CODEWITHSUNDEEP
javapow
DJ Quimby
DJ Quimby

Reputation: 3699

Java's Math.Pow() function returning confusing results

I am working with the Math.pow() function, and have the following code:

double monthlyRate = (0.7d / 12);
int loanLength = 3;

double powerTest = Math.pow(1.00583, 36);
double powerResult = Math.pow((1 + monthlyRate),(loanLength * 12));

When this is run through the debugger, the values become

powerTest => 1.2327785029794363
powerResult => 7.698552870922063

The first is the correct one. I've stepped into the Math.pow function on both of the assignment lines. For powerTest, the parameters for Math.pow are double a => 1.00583 double b => 36.0

For powerResult, they are double a => 1.0058333333333333 double b => 36.0

I know that this is an issue with the way floating point math is performed by the machine, I'm just not sure how to correct it. I tried doing the following before the calculation with poor results:

monthlyRate = Math.round(monthlyRate * 1000) / 1000;

Upvotes: 0

Views: 4460

Answers (7)

vz0
vz0

Reputation: 32923

On Java you can use BigDecimal to perform money operations resulting on accurate numbers: it is the Sun/Oracle recommended way to store money numbers.

// I'm using Strings for most accuracy

BigDecimal monthlyRate = new BigDecimal("0.7").divide(new BigDecimal(12));
int loanLength = 3;

BigDecimal powerTest = new BigDecimal("1.00583").pow(36);
BigDecimal powerResult = BigDecimal.ONE.add(monthlyRate).pow(loanLength * 12);

Upvotes: 1

fmucar
fmucar

Reputation: 14548

monthlyRate = ((double)Math.round(monthlyRate * 1000)) / 1000;

Upvotes: 0

mb14
mb14

Reputation: 22596

Obviously a such big difference in the resut(1.23... and 7.70) is not related to the way floats are coded but more than you made a mistake somewhere 1+0.7/12 = 1.0583 is different from 1.00583 ;-).

Upvotes: 2

Manidip Sengupta
Manidip Sengupta

Reputation: 3611

Your expression 0.7d/12 = 0.0583, in the powerTest expression you are using 0.00583.

Upvotes: 3

Mark Elliot
Mark Elliot

Reputation: 77024

I think part of the problem is that 0.7/12 ~ 0.058333, and 1.0583 > 1.00583. My bet is this is the true source of your discrepancy, the floating point adjustments have little to do with it.

Upvotes: 2

axtavt
axtavt

Reputation: 242686

1 + monthlyRate is 1.0583..., not 1.00583.

Upvotes: 9

Nikita Rybak
Nikita Rybak

Reputation: 67986

Math.round(monthlyRate * 1000) / 1000.0;

You were using integer division.

Upvotes: 1

Related Questions