my brain isn't working today folks, need help with a math formula

[KodeShark]

Alrighty, I am trying to:
I have a float data type, Misc, that starts at a predetermined value...in this case 0.0.

My program loops it in increments of 0.1 until 2.5 is met, at which point it starts over at 0.0. How can I have it count backwards in increments of 0.1 during execution when it reaches 2.5? (So I can get a swing effect instead of start to finish~

static int Counter;

FLOAT fAngle = Misc	+= 0.1f;

	while (Misc == 0){
		for ( Counter = 1; Counter <= 25; Counter++ ){
			Misc = Misc + 0.1;
		}
	}
	
	while (Misc == 25){
		for ( int Counter = 25; Counter >= 1; Counter-- ){
			Misc = Misc - 0.1;
		}
	}

This isn't working. There is a flaw somewhere I guess. Anyways, help me if you can.:afrog:

[PT Exorcist]

if (fAngle == 2.5) {
Angle -= 0.1;
}

??

[KodeShark]

that would put it at 2.4~

[PT Exorcist]

you have 25 instead of 2.5.
is that it?

[KodeShark]

nah, i did that for the "for" loop. I didn't want to use a float variable...i usually use int for loops (so i had 1.0 instead of .1)

[NickMeuir]

Off the top of my head:

float swing = 45.0;
bool left = false;

if(left)
{
	swing += 1.0f;
	if(swing >= 45.0f)
	{
		left = true;
	}
}
else
{
	swing -= 1.0f;
	if(swing <= -45.0f)
	{
		left = false;
	}
}

Something like that.

[SLH]

The standard game programming method would be to give it a rate of change.

Start it and +vb 0.1, then when it reaches 2.5 change it to -ve 0.1
Then juse increnemt the angle by that each time.

[Halsafar]

A basic example of what SLH explained I believe.
Incorperate that idea into a for loop how you like.

int delim = 0.1;
int num=0;

num =+ delim;

if (num == 2.5) delim = -0.1;
if (num == 0) delim = 0.1; 

[KodeShark]

Thanks guys. I guess this thread got taken out of my "to watch" list. Ill go work with it now~ =)

[CornedBee]

Avoid == comparisons with floating point numbers, they might unexpectedly fail, perhaps just when porting from one platform to another. Always use >, >= etc.

[Halsafar]

Really?
== can fail...I guess putting a hint of thought to the subject I can see exactly how it may fail.

floats already have a loss of accuracy...I suppose some platforms may signify your float 2.5 to 3 when it see's == cuz it expects a whole or integeral number.

I'll get out of that habit myself.

[CornedBee]

No, conversion to integral is not the problem. The problem is different.
Floating point numbers can represent very small numbers with high accuracy or very large numbers with low accuracy, with a linear curve between the two extremes (average numbers with average accuracy, that is). That is the nature of the floating point.

Now, let's say that you're incrementing from 0.0 to 10.0 in steps of 0.1, ending the increment by testing == 10.0. 10.0 could be exactly representable by a float, so you think there's nothing that can go wrong, right? Wrong. For one, this is not portable. For another, perhaps 9.9 is not representable, and you get 9.899123921483 instead. The next increment takes you to 9.999123921483. Again, this is not representable and gets rounded to 10.0. Lucky one. The increment becomes a decrement. Taking 0.1 away from 10.0 yields 9.9 again, so you get 9.899123921483. Taking 0.1 away again gets you 9.799123921483. Because there's higher accuracy at smaller numbers, this stays representable. And so does every such number down to 0.099123921483 (instead of 0.1) and -0,000876078517 (instead of 0.0). Because your lower test was == 0.0, this is not true. You're past the limit, the loop becomes infinite and unbounded. Very bad.

[SLH]

What data type would you suggest as an alternative?

I need to represent +ve and -ve numbers which would be coordinates of objects. I'd also need to do for loops and comparisons and such, as in your post above.

At the moment i'm using double, and checking things like >= and <=, but sometimes i'd like to check for equality.

[CornedBee]

Usually you don't need another datatype. Just don't use ==.

If you really need known precision, fixed point math or an arbitrary precision math library comes to mind.

[SLH]

Ok, so i could just make a function like this:

bool FloatEquals(float f1, float f2, float tollerence)
{
    return (f1>=f2-tollerence) && (f1<=f2+tollerence)
}

[NoteMe]

Originally posted by SLH
Ok, so i could just make a function like this:

bool FloatEquals(float f1, float f2, float tollerence)
{
    return (f1>=f2-tollerence) && (f1<=f2+tollerence)
}

Yeah thats is more or less the way I do it....here is an example:

inline bool deltaRangeTest(T a, T b, T epsilon)
{
    return (absoluteValue(a - b) < epsilon) ? true : false;
}

[Arc]

What if you used fixed and setprecision(1) to force the decaimal places to stop at 1? Would it still then be inacurate?

[CornedBee]

fixed and setprecision only affect the way the numbers are converted to strings on output, they only have meaning when you're writing the numbers. What is saved in the variable is not affected.

[SLH]

Originally posted by NoteMe
Yeah thats is more or less the way I do it....here is an example:

inline bool deltaRangeTest(T a, T b, T epsilon)
{
    return (absoluteValue(a - b) < epsilon) ? true : false;
}

I guess this would be equivilent (and faster)

inline bool deltaRangeTest(T a, T b, T epsilon)
{
    return (absoluteValue(a - b) < epsilon);
}