Java wierdness [now all is right with the world], Something wrong with my compy maybe

Discovered this when writing a fairly basic program for university coursework, and running into bugs. Values of primitive type double are acting strange under certain conditions. Like this:

very simple code:

The seemingly inexplicable results are

Maybe a virus or something is making my Java thingies act wierd?

This post has been edited by Citrusponge on Nov 8 2005, 03:54 AM

maybe your computer can't add

is there some way to have fixed decimal places?

This post has been edited by Xgoff on Nov 4 2005, 12:02 PM

That's the nature of floating point. It's imprecise with even well behaved decimal numbers. You'll have to round.

In fact do your math in integer and divide after. Best solution.

So this is something that's meant to happen? Java sux then >:[
You sure? I can't just use integers here for my program. Some of the inputs *have* to be type double. That's not up to me. I can't convert them into integers without the compiler shouting POSSIBLE LOSS OF PRECISION at me
This especially sucks because without this problem my code would be perfect :<

*thinks of an idea which may or may not work*

well your "possible loss of precision" would only be one one-hundred-quadrillionth off anyway so it probably wouldn't matter

The compiler forces me to change it anyway :/

Doesn't matter now though, my slightly altered code works perfectly now. Thanks anyway

that's due to the round-off error, it's just a simple fact of binary that it can't hold numbers like 0.3 exactly without an infinite amount of space

some more basic programming languages will round it off transparently but java isn't one of them, so what you want to do to avoid this is check if the variable you're testing is almost exactly what you want, say in a range of about 0.0000000000000010

or you could just do what r-ii says and use integers more, but that's for pansies ;)

also while we're at it does java do fixed-point numbers too?

This post has been edited by Quicky on Nov 5 2005, 01:46 PM

Java has two types for real numbers, 'double' and 'float'.
My book says float is 'single-precision floating point' e.g. 43.889F
And double is 'double-precision floating point' e.g. 45.63 or 2.4e5

I have no idea what that means

Except the F after the float specifies that it is indeed of type float.

I don't like floating points.

<<edit>>
For anyone who cares:

Float is even worse.

Similar, should-be-predictable code:

prints this:

This post has been edited by Citrusponge on Nov 5 2005, 02:02 PM

real numbers in java are represented with the exponent-coefficient (real men call it the mantissa ;) model, which is explained right here

floats are 32-bit numbers with a 24-bit mantissa and an 8-bit exponent, doubles are 64-bit with a 53-bit mantissa and 11-bit exponent, so in short doubles give you much more precision and a wider range of numbers

man i'm on a roll

This post has been edited by Quicky on Nov 5 2005, 02:01 PM

Ooh, we learnt about mantissa and exponent and shi in college. We didn't learn that these silly problems come up :|

Yay Wikipedia

Take assembler-level programming and you'll learn real men use integers whenever possible. They're many times faster than floating point arithmetic.

And you can do fixed-point decimal in integer as well.

The only good way to escape round off error without always needing to round is to stay away from floating point arithmentic. Calculators use BCD arithmetic instead, which has no round-off error. But that's another beast I won't get into.

Hey d00ds, I'm posting this from uni \o/

More Java funfunfun:

I wrote this code to do the same thing as that other code, but hopefully without the wierd precision errors. I used a similar technique to make my assignment code work, and it worked just fine. But this code... does not work at all. It just makes the runtime environment act like it's trying to work until I kill it. I can't see any problems with it, and the compiler has no complaints. Anyone see anything bad about this?

I don't see anything wrong with it. You're sure that's the code causing the environment to hang?

I'm pretty sure. The same thing happens on both the computer I was using at university and this home one here. I copy-pasted the code just now :/

Changing certain statements lets the code run as expected (but not as I wanted)... I'll experiment more with this.

Semicolons don't go after While () statements.
If that's a direct copy and paste from your program and not a copy error.

If that did in fact run like that, that would EASILY cause your infinite loop of do-nothing.

OOhhh silly me

woohoo it counted right too \o/

This post has been edited by Citrusponge on Nov 8 2005, 03:53 AM