Floating Point things work.....Most of the time

Question

Possible Duplicate:
Correcting floating point numbers

float randNum = arc4random()%(100)-1;

I read somewhere that this will give me random numbers between 1-100. Or something close to that.

This line seems to work all the time, but I convert this number to an NSString for storage/printing to text, and then convert it back to a float for ordering and other calculations. But when I get that float back sometimes it looks something like gibberish; like this in the variable view:

9   float   9.75303731e-41

Why would converting to an NSString and back to a float ruin the float value I'm using? (e.g. What could I have screwed up? Or should I be using CGFloat instead?)

I think this is all the pertinent code:

NSMutableArray *stringArray = [[NSMutableArray alloc] init];
floatArray[30];
// put three random floats into an NSMutableArray of NSStrings
for(int i = 0; i < 3; i++)
{
    float randNum = arc4random()%(100)-1;
    NSString *randString = [NSString stringWithFormat:@"%.3f", randNum];
    [stringArray addObject:randString];
}

// convert NSStrings back to float
for(NSString *string in stringArray)
{
    float temp = [string floatValue];
    floatArray[iterator] = temp;
}

Thanks in advance for any help/advice.

EDIT: When I step through the code it looks like the float value looks sane until the line "float temp = [string floatValue]", which is where the value seems to be garbaged.

Instead of converting to NSString, why dont you convert to NSNumber and save it in array and then get it back from NSNumber object? — iDev, Nov 20 '12 at 17:56
ACB - Thanks for the suggestion. I will try that out. Inafziger - Thanks for posting the possible duplicate. I will try out ACB's suggestion. — user990769, Nov 20 '12 at 17:59
Why are you subtracting 1? `arc4random()%(100)` returns a number from 0-99, so you should _add_ 1 to shift up to 1-100. — Barmar, Nov 20 '12 at 18:04

score 1 · Answer 1 · answered Nov 20 '12 at 18:00

Why are you using float when the result from arc4random() is a uint32_t? Switching to integer types would almost certainly get around all this, as I suspect the problem is because of the conversion to string form allowing only 3 significant digits. What happens if you use %.15f as your format?

Sidenote: use arc4random_uniform() - it's simpler and guaranteed to be a uniformly random distribution within that range.

Floating Point things work.....Most of the time

1 Answers1