Win a copy of Functional Reactive Programming this week in the Other Languages forum!

# float vs double equality

Omkar chary
Greenhorn
Posts: 11
I am preparing for my Job.
Going to write OCJP 6 certification.
Here i got a weird problem.
I am comparing floating point and double literals.
In SCJP kathy seirra book, kathy mentioned: a float value never equals to double value, in comparison always we get false.

Winston Gutkowski
Bartender
Posts: 10527
64
Omkar onky wrote:In SCJP kathy seirra book, kathy mentioned: a float value never equals to double value, in comparison always we get false.

I think you might want to double-check what she actually said in the book, because if she did, then it's plainly wrong - as you have conclusively proved. In most comparisons involving numeric types of different sizes, the smaller one will be converted to the larger before the comparison takes place, and since floating-point numbers are notorious for not holding exact values - particularly for fractions involving 10ths - the double version will be different from a float expanded to a double.

I suspect that if you try:
System.out.println( 0.75 == 0.75f );
and
System.out.println( 0.0625 == 0.0625f );
you'll find that they print out true as well.

I suggest you read this. It may well help to explain some of this stuff.

Winston

Omkar chary
Greenhorn
Posts: 11
I found these comparisons also giving true!
System.out.println(2.25f==2.25);
System.out.println(2.75f==2.75);
System.out.println(2.125f==2.125);
System.out.println(2.625f==2.625);

.25, 75, 125, 625 are the factors of 5. giving true.
But 175 is also a factor of 5.
System.out.println(2.175f == 2.175);//false

what i am trying to ask you is, is there any way to
identify?

Winston Gutkowski
Bartender
Posts: 10527
64
Omkar onky wrote:what i am trying to ask you is, is there any way to identify?

Yes, but it's tricky, and I'm not sure that that's what you actually need to do.

And BTW, those values that give true are not factors of 5 but fractions involving some power of 2. Integer amounts (within certain limits) will also compare equal.

What is more likely to work is to cast the larger type (the double) to the smaller (float), viz:
System.out.println(0.1f == (float) 0.1);

I suspect strongly that this is not guaranteed to work because of the inexact nature of FP values and vagaries of rounding; although you may have trouble finding values for which it doesn't. For more details, you need to check out the relevant chapter in the JLS.

HIH

Winston

Omkar chary
Greenhorn
Posts: 11
According to your knowledge is there any
trick to identify which type of values direct comparison gives true?
I am just eager to know...Mr.Winston.

Jesper de Jong
Java Cowboy
Saloon Keeper
Posts: 15484
43
Floating-point values in a float or double are stored as binary fractions, in the IEEE-754 single and double precision formats. Note that these formats cannot store any number with arbitrary precision. Only numbers that consist of combinations of (positive or negative) powers of 2 that fit in the range of the formats can be stored exactly.

It's interesting to know the exact details of how floating-point numbers are stored, but the SCJP or OCJP exam does not require you to know this into this level of detail.

There are not going to be questions on the exam that expect you to know that 2.25 == 2.25f but 2.1 != 2.1f.