round scientific notation values?

Joseph_Sliker

Hi;

I am running a script that gets me the slope and intercept for an ordered set of values (simple linear regression).

The output of these calculations pretty much always comes back as scientific notation (ie. "-1.2550405687496E-08"). I have been trying to use round() or number_format() directly on these values to produce a narrower display column but apparently this does not work because the number is a string that contains non-numeric characters.

Has anyone else encountered this problem and figured out how to round such values?

Thanks.

laserlight

Perhaps you can make use of [man]sprintf/man by specifying the precision.

Joseph_Sliker

I can't seem to comprehend what is happening with sprintf() and I just end up getting a bunch of "%" and zeros.

I've come up with a temporary solution that may be adequate...even if it's not right.

$tempval=floatval($m); // the offending string...typically in scientific notation
$chopped=substr($tempval,0,8); //hack to get past scientific notation. (the string is typically about 17 characters long....way more precise than needed for the display.
$fm=number_format(round($chopped, 3),3);  // rounded/formatted number

I'll keep trying to do it "correctly" but I needed something in place for end of schoolyear reporting.

laserlight

If you are trying to reduce it to 3 decimal places, then write:

$fm = sprintf('&#37;.3E', $m);

Weedpacket

Joseph Sliker wrote:
apparently this does not work because the number is a string that contains non-numeric characters.

It shouldn't be - floating-point numbers are stored as numbers until some string operation needs them to be coerced into being a string; none of the mathematical functions do that except those that do base conversion ([man]number_format[/man] on the other hand is a string function).

As laserlight's suggestion hints, adjusting the precision of the value ought to be limited to the display of results (so that it doesn't appear to have higher precision than the original data); doing it at display time also avoids the problem that floating-point numbers aren't completely accurate real numbers - on my machine [font=monospace]echo 0.1;[/font] outputs [font=monospace]0.100000000000000005551115[/font], and if I [font=monospace]echo round(0.1, 1);[/font] it still outputs [font=monospace]0.100000000000000005551115[/font], simply because that's the closest it can come to storing one-tenth as a floating-point number.