Unfortunately, none of the previous answers (including the accepted one) works for all possible inputs.
1) sprintf('%1.'.$precision.'f', $val)
Fails with a precision of 2 : 14.239 should return 14.23 (but in this case returns 14.24).
2) floatval(substr($val, 0, strpos($val, '.') + $precision + 1))
Fails with a precision of 0 : 14 should return 14 (but in this case returns 1)
3) substr($val, 0, strrpos($val, '.', 0) + (1 + $precision))
Fails with a precision of 0 : -1 should return -1 (but in this case returns '-')
4) floor($val * pow(10, $precision)) / pow(10, $precision)
Although I used this one extensively, I recently discovered a flaw in it ; it fails for some values too. With a precision of 2 : 2.05 should return 2.05 (but in this case returns 2.04 !!)
So far the only way to pass all my tests is unfortunately to use string manipulation. My solution based on rationalboss one, is :
function floorDec($val, $precision = 2) {
if ($precision < 0) { $precision = 0; }
$numPointPosition = intval(strpos($val, '.'));
if ($numPointPosition === 0) { //$val is an integer
return $val;
}
return floatval(substr($val, 0, $numPointPosition + $precision + 1));
}
This function works with positive and negative numbers, as well as any precision needed.