in the format string
         // so create a new one that we won't use at all just to get the
         // decimal places in it
         decimalPlaces = (new DecimalFormat(format)).getMaximumFractionDigits();
    } else {
        formatter = new DecimalFormat( format );
        decimalPlaces = formatter.getMaximumFractionDigits();
    }

    // get the result as number
    Double result = null;
    try {
        result = formatter.parse( value ).doubleValue();
    } catch( ParseException ex ) {
        // not bothered at minute
    }

    // round the Double to the precision specified in the format string


    BigDecimal bd = new BigDecimal(result );
    Double roundedValue = bd.setScale( decimalPlaces, RoundingMode.HALF_UP ).doubleValue();

    // output summary
    System.out.println("\tValue = " + value);
    System.out.println( locale == null  ? "\tLocale not specified" : "\tLocale = " + locale.toString());
    System.out.println( format == null || format.length() == 0 ? "\tFormat = Not specified" : "\tFormat = " + format);
    System.out.println("\tResult (Double) = " + result);
    System.out.println("\tRounded Result (Double) (" + decimalPlaces + "dp) = " + roundedValue);
    System.out.println("");
}

This produces the following output:

Current Locale is nl_BE
Output numbers using the German locale

    Value = 1.000,234567
    Locale = de
    Format = ###,##0.0000
    Result (Double) = 1000.234567
    Rounded Result (Double) (4dp) = 1000.2346

    Value = 1,2345678
    Locale = de
    Format = ###,##0.0000
    Result (Double) = 1.2345678
    Rounded Result (Double) (4dp) = 1.2346

    Value = 1.222.333,234567
    Locale = de
    Format = ###,##0.0000
    Result (Double) = 1222333.234567
    Rounded Result (Double) (4dp) = 1222333.2346

Output numbers using the UK locale

    Value = 1.000,234567
    Locale = en_GB
    Format = ###,##0.0000
    Result (Double) = 1.0
    Rounded Result (Double) (4dp) = 1.0

    Value = 1,2345678
    Locale = en_GB
    Format = ###,##0.0000
    Result (Double) = 1.2345678E7
    Rounded Result (Double) (4dp) = 1.2345678E7

    Value = 1.222.333,234567
    Locale = en_GB
    Format = ###,##0.0000
    Result (Double) = 1.222
    Rounded Result (Double) (4dp) = 1.222



Output numbers using new DecimalFormat( ###,##0.0000 )

    Value = 1.000,234567
    Locale not specified
    Format = ###,##0.0000
    Result (Double) = 1000.234567
    Rounded Result (Double) (4dp) = 1000.2346

    Value = 1,2345678
    Locale not specified
    Format = ###,##0.0000
    Result (Double) = 1.2345678
    Rounded Result (Double) (4dp) = 1.2346

    Value = 1.222.333,234567
    Locale not specified
    Format = ###,##0.0000
    Result (Double) = 1222333.234567
    Rounded Result (Double) (4dp) = 1222333.2346

Answer 4

DecimalFormat is used for two distinct purposes: parsing input and formatting output. If you want to do both, you'll have to use the format object twice.

If you want to take that value and format the output, restricting the number of significant digits, you need to use the format object again. This time it uses your formatting rules to create an output string from a numeric value:

String output = formatterUK.format(valCEWithUKFormat.doubleValue() );

This will give you the output of 1,235

It seems you want this numeric value to be presented in the 1.235 format. To do this, you should format the output using a specific locale (if yours uses a different format).

HOWEVER, I would recommend approaching this problem differently:

String text = "1,234567";
NumberFormat nf_in = NumberFormat.getNumberInstance(Locale.GERMANY);
double val = nf_in.parse(text).doubleValue();

NumberFormat nf_out = NumberFormat.getNumberInstance(Locale.UK);
nf_out.setMaximumFractionDigits(3);
String output = nf_out.format(val);

A few notes:

• Input parsing should be kept separate from output formatting. Especially when you start throwing in multiple Locales.
• Allow the standard library to do the heavy lifting for determining what a valid input value is for a given Locale. You just need to select an appropriate Locale (I chose GERMANY, but this would obviously work with others). Always use Locales when possible. Don't try to recreate formatting strings.
• Always store your input value SEPARATE from any output formatting. IE, if you want to show only three digits in the output, that's fine, but store the whole double value anyway.