ChatGPT解决这个技术问题 Extra ChatGPT

How to round a number to n decimal places in Java

What I would like is a method to convert a double to a string which rounds using the half-up method - i.e. if the decimal to be rounded is 5, it always rounds up to the next number. This is the standard method of rounding most people expect in most situations.

I also would like only significant digits to be displayed - i.e. there should not be any trailing zeroes.

I know one method of doing this is to use the String.format method:

String.format("%.5g%n", 0.912385);

returns:

0.91239

which is great, however it always displays numbers with 5 decimal places even if they are not significant:

String.format("%.5g%n", 0.912300);

returns:

0.91230

Another method is to use the DecimalFormatter:

DecimalFormat df = new DecimalFormat("#.#####");
df.format(0.912385);

returns:

0.91238

However as you can see this uses half-even rounding. That is it will round down if the previous digit is even. What I'd like is this:

0.912385 -> 0.91239
0.912300 -> 0.9123

What is the best way to achieve this in Java?


f
fishinear

Use setRoundingMode, set the RoundingMode explicitly to handle your issue with the half-even round, then use the format pattern for your required output.

Example:

DecimalFormat df = new DecimalFormat("#.####");
df.setRoundingMode(RoundingMode.CEILING);
for (Number n : Arrays.asList(12, 123.12345, 0.23, 0.1, 2341234.212431324)) {
    Double d = n.doubleValue();
    System.out.println(df.format(d));
}

gives the output:

12
123.1235
0.23
0.1
2341234.2125

EDIT: The original answer does not address the accuracy of the double values. That is fine if you don't care much whether it rounds up or down. But if you want accurate rounding, then you need to take the expected accuracy of the values into account. Floating point values have a binary representation internally. That means that a value like 2.7735 does not actually have that exact value internally. It can be slightly larger or slightly smaller. If the internal value is slightly smaller, then it will not round up to 2.7740. To remedy that situation, you need to be aware of the accuracy of the values that you are working with, and add or subtract that value before rounding. For example, when you know that your values are accurate up to 6 digits, then to round half-way values up, add that accuracy to the value:

Double d = n.doubleValue() + 1e-6;

To round down, subtract the accuracy.


This is probably the best solution presented so far. The reason I didn't spot this facility when I first looked at the DecimalFormat class is that it was only introduced in Java 1.6. Unfortunately I'm restricted to using 1.5 but it will be useful to know for the future.
I tried this with: "#.##", rounding HALF_UP. 256.335f -> "256.33" ...(example comes from comments to @asterite's answer).
Please be carefull as DecimalFormat depend on your current Local configuration, you may not get a dot as a separator. I personnally prefer Asterite's answer below
Also be aware that you should not expect DecimalFormat to be thread-safe. As per Java docs: Decimal formats are generally not synchronized. It is recommended to create separate format instances for each thread. If multiple threads access a format concurrently, it must be synchronized externally.
how do i make it so that it does a proper rounding so it will not round 0.0004 to 0.001
c
confile

Assuming value is a double, you can do:

(double)Math.round(value * 100000d) / 100000d

That's for 5 digits precision. The number of zeros indicate the number of decimals.


UPDATE: I just confirmed that doing this IS WAY faster than using DecimalFormat. I looped using DecimalFormat 200 times, and this method. DecimalFormat took 14ms to complete the 200 loops, this method took less than 1ms. As I suspected, this is faster. If you get paid by the clock cycle, this is what you should be doing. I'm surprised Chris Cudmore would even say what he said to be honest. allocating objects is always more expensive than casting primitives and using static methods (Math.round() as opposed to decimalFormat.format()).
This technique fails in over 90% of cases. -1.
Indeed, this fails: Math.round(0.1 * Math.pow(10,20))/Math.pow(10,20) == 0.09223372036854775.
Be very careful when using this method (or any rounding of floating points). It fails for something as simple as 265.335. The intermediate result of 265.335 * 100 (precision of 2 digits) is 26533.499999999996. This means it gets rounded down to 265.33. There simply are inherent problems when converting from floating point numbers to real decimal numbers. See EJP's answer here at stackoverflow.com/a/12684082/144578
@SebastiaanvandenBroek: Wow I never knew it was that easy to get a wrong answer. However, if one is working with non-exact numbers, one must recognize that any value is not exact. 265.335 really means 265.335 += tolerance, where tolerance depends on previous operations and range of input values. We do not know the true, exact value. At the edge values, either answer is arguably correct. If we need to be exact, we shouldn't work in double. The fail here isn't in converting back to double. Its in OP thinking he can rely on the incoming 265.335 as being exactly that.
N
Nav
new BigDecimal(String.valueOf(double)).setScale(yourScale, BigDecimal.ROUND_HALF_UP);

will get you a BigDecimal. To get the string out of it, just call that BigDecimal's toString method, or the toPlainString method for Java 5+ for a plain format string.

Sample program:

package trials;
import java.math.BigDecimal;

public class Trials {

    public static void main(String[] args) {
        int yourScale = 10;
        System.out.println(BigDecimal.valueOf(0.42344534534553453453-0.42324534524553453453).setScale(yourScale, BigDecimal.ROUND_HALF_UP));
    }

That's my preferred solution. Even shorter: BigDecimal.valueOf(doubleVar).setScale(yourScaleHere, BigDecimal.ROUND_HALF_UP); BigDecimal.valueOf(double val) actually calls Double.toString() under the hood ;)
Nice. Don't cut corners and use new BigDecimal(doubleVar) as you can run into issues with rounding of floating points
@Edd, interestingly, the rounding issue occurs in the case SebastiaanvandenBroek mentions in comment to asterite's answer. double val = 265.335;, BigDecimal.valueOf(val).setScale(decimals, BigDecimal.ROUND_HALF_UP).toPlainString(); => 265.34, but (new BigDecimal(val)).setScale(decimals, BigDecimal.ROUND_HALF_UP).toPlainString(); => 265.33.
@ToolmakerSteve That's because using new BigDecimal with the double takes the double value directly and attempts to use that to create the BigDecimal, whereas when using BigDecimal.valueOf or the tostring form parses it to a string first (a more exact representation) before the conversion.
BigDecimal.ROUND_HALF_UP is deprecated since 9. You can use: RoundingMode.HALF_UP instead.
M
Milhous

You can also use the

DecimalFormat df = new DecimalFormat("#.00000");
df.format(0.912385);

to make sure you have the trailing 0's.


I believe one of the goals of the question was that "there should not be any trailing zeroes".
For this question, the op didn't want zeros, but this is exactly what I wanted. If you have a list of numbers with 3 decimal places, you want them to all have the same digits even if it's 0.
You forgot to specify RoundingMode.
@IgorGanapolsky by default Decimal mode uses RoundingMode.HALF_EVEN.
u
user207421

As some others have noted, the correct answer is to use either DecimalFormat or BigDecimal. Floating-point doesn't have decimal places so you cannot possibly round/truncate to a specific number of them in the first place. You have to work in a decimal radix, and that is what those two classes do.

I am posting the following code as a counter-example to all the answers in this thread and indeed all over StackOverflow (and elsewhere) that recommend multiplication followed by truncation followed by division. It is incumbent on advocates of this technique to explain why the following code produces the wrong output in over 92% of cases.

public class RoundingCounterExample
{

    static float roundOff(float x, int position)
    {
        float a = x;
        double temp = Math.pow(10.0, position);
        a *= temp;
        a = Math.round(a);
        return (a / (float)temp);
    }

    public static void main(String[] args)
    {
        float a = roundOff(0.0009434f,3);
        System.out.println("a="+a+" (a % .001)="+(a % 0.001));
        int count = 0, errors = 0;
        for (double x = 0.0; x < 1; x += 0.0001)
        {
            count++;
            double d = x;
            int scale = 2;
            double factor = Math.pow(10, scale);
            d = Math.round(d * factor) / factor;
            if ((d % 0.01) != 0.0)
            {
                System.out.println(d + " " + (d % 0.01));
                errors++;
            }
        }
        System.out.println(count + " trials " + errors + " errors");
    }
}

Output of this program:

10001 trials 9251 errors

EDIT: To address some comments below I redid the modulus part of the test loop using BigDecimal and new MathContext(16) for the modulus operation as follows:

public static void main(String[] args)
{
    int count = 0, errors = 0;
    int scale = 2;
    double factor = Math.pow(10, scale);
    MathContext mc = new MathContext(16, RoundingMode.DOWN);
    for (double x = 0.0; x < 1; x += 0.0001)
    {
        count++;
        double d = x;
        d = Math.round(d * factor) / factor;
        BigDecimal bd = new BigDecimal(d, mc);
        bd = bd.remainder(new BigDecimal("0.01"), mc);
        if (bd.multiply(BigDecimal.valueOf(100)).remainder(BigDecimal.ONE, mc).compareTo(BigDecimal.ZERO) != 0)
        {
            System.out.println(d + " " + bd);
            errors++;
        }
    }
    System.out.println(count + " trials " + errors + " errors");
}

Result:

10001 trials 4401 errors

The trick is that in all of your 9251 errors, the printed result is still correct.
@DidierL It doesn't surprise me. I had the very good fortune of doing 'Numerical Methods' as my very first computing course and being introduced right at the start to what floating-point can and cannot do. Most programmers are pretty vague about it.
All you are doing is refuting that floating doesn't represent many decimal values exactly, which I would hope we all understand. Not that rounding does causes a problem. As you admit, the numbers still print as expected.
Your test is broken, take round() out and the test fails 94% of the time. ideone.com/1y62CY prints 100 trials 94 errors You should start with a test which passes, and show that introducing rounding breaks the test.
Refutation, refuted here. Using Math.round for this range of double as no errors ideone.com/BVCHh3
B
Basil Bourque

Suppose you have

double d = 9232.129394d;

you can use BigDecimal

BigDecimal bd = new BigDecimal(d).setScale(2, RoundingMode.HALF_EVEN);
d = bd.doubleValue();

or without BigDecimal

d = Math.round(d*100)/100.0d;

with both solutions d == 9232.13


I think this is the best solution for Java 1.5 users (and below). One comment tho, don't use the HALF_EVEN rounding mode since it has diff behavior for odd and even numbers (2.5 rounds to 2 while 5.5 rounds to 6, for example), unless this is what you want.
The first solution is correct: the second one doesn't work. See here for proof.
@EJP: Even the first solution with RoundingMode.HALF_UP is wrong. Try it with 1.505. The right way is to use BigDecimal.valueOf(d).
Matthias Braun, the solution is fine, hence 31 ups.. 1.505 decimal is stored in floating point double as 1.50499998 if you want to take 1.505 and convert from double to decimal, then you have to convert it to Double.toString(x) first then put it into a BigDecimal(), but that is extremely slow, and defeats the purpose of using double for speed in the first place.
Ran a loop of 100k with BigDecimal (took 225 ms) and Math.round (2 ms) way and here is the timing...Time Taken : 225 milli seconds to convert using to : 9232.13 Time Taken : 2 milli seconds to convert to : 9232.13 techiesinfo.com
J
JibW

You can use the DecimalFormat class.

double d = 3.76628729;

DecimalFormat newFormat = new DecimalFormat("#.##");
double twoDecimal =  Double.valueOf(newFormat.format(d));

Any reason why Double.valueOf() was chosen over Double.parseDouble()? The valueOf() method returns a Double object, while parseDouble() will return a double primitive. With the way the current code is written, you also apply auto-unboxing to the return to cast it to the primitive that your twoDouble variable expects, an extra bytecode operation. I'd change the answer to use parseDouble() instead.
Double.parseDouble() needs String input.
M
Md. Farhan Memon

Real's Java How-to posts this solution, which is also compatible for versions before Java 1.6.

BigDecimal bd = new BigDecimal(Double.toString(d));
bd = bd.setScale(decimalPlace, BigDecimal.ROUND_HALF_UP);
return bd.doubleValue();

UPDATE: BigDecimal.ROUND_HALF_UP is deprecated - Use RoundingMode

BigDecimal bd = new BigDecimal(Double.toString(number));
bd = bd.setScale(decimalPlaces, RoundingMode.HALF_UP);
return bd.doubleValue();

C
Chris Cudmore
double myNum = .912385;
int precision = 10000; //keep 4 digits
myNum= Math.floor(myNum * precision +.5)/precision;

yes this is exactly what math.round does for positive numbers, but have you tried this with negative numbers? people are using math.round in the other solutions to also cover the case of negative numbers.
Note: Math.floor(x + 0.5) and Math.round(x)
P
Paŭlo Ebermann

@Milhous: the decimal format for rounding is excellent:

You can also use the DecimalFormat df = new DecimalFormat("#.00000"); df.format(0.912385); to make sure you have the trailing 0's.

I would add that this method is very good at providing an actual numeric, rounding mechanism - not only visually, but also when processing.

Hypothetical: you have to implement a rounding mechanism into a GUI program. To alter the accuracy / precision of a result output simply change the caret format (i.e. within the brackets). So that:

DecimalFormat df = new DecimalFormat("#0.######");
df.format(0.912385);

would return as output: 0.912385

DecimalFormat df = new DecimalFormat("#0.#####");
df.format(0.912385);

would return as output: 0.91239

DecimalFormat df = new DecimalFormat("#0.####");
df.format(0.912385);

would return as output: 0.9124

[EDIT: also if the caret format is like so ("#0.############") and you enter a decimal, e.g. 3.1415926, for argument's sake, DecimalFormat does not produce any garbage (e.g. trailing zeroes) and will return: 3.1415926 .. if you're that way inclined. Granted, it's a little verbose for the liking of some dev's - but hey, it's got a low memory footprint during processing and is very easy to implement.]

So essentially, the beauty of DecimalFormat is that it simultaneously handles the string appearance - as well as the level of rounding precision set. Ergo: you get two benefits for the price of one code implementation. ;)


If you really want decimal numbers for calculation (and not only for output), do not use a binary-based floating point format like double. Use BigDecimal or any other decimal-based format.
C
Community

Here is a summary of what you can use if you want the result as String:

DecimalFormat#setRoundingMode(): DecimalFormat df = new DecimalFormat("#.#####"); df.setRoundingMode(RoundingMode.HALF_UP); String str1 = df.format(0.912385)); // 0.91239 BigDecimal#setScale() String str2 = new BigDecimal(0.912385) .setScale(5, BigDecimal.ROUND_HALF_UP) .toString();

Here is a suggestion of what libraries you can use if you want double as a result. I wouldn't recommend it for string conversion, though, as double may not be able to represent what you want exactly (see e.g. here):

Precision from Apache Commons Math double rounded = Precision.round(0.912385, 5, BigDecimal.ROUND_HALF_UP); Functions from Colt double rounded = Functions.round(0.00001).apply(0.912385) Utils from Weka double rounded = Utils.roundDouble(0.912385, 5)


A
Amit

You could use the following utility method-

public static double round(double valueToRound, int numberOfDecimalPlaces)
{
    double multipicationFactor = Math.pow(10, numberOfDecimalPlaces);
    double interestedInZeroDPs = valueToRound * multipicationFactor;
    return Math.round(interestedInZeroDPs) / multipicationFactor;
}

@mariolpantunes: It will fail. Try this: round(1.005,2); or round(0.50594724957626620092, 20);
It works. But uninformatively float and doubles are approximations. Let us consider your first example. If you print the output of interestedInZeroDPs before Math.round it will print 100.49999999999999. You lost precision as such Math.round round it as 100. Due to the nature or floats and doubles there are borderlines cases when it does not work properly (more information here en.wikipedia.org/wiki/Floating_point#Accuracy_problems)
double is a fast! decimal is slow. computers don't bother processing their thinking in decimal notation. you have to give up some decimal precision to keep floating point double fast.
@hamish The question is about precision, not about speed.
v
vitalh

A succinct solution:

   public static double round(double value, int precision) {
      int scale = (int) Math.pow(10, precision);
      return (double) (Math.round(value * scale) / scale);
  }

See also, https://stackoverflow.com/a/22186845/212950 Thanks to jpdymond for offering this.

Edit: Added round brackets. Casts the whole result to double, not the first argument only!


E
Easwaramoorthy K

You can use BigDecimal

BigDecimal value = new BigDecimal("2.3");
value = value.setScale(0, RoundingMode.UP);
BigDecimal value1 = new BigDecimal("-2.3");
value1 = value1.setScale(0, RoundingMode.UP);
System.out.println(value + "n" + value1);

Refer: http://www.javabeat.net/precise-rounding-of-decimals-using-rounding-mode-enumeration/


J
Jorgesys

To achieve this we can use this formatter:

 DecimalFormat df = new DecimalFormat("#.00");
 String resultado = df.format(valor)

or:

DecimalFormat df = new DecimalFormat("0.00"); :

Use this method to get always two decimals:

   private static String getTwoDecimals(double value){
      DecimalFormat df = new DecimalFormat("0.00"); 
      return df.format(value);
    }

Defining this values:

91.32
5.22
11.5
1.2
2.6

Using the method we can get this results:

91.32
5.22
11.50
1.20
2.60

demo online.


H
HarshitMadhav

If you really want decimal numbers for calculation (and not only for output), do not use a binary-based floating point format like double.

Use BigDecimal or any other decimal-based format.

I do use BigDecimal for calculations, but bear in mind it is dependent on the size of numbers you're dealing with. In most of my implementations, I find parsing from double or integer to Long is sufficient enough for very large number calculations.

In fact, I've recently used parsed-to-Long to get accurate representations (as opposed to hex results) in a GUI for numbers as big as ################################# characters (as an example).


L
Li Ying

Try this: org.apache.commons.math3.util.Precision.round(double x, int scale)

See: http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/util/Precision.html

Apache Commons Mathematics Library homepage is: http://commons.apache.org/proper/commons-math/index.html

The internal implemetation of this method is:

public static double round(double x, int scale) {
    return round(x, scale, BigDecimal.ROUND_HALF_UP);
}

public static double round(double x, int scale, int roundingMethod) {
    try {
        return (new BigDecimal
               (Double.toString(x))
               .setScale(scale, roundingMethod))
               .doubleValue();
    } catch (NumberFormatException ex) {
        if (Double.isInfinite(x)) {
            return x;
        } else {
            return Double.NaN;
        }
    }
}

佚名

Since I found no complete answer on this theme I've put together a class that should handle this properly, with support for:

Formatting: Easily format a double to string with a certain number of decimal places

Parsing: Parse the formatted value back to double

Locale: Format and parse using the default locale

Exponential notation: Start using exponential notation after a certain threshold

Usage is pretty simple:

(For the sake of this example I am using a custom locale)

public static final int DECIMAL_PLACES = 2;

NumberFormatter formatter = new NumberFormatter(DECIMAL_PLACES);

String value = formatter.format(9.319); // "9,32"
String value2 = formatter.format(0.0000005); // "5,00E-7"
String value3 = formatter.format(1324134123); // "1,32E9"

double parsedValue1 = formatter.parse("0,4E-2", 0); // 0.004
double parsedValue2 = formatter.parse("0,002", 0); // 0.002
double parsedValue3 = formatter.parse("3423,12345", 0); // 3423.12345

Here is the class:

import java.math.RoundingMode;
import java.text.DecimalFormat;
import java.text.DecimalFormatSymbols;
import java.text.ParseException;
import java.util.Locale;

public class NumberFormatter {

    private static final String SYMBOL_INFINITE           = "\u221e";
    private static final char   SYMBOL_MINUS              = '-';
    private static final char   SYMBOL_ZERO               = '0';
    private static final int    DECIMAL_LEADING_GROUPS    = 10;
    private static final int    EXPONENTIAL_INT_THRESHOLD = 1000000000; // After this value switch to exponential notation
    private static final double EXPONENTIAL_DEC_THRESHOLD = 0.0001; // Below this value switch to exponential notation

    private DecimalFormat decimalFormat;
    private DecimalFormat decimalFormatLong;
    private DecimalFormat exponentialFormat;

    private char groupSeparator;

    public NumberFormatter(int decimalPlaces) {
        configureDecimalPlaces(decimalPlaces);
    }

    public void configureDecimalPlaces(int decimalPlaces) {
        if (decimalPlaces <= 0) {
            throw new IllegalArgumentException("Invalid decimal places");
        }

        DecimalFormatSymbols separators = new DecimalFormatSymbols(Locale.getDefault());
        separators.setMinusSign(SYMBOL_MINUS);
        separators.setZeroDigit(SYMBOL_ZERO);

        groupSeparator = separators.getGroupingSeparator();

        StringBuilder decimal = new StringBuilder();
        StringBuilder exponential = new StringBuilder("0.");

        for (int i = 0; i < DECIMAL_LEADING_GROUPS; i++) {
            decimal.append("###").append(i == DECIMAL_LEADING_GROUPS - 1 ? "." : ",");
        }

        for (int i = 0; i < decimalPlaces; i++) {
            decimal.append("#");
            exponential.append("0");
        }

        exponential.append("E0");

        decimalFormat = new DecimalFormat(decimal.toString(), separators);
        decimalFormatLong = new DecimalFormat(decimal.append("####").toString(), separators);
        exponentialFormat = new DecimalFormat(exponential.toString(), separators);

        decimalFormat.setRoundingMode(RoundingMode.HALF_UP);
        decimalFormatLong.setRoundingMode(RoundingMode.HALF_UP);
        exponentialFormat.setRoundingMode(RoundingMode.HALF_UP);
    }

    public String format(double value) {
        String result;
        if (Double.isNaN(value)) {
            result = "";
        } else if (Double.isInfinite(value)) {
            result = String.valueOf(SYMBOL_INFINITE);
        } else {
            double absValue = Math.abs(value);
            if (absValue >= 1) {
                if (absValue >= EXPONENTIAL_INT_THRESHOLD) {
                    value = Math.floor(value);
                    result = exponentialFormat.format(value);
                } else {
                    result = decimalFormat.format(value);
                }
            } else if (absValue < 1 && absValue > 0) {
                if (absValue >= EXPONENTIAL_DEC_THRESHOLD) {
                    result = decimalFormat.format(value);
                    if (result.equalsIgnoreCase("0")) {
                        result = decimalFormatLong.format(value);
                    }
                } else {
                    result = exponentialFormat.format(value);
                }
            } else {
                result = "0";
            }
        }
        return result;
    }

    public String formatWithoutGroupSeparators(double value) {
        return removeGroupSeparators(format(value));
    }

    public double parse(String value, double defValue) {
        try {
            return decimalFormat.parse(value).doubleValue();
        } catch (ParseException e) {
            e.printStackTrace();
        }
        return defValue;
    }

    private String removeGroupSeparators(String number) {
        return number.replace(String.valueOf(groupSeparator), "");
    }

}

a
ashr

Just in case someone still needs help with this. This solution works perfectly for me.

private String withNoTrailingZeros(final double value, final int nrOfDecimals) {
return new BigDecimal(String.valueOf(value)).setScale(nrOfDecimals,  BigDecimal.ROUND_HALF_UP).stripTrailingZeros().toPlainString();

}

returns a String with the desired output.


Please include your reason for downvoting in the comment, otherwise that's what we call intimidation.
m
marco

I agree with the chosen answer to use DecimalFormat --- or alternatively BigDecimal.

Please read Update below first!

However if you do want to round the double value and get a double value result, you can use org.apache.commons.math3.util.Precision.round(..) as mentioned above. The implementation uses BigDecimal, is slow and creates garbage.

A similar but fast and garbage-free method is provided by the DoubleRounder utility in the decimal4j library:

 double a = DoubleRounder.round(2.0/3.0, 3);
 double b = DoubleRounder.round(2.0/3.0, 3, RoundingMode.DOWN);
 double c = DoubleRounder.round(1000.0d, 17);
 double d = DoubleRounder.round(90080070060.1d, 9);
 System.out.println(a);
 System.out.println(b);
 System.out.println(c);
 System.out.println(d);

Will output

 0.667
 0.666
 1000.0
 9.00800700601E10

See https://github.com/tools4j/decimal4j/wiki/DoubleRounder-Utility

Disclosure: I am involved in the decimal4j project.

Update: As @iaforek pointed out DoubleRounder sometimes returns counterintuitive results. The reason is that it performs mathematically correct rounding. For instance DoubleRounder.round(256.025d, 2) will be rounded down to 256.02 because the double value represented as 256.025d is somewhat smaller than the rational value 256.025 and hence will be rounded down.

Notes:

This behaviour is very similar to that of the BigDecimal(double) constructor (but not to valueOf(double) which uses the string constructor).

The problem can be circumvented with a double rounding step to a higher precision first, but it is complicated and I am not going into the details here

For those reasons and everything mentioned above in this post I cannot recommend to use DoubleRounder.


Do you have metrics showing how efficient is your solution compared to the other ones?
I haven't compared it with other solutions but there is a jmh benchmark available in the source code: github.com/tools4j/decimal4j/blob/master/src/jmh/java/org/… I did run the benchmark on a VM, the results are available as csv file here: github.com/tools4j/decimal4j/wiki/Performance
DoubleRounder fails for following cases: DoubleRounder.round(256.025d, 2) - expected: 256.03, actual: 256.02 or for DoubleRounder.round(260.775d, 2) - expected: 260.78, actual: 260.77.
@iaforek: this is correct, because DoubleRounder performs mathematically correct rounding. However I admit that this is somewhat counterintuitive and will hence update my answer accordingly.
A
Alain Cruz

So after reading most of the answers, I realized most of them won't be precise, in fact using BigDecimal seems like the best choice, but if you don't understand how the RoundingMode works, you will inevitable lose precision. I figured this out when working with big numbers in a project and thought it could help others having trouble rounding numbers. For example.

BigDecimal bd = new BigDecimal("1363.2749");
bd = bd.setScale(2, RoundingMode.HALF_UP);
System.out.println(bd.doubleValue());

You would expect to get 1363.28 as an output, but you will end up with 1363.27, which is not expected, if you don't know what the RoundingMode is doing. So looking into the Oracle Docs, you will find the following description for RoundingMode.HALF_UP.

Rounding mode to round towards "nearest neighbor" unless both neighbors are equidistant, in which case round up.

So knowing this, we realized that we won't be getting an exact rounding, unless we want to round towards nearest neighbor. So, to accomplish an adequate round, we would need to loop from the n-1 decimal towards the desired decimals digits. For example.

private double round(double value, int places) throws IllegalArgumentException {

    if (places < 0) throw new IllegalArgumentException();

    // Cast the number to a String and then separate the decimals.
    String stringValue = Double.toString(value);
    String decimals = stringValue.split("\\.")[1];

    // Round all the way to the desired number.
    BigDecimal bd = new BigDecimal(stringValue);
    for (int i = decimals.length()-1; i >= places; i--) {
        bd = bd.setScale(i, RoundingMode.HALF_UP);
    }

    return bd.doubleValue();
}

This will end up giving us the expected output, which would be 1363.28.


S
Suragch

I came here just wanting a simple answer on how to round a number. This is a supplemental answer to provide that.

How to round a number in Java

The most common case is to use Math.round().

Math.round(3.7) // 4

Numbers are rounded to the nearest whole number. A .5 value is rounded up. If you need different rounding behavior than that, you can use one of the other Math functions. See the comparison below.

round

As stated above, this rounds to the nearest whole number. .5 decimals round up. This method returns an int.

Math.round(3.0); // 3
Math.round(3.1); // 3
Math.round(3.5); // 4
Math.round(3.9); // 4

Math.round(-3.0); // -3
Math.round(-3.1); // -3
Math.round(-3.5); // -3 *** careful here ***
Math.round(-3.9); // -4

ceil

Any decimal value is rounded up to the next integer. It goes to the ceiling. This method returns a double.

Math.ceil(3.0); // 3.0
Math.ceil(3.1); // 4.0
Math.ceil(3.5); // 4.0
Math.ceil(3.9); // 4.0

Math.ceil(-3.0); // -3.0
Math.ceil(-3.1); // -3.0
Math.ceil(-3.5); // -3.0
Math.ceil(-3.9); // -3.0

floor

Any decimal value is rounded down to the next integer. This method returns a double.

Math.floor(3.0); // 3.0
Math.floor(3.1); // 3.0
Math.floor(3.5); // 3.0
Math.floor(3.9); // 3.0

Math.floor(-3.0); // -3.0
Math.floor(-3.1); // -4.0
Math.floor(-3.5); // -4.0
Math.floor(-3.9); // -4.0

rint

This is similar to round in that decimal values round to the closest integer. However, unlike round, .5 values round to the even integer. This method returns a double.

Math.rint(3.0); // 3.0
Math.rint(3.1); // 3.0
Math.rint(3.5); // 4.0 ***
Math.rint(3.9); // 4.0
Math.rint(4.5); // 4.0 ***
Math.rint(5.5); // 6.0 ***

Math.rint(-3.0); // -3.0
Math.rint(-3.1); // -3.0
Math.rint(-3.5); // -4.0 ***
Math.rint(-3.9); // -4.0
Math.rint(-4.5); // -4.0 ***
Math.rint(-5.5); // -6.0 ***

you are only solving the particular case of rounding to 0 decimals. The original question is more generic.
C
Craigo

If you're using a technology that has a minimal JDK. Here's a way without any Java libs:

double scale = 100000;    
double myVal = 0.912385;
double rounded = (int)((myVal * scale) + 0.5d) / scale;

This would fail in cases where the myVal is not less that 1 and with zeros after decimal beyond scale value. Say you have myVal = 9.00000000912385; The above will return 9.0. I think we should provide a solution that works in all cases of myVal. Not specifically for the value you stated.
@user102859 In your example, 9.0 is the correct result. I don't understand how this would fail.
M
Md. Jamal Uddin

here is my answer:

double num = 4.898979485566356;
DecimalFormat df = new DecimalFormat("#.##");      
time = Double.valueOf(df.format(num));

System.out.println(num); // 4.89

A
Amr Ali

Here is a better function that rounds edge cases like 1.005 correctly.

Simply, we add the smallest possible float value (= 1 ulp; unit in the last place) to the number before rounding. This moves to the next representable value after the number, away from zero.

This is a little program to test it: ideone.com

/**
 * Round half away from zero ('commercial' rounding)
 * Uses correction to offset floating-point inaccuracies.
 * Works symmetrically for positive and negative numbers.
 */
public static double round(double num, int digits) {

    // epsilon correction
    double n = Double.longBitsToDouble(Double.doubleToLongBits(num) + 1);
    double p = Math.pow(10, digits);
    return Math.round(n * p) / p;
}

// test rounding of half
System.out.println(round(0.5, 0));   // 1
System.out.println(round(-0.5, 0));  // -1

// testing edge cases
System.out.println(round(1.005, 2));   // 1.01
System.out.println(round(2.175, 2));   // 2.18
System.out.println(round(5.015, 2));   // 5.02

System.out.println(round(-1.005, 2));  // -1.01
System.out.println(round(-2.175, 2));  // -2.18
System.out.println(round(-5.015, 2));  // -5.02

C
Chris Gerken

The code snippet below shows how to display n digits. The trick is to set variable pp to 1 followed by n zeros. In the example below, variable pp value has 5 zeros, so 5 digits will be displayed.

double pp = 10000;

double myVal = 22.268699999999967;
String needVal = "22.2687";

double i = (5.0/pp);

String format = "%10.4f";
String getVal = String.format(format,(Math.round((myVal +i)*pp)/pp)-i).trim();

D
Drew Noakes

If you're using DecimalFormat to convert double to String, it's very straightforward:

DecimalFormat formatter = new DecimalFormat("0.0##");
formatter.setRoundingMode(RoundingMode.HALF_UP);

double num = 1.234567;
return formatter.format(num);

There are several RoundingMode enum values to select from, depending upon the behaviour you require.


S
Se Song

DecimalFormat is the best ways to output, but I don't prefer it. I always do this all the time, because it return the double value. So I can use it more than just output.

Math.round(selfEvaluate*100000d.0)/100000d.0;

OR

Math.round(selfEvaluate*100000d.0)*0.00000d1;

If you need large decimal places value, you can use BigDecimal instead. Anyways .0 is important. Without it the rounding of 0.33333d5 return 0.33333 and only 9 digits are allows. The second function without .0 has problems with 0.30000 return 0.30000000000000004.


E
Enamul Haque

I have used bellow like in java 8. it is working for me

    double amount = 1000.431;        
    NumberFormat formatter = new DecimalFormat("##.00");
    String output = formatter.format(amount);
    System.out.println("output = " + output);

Output:

output = 1000.43

D
Dmitry Fisenko

the following method could be used if need double

double getRandom(int decimalPoints) {
    double a = Math.random();
    int multiplier = (int) Math.pow(10, decimalPoints);
    int b = (int) (a * multiplier);
    return b / (double) multiplier;
}

for example getRandom(2)