Adding 0.05 eventually produces garbage

[Pigmaster]

Hi,

I start of with a variable called $SPS with a value of 0.00

I run a loop adding 0.05 to $SPS and echo out the result which will produce..
0.00
0.05
0.10
0.15 etc.

However after a while the number suddenly starts coming out as as xx.4999999999999999
xx.54999999999999999

Any ideas why?

Thanks

[zabmilenko]

Looks like an obscure rounding issue in the way php handles floats...

Try using: round($SPS, 2); on each consecutive increase.

EDIT: Can you post a section of code demonstrating how you are doing it?

[Pigmaster]

I have a file called test.pfd which contains large amounts data (20 samples per second) and around 112 seconds, so roughly 2240 data units.

I want to produce a a graph, so need to generate the xaxis time which is 0.05 seconds (20 samples per second) intervals

This code produces a result like after several loops
46.1
46.15
46.2
46.25
46.3
46.35
46.4
46.449999999999
46.499999999999
46.549999999999
46.599999999999
46.649999999999
46.699999999999
46.749999999999
46.799999999999

$handle = @fopen("test.pfd", "r");
$Altitude ="";
$SPSn = 0.00;
$SPSt = "0.00";
$ImportStart = "False";
if ($handle) {
   while (!feof($handle)) {
       $buffer = trim(fgets($handle, 4096));

   if ($buffer == "**!!Data:") {
   		$ImportStart = "True";
   }
 	if ($ImportStart == "True") {
   		$Altitude .=  ",".$buffer;
   		$SPSn = $SPSn  + 0.05;
   		echo $SPSn."<br>";
   }
}
}

I have now changed the line to read $SPSn = round($SPSn, 2) + 0.05; and all is fine with the rounding. but why?

Thanks

[Weedpacket]

Hardly that obscure; it is inherent in the use of floating point arithmetic (a search of the forums for that term is likely to turn up some other threads that mention it). Indeed, there is a big notice about it in the manual on its page about the [man]float[/man] type.

Pigmaster wrote:

and all is fine with the rounding. but why?

Because you're rounding, and so truncating the errors before they accumulate too far.

A better approach is to do the loop in integers, and scale by 100 for the echo.

[zabmilenko]

I simply don't know. At the php page http://us2.php.net/float there are lots of plausible explainations about it, but nothing solid.

In a nutshell (the way I gather it), floats have to be converted to binary (of course) in memory. Depending on the number of 1s and 0s before and after the decimal point, the binary representation can be waaaaay different. Sometimes when doing math operations on these different representation, precision gets lost.

Sorry I don't have a better answer for you, but if round() does the trick I'd go for it, except with this alteration:

$SPSn = round($SPSn + 0.05, 2);

This way you round AFTER the addition, as a preventative measure of course.