PHP : Function Reference : Mathematical Functions : round PHP Examples Tutorials References

[quote]I'm sure its been done before, but here's my example of a round up function. This function allows you to specify the number of decimal places to round up to. Eg, when rounding up to 3 decimal places, this function adds 0.0005 to the original value and performs round(), or for 6 decimal places, adds 0.0000005 and performs round(), etc... Hopefully some of you will find it useful: <?php function roundup ($value, $dp) { // Offset to add to $value to cause round() to round up to nearest significant digit for '$dp' decimal places $offset = pow (10, -($dp + 1)) * 5; return round ($value + $offset, $dp); } ?> Please post if you have any comments or improvements on this :-) [/quote] This can be much more efficient: <?php function roundup ($value, $dp) { return ceil($value*pow(10, $dp))/pow(10, $dp); } function rounddown ($value, $dp) { return floor($value*pow(10, $dp))/pow(10, $dp); } ?> Much cleaner, isn't it ;)?

With regard to chafy at alo dot bg's not just previous; if the number 5.555 is displayed with sufficient precision (more than about 16 digits) it will probably turn out to have been stored as something like 5.554999999....9997, which would round down to 5.55. Floating-point numbers should ALWAYS be regarded as being a little bit fuzzy. It's a consequence of the fact that they're stored in binary but humans want them displayed in decimal: you can't write one-tenth exactly as a binary expansion any more than you can write one-seventh exactly as a decimal one. See the warning note on the floating-point type page.

Unfortunately, ranzoategui's (and my revision) round version did not work well (try to round 8.075 to 2 decimals for example). I found out the best way to achieve a good round is to add a tiny number to the value before rounding. This seems to work fine in all cases (and luckily it's faster than previous methods): <? function stdround($num, $d=0) { return round($num + 0.0001 / pow(10, $d), $d); } ?>

To round any number to a given number of significant digits, use log10 to find out its magnitude: round($n, ceil(0 - log10($n)) + $sigdigits); Or when you have to display a per-unit price which may work out to be less than a few cents/pence/yen you can use: // $exp = currency decimal places - 0 for Yen/Won, 2 for most others $dp = ceil(0 - log10($n)) + $sigdigits; $display = number_format($amount, ($exp>$dp)?$exp:$dp); This always displays at least the number of decimal places required by the currency, but more if displaying the unit price with precision requires it - eg: 'English proofreading from $0.0068 per word', 'English beer from $6.80 per pint'.

to mark at customcartons dot com dot au found your roundup very useful, the only problem is if the value passed is 0.00 as it roundsup to 0.01 which may not be desired. I modified the return statement to: return ($value>0) ? round($value+$offset,$dp) : '0.00'; regards James

This isn't a bug. Rounding of "halves" is usally done to the even number. This means that on average half of the "halves" are rounded up and half of the "halves" are rounded down. So that any "average" calculated should not be affected by the rounding.

This is a slighty shorter and faster version of ranzoategui's round function: <? function stdround($num, $d=0) { $d = pow(10, $d); return round(floor($num * $d * 10) / 10) / $d; } ?> This function behaves exactly like round(), except for the fact that it doesn't apply a so called bankers' rounding algorithm, but a more classic approach (where a 5 is always rounded up). For more info on Banker's rounding: http://en.wikipedia.org/wiki/Significance_arithmetic

There are many ways to round floating-point numbers. And many valid (!) algorithms. This is because the algorithm depends on the domain of value and sometimes on... country. For example as far as I know in my country money is rounded in the same way as in math calculations: 1.4 => 1 1.5 => 2 1.6 => 2 and 2.5 => 3 A lot of rounding algorithms (and very easy to implement in every language) can be found here: http://www.merlyn.demon.co.uk/pas-chop.htm

the same is true for round_down function round_down($value){ list($val,$dummy) = explode(".",$value); return $val; }

The round() function may indeed work properly with half-values (eg. 1.5), but this little method will give you peace of mind. Add some "fuzz" to your function with a miniscule delta value. $delta = 0.00001; $x = round($x+$delta); This is fine, unless $x has a value of 1.49999 ... if you worried about that, use this method instead: if(($x-floor($x))==0.5){ $x+=$delta; } $x = round($x); you can change your "optimistic" delta into a "pessimistic" delta by subtracting instead of adding. Cheers, Ian Ring

the result of this function always depends on the underlying C function. There have been a lot of compiler bugs and floating-point precission problems involving this function. Right now the following code: <?php echo round(141.075, 2); ?> returns: 141.07 on my machine. So never really trust this function when you do critical calculations like accounting stuff! Instead: use only integers or use string comparisons.

The function round numbers to a given precision. function toFixed($number, $round=2) { $tempd = $number*pow(10,$round); $tempd1 = round($tempd); $number = $tempd1/pow(10,$round); return $number; } echo round(5.555,2); //retunr 5.55 - I don\t know why echo toFixed(5.555,2); //return 5.56

The function from Felix Schwarz returns wrong results for values near 0. bccomp should be called with param $scale. function bcround ($number, $decimals, $precision) { $precision_add = str_repeat("5",$precision); if (bccomp($number,0, $decimals+$precision) > 0) { return (bcadd (bcadd ($number, "0.".str_repeat("0",$decimals).$precision_add, $decimals+$precision), 0, $decimals)); } else { return (bcadd (bcsub ($number, "0.".str_repeat("0",$decimals).$precision_add, $decimals+$precision), 0, $decimals)); } }

The function below regards a higher number of digits for rounding as the number of digits you want to round! At least it rounds a Value to the number of digits you want to: function MyRound($iValue, $iDigits, $iPrecision){ $iDigits = intval($iDigits); $iPrecision = intval($iPrecision); if($iDigits > $iPrecision){ $iPrecision = $iDigits; } for($i = $iPrecision; $i >= $iDigits; $i--){ $iValue = round($iValue, $i); } // for($i = $iPrecision; $i >= $iDigits; $i--) -- END return $iValue; }

The bankers' rounding behaviour mentioned in the notes below seems to have been changed to normaly rounding. From the comments in http://bugs.php.net/bug.php?id=24142 it was changed in 4.3.0 but various bugs seem to have been introduced and it wasn't stable again until 4.3.5 (http://bugs.php.net/bug.php?id=25694)

Surprisingly, the round() cannot work with 1.255, echo round(1.255, 2); echo " ";// 1.25

Sometimes you will want to make a number divisible by another (like rounding by 5s, or 10s, or 7s, or 1438s, or whatever) instead of just making it divisble by one such as round() with only one argument. The function that follows is fairly simply and behaves mostly as expected. <?php function divby($val, $by) { if ($val < 0) $val += -$val % $by; else $val -= $val % $by; return $val; } ?> divby(23, 10) will return 20 divby(77, 5) will return 75 divby(138, 7) will reuturn 133 divby(-23, 10) will return 20 note: the only way to return a negative value is to make BOTH values negative, for example: divby(-25, -10) will return -20 note also: if the value of $by is greater than $val then the function will return 0, for example: divby(3, 4) will return 0 divby(4, 3) will return 3

smithaapc's trueRound() function is also broken. For example: trueRound(1.11, 3) //returns 1.1, should return 1.11 Here's a function that uses PHP's round() to calculate significant figures: <?php function sigFig($value, $sigFigs) { $exponent = floor(log10($value) + 1); $significand = $value / pow(10, $exponent); $significand = round($significand * pow(10, $sigFigs)) / pow(10, $sigFigs); $value = $significand * pow(10, $exponent); return($value); } //Examples sigFig(123.45, 6); // 123.45 (doesn't pad with zeros) sigFig(123.45, 5); // 123.45 sigFig(123.45, 4); // 123.5 sigFig(123.45, 3); // 123 sigFig(123.45, 2); // 120 (rounds to 2 figures) sigFig(123.45, 1); // 100 sigFig(123.45, 0); // 0 (!) ?>

Since PHP no longer does Banker's Rounding, I came up with an implementation here: http://boonedocks.net/mike/archives/106-Bankers-Rounding-for-PHP.html It hasn't really been hammered on, so testing and comments are welcome.

Since everybody here seems to be occupied wanking over their selfmade round() functions, a little note about the actual php-implementation: If you round() a _very_ small number to a reasonable number of decimals, it will retain its sign! IE: round(1.4 - 1.4, 4) returns "-0" instead of just "0"! Dont ask me how to properly get rid of this. Im just gonna cut it off using preg_replace().

round() is broken, plain and simple. The solution must take into account all of the digits of a given number during rounding. Truncating the number before the rounding eliminates valuable information. Consider: myRound(12.4444444443, 2) = 12.45 ... not quite. Try this out instead: function trueRound($value, $figs=1) { if(is_numeric($value) && is_int($figs) && $figs > 0) { // must adjust $figs down 1 for computing significant figures $figs--; // $prec ensures all of the original digits are included in the sprintf format $prec = strpos($value, '.') === false ? strlen($value)-1 : strlen($value)-2; $sci = sprintf('%.'.$prec.'e', $value); list($sig, $mant) = explode('e', $sci); // $base includes all of the significant digits, $test contains all the remaining digits list($base, $test) = explode('.', $sig * pow(10, $figs)); // All 4's on the left side of $test will not be significant in the rounding $test = preg_replace('/^4+/', '', $test); // The first digit in $test that is not a 4 will determine the direction of the rounding $test = preg_replace('/^(\d)\d+/', '$1', $test); if($test > 4) { // If the digit is > 4, then simply round up by adding 1 ... $base += 1; } // ... then simply adjust the decimal point $value = $base * pow(10, $mant-$figs); } return($value); } Note this function rounds up to a number of significant figures (all digits) rather than significant digits (digits after the decimal point). Rounding down could be easily accomplished by passing a switch along with the other parameters. Mike

Round down.. A very simple way to round down is the following.. $numbera=41; $numberb=3; $sum=$numbera/$numberb; echo $sum; // Will give 13.66666..... $rounddownsum=(int)($numbera/$numberb); echo $rounddownsum; // Will give 13 ;o) If you wana round decimals simply multiply the variables with 10...100...1000 or whatever before performing this action.. Then divide back after calculations.. then ex. $numbera=41; $numberb=3; $sum=$numbera/$numberb; $sum=(int)($sum*1000); $sum=$sum/1000; echo $sum; // Will give 13.666 Hope that helps someone... ;o)

Ron, very good observation, in some occasions the behavior of an accounting system in which I participated, the following problem occurred: reduces it of two exact values did not give zero due to that millionths of difference in the decimals they existed (not yet itself if due to the own algorithms of the microprocessor or of the language, did not I have the opportunity to do a bench, and the surest thing is that this Reason that may not have been able to notice me previously). Your observation I have it present in my code.

Oh sorry, I had mistake in a simple version of the script. Here is a working and tested version: <?php function divby($val, $by) { $val -= $val % $by; return $val; } ?> But in blackhole's description is a little mistake, because if I try: divby(-23, 10); it will return -20, because $val is negative so the result must be negative too.

Note, there is a bug in what appears to be the Windows implementation. See http://bugs.php.net/bug.php?id=36008 for more information and a fix (convert the number to a string before rounding).

my round function function rounding($n,$d){ $base=intval($n); $n=number_format($n,25); $dotpos=strpos($n,"."); $strlen=strlen($n); $b= substr($n,$dotpos+1,$strlen-$dotpos); $j=1; $p=0; for($i=strlen($b);$i>$d;$i--){ $c=substr($b,-$j,1)+$p; if($c<5){ $p=0; }else{ $p=1; } $j++; } return $base+((substr($b,0,$d)+$p)/pow(10,$d)); }

My name is Ionut Socol and i'm from Romania A small script that allow you to round up, round down or just return the integer number. The $direction value cand be: 0 - rounding down 1 - rounding up any thing else for returning the integer value <?php function rounding($no,$direction) { $skip=0; if(is_float($no) and $direction = 1) { $exploded = explode(".",$no); $nrr = $exploded[0]+1; $skip=1; } if(is_float($no) and $direction = 0) { $exploded = explode(".",$no); $nrr = $exploded[0]; $skip=1; } if(!is_float($no) and $skip ==1) { $nrr = $nrr; } else { $nrr = floor($nrr); } return $nrr; } ?>

Many have thus far mentioned problems encountered when trying to add a small fuzz factor to a number such as 1.499999999. This is the way I get around that problem using , allbeit probably less efficient than assuming a small possiblitiy for error: $numberToRound = 1.5; //Convert to string. $numberToRound = "$numberToRound"; //iff number ends in a "5", add fuzz if (eregi("$5", $pages)) $pages += .000001; $round = round($pages, 0);

Just a quick note on rounding to the nearest "half" instead of nearest "whole". I my case, I'm showing a user rating with stars and need to show a star for each whole, and half a star for remainder values closer to half than whole. // replace 3.45 with any rating $_actual_rating = 3.45; $_rating_round_half = round($_actual_rating / 0.5) * 0.5; $_remainder = fmod($_rating_round_half, 1); $_rating = (int) $_rating_round_half - $_remainder; $_half_star = $_remainder == 0.5; In the above example, $_rating will contain a whole number for the number of stars to display, and $_half_star will be a boolean to determine whether or not to display half a star as well.

Javascript round_number function roundnumber (n, d) { n = n - 0; if (d == null) d = 2; var f = Math.pow(10, d); n += Math.pow(10, - (d + 1)); n = Math.round(n * f) / f; n += Math.pow(10, - (d + 1)); n += ''; return d == 0 ? n.substring(0, n.indexOf('.')) : n.substring(0, n.indexOf('.') + d + 1); } PHP round_number; function roundnumber($n, $d = 0) { $n = $n - 0; if ($d === NULL) $d = 2; $f = pow(10, $d); $n += pow(10, - ($d + 1)); $n = round($n * $f) / $f; $n += pow(10, - ($d + 1)); $n += ''; if ( $d == 0 ) : return substr($n, 0, strpos($n, '.')); else : return substr($n, 0, strpos($n, '.') + $d + 1); endif; }

Instead of writing roundDown() and roundUp() functions in php, you might want to consider floor() and ceil(), as they are probably much, much faster.

In response to maxteiber at gmail dot com, regarding rounding 141.075 to 2 decimal places: I can justify your computer for rounding it to 141.07. It's following a rounding rule similar to the "significant figures" rules in science. Try rounding 141.0751 or 141.0756.

In response to boonedocks at gmail dot com Your calculations are slightly incorrect. You only use the formula when the 1/1000 position is equal to 5, otherwise you follow standard rounding proceedures. I wrote a quick snippet which works with the input as a string, uses inline substrings and substitutions. Works VERY fast. <?php function bankers_round ($moneyfloat = null) { $money_str = sprintf("%01.3f", round($moneyfloat, 3)); // convert to rounded (to the nearest thousandth) string $thous_pos = strlen($money_str)-1; // Thousandth string position $hundt_pos = strlen($money_str)-2; // Hundredth string position if ($money_str[$thous_pos] === "5") $money_str[$thous_pos] = ((int)$money_str[$hundt_pos] & 1) ? "9" : "0"; // use a bitwise test, its faster than modulus // The above statement assists round() by setting the thousandth position to 9 or zero if the hundredth is even or odd (respectively) return round($money_str, 2); // Return rounded value } ?> Hitlers Pet Gerbil

If you'd only want to round for displaying variables (not for calculating on the rounded result) then you should use printf with the float: printf ("%6.2f",3.39532); This returns: 3.40 .

If you want more speed, and you dont need any decimals returned, use this: <?php $number = 254.51263; $round = (int)($number+0.5); // 255 ?> This way is 32% faster than calling the round() function in this situation (tested!).

If you want a complete and simple solution to round for n digits you should use this function at the bottom. (you may want this to be even clearer with nice defines like below but it is not obliged if you use -1/0/1) <?php define(FLOOR,-1); define(NEARER,0); define(CEIL,1); ?> $myNiceValue=MyApprox($value,3); <- nearest value for 3 digits $myNiceValue=MyApprox($value,3,FLOOR); <- floor value ex: MyApprox(2.36745,3); // 2.37 MyApprox(1462.36745,3); // 146 (because nearer) MyApprox(1465.36745,3); // 147 (because nearer) MyApprox(1462.36745,3,FLOOR); // 146 MyApprox(1462.36745,3,CEIL); // 147 MyApprox(-1462.36745,3); // -146 (because nearer) MyApprox(-1465.36745,3); // -147 (because nearer) MyApprox(-1462.36745,3,FLOOR); // -147 <-- like PHP but for n digit MyApprox(-1462.36745,3,CEIL); // 146 <-- like PHP but for n digit The last argument ($inverseNeg) is provided only if you want to switch floor and ceil for negative value to overide the common behaviour of php. (little chances it would be useful for you) This must work for php > 4.0.5 if you want to make it work for < versions you may only have to change the str_replace to do not use arrays but 2 str_replace instead. It may be really interesting to have this kind of functions directly in php ! Plz note that this one is probably not the quicker code possible I coded it quickly. (if you need speed it's better to use directly the method you want anyway) Note that it's simple to use a "common" behavior in your script(s) : define(APPROX_METHOD,-1) MyApprox($value,$n,APPROX_METHOD); ... So you can change once for all the method or even by choice of your user. <?php function MyApprox($fValue,$sigdigits=0,$mode=0,$inverseNeg=0) // mode=-1 gives floor // mode=0 (default) gives automatic Round to nearest digit : >5 -> +1 // mode=1 gives ceil // inverseNeg=0 (default)-> PHP floor/ceil behavior, floor(-2.3)=-6, floor towards -INF // inverseNeg=1 -> inverse floor/ceil behavior means floor(-2.3)=-5, floor towards 0 in case of need { $sValue=''.(float)($fValue); $lenght=strlen($sValue); $isNeg=$fValue<0; if(($sigdigits<=0)||($sigdigits==False)) { if(($inverseNeg)&&($isNeg)) $mode=-$mode; if($mode==-1) return floor($fValue); if($mode==1) return ceil($fValue); return round($fValue); } $posDot=strpos($sValue,'.'); $sResult=str_replace(array('.','-'),'', $sValue); $fResult=0.0; if(($mode==0)&&(intval($sResult{$sigdigits})>=5)) $fResult=1.0; $sResult=substr($sResult,0,$sigdigits); $fResult+=(float)$sResult; if($isNeg) { if($inverseNeg) { if($mode==1) $fResult+=1; } else if($mode==-1) $fResult+=1; } else if($mode==1) $fResult+=1; if($posDot==False) $posDot=$lenght; $numb=$sigdigits-($posDot-$isNeg); if($numb>0) for($i=0;$i<$numb;$i++) $fResult/=10.0; else if($numb<0) for($i=0;$i<$numb;$i++) $fResult*=10.0; if($isNeg) return -$fResult; else return $fResult; } ?>

If you didn't want to use 'ceil()' or 'floor()' as it only rounds to a whole number u could do this: <?php $actual_value = 3.352; $number = 0.01; //how many decimal places you want it to be $temp1 = $actual_value * 2; $temp2 = $temp1 + $number; //'+ $number' if rounding up '- $number' if rounding down $temp3 = $temp2 / 2; $new_value = round($temp3, 2); echo $new_value; // 3.36 ?>

I'm sure its been done before, but here's my example of a round up function. This function allows you to specify the number of decimal places to round up to. Eg, when rounding up to 3 decimal places, this function adds 0.0005 to the original value and performs round(), or for 6 decimal places, adds 0.0000005 and performs round(), etc... Hopefully some of you will find it useful: <?php function roundup ($value, $dp) { // Offset to add to $value to cause round() to round up to nearest significant digit for '$dp' decimal places $offset = pow (10, -($dp + 1)) * 5; return round ($value + $offset, $dp); } ?> Please post if you have any comments or improvements on this :-)

I just found out then that even if you round a double (3.7) to an integer (4), it's data type remains as 'double'. So it's always good to use the settype() function when using the round() function to prevent any problems with your scripts.

I have made the round_up-function much more simpler and useable of any values (float or int) -> if no decimal values then the value is taken else we add one to the value in front of the dot ;) function round_up($value){ list($val,$dummy) = explode(".",$value); return $dummy?$val+1:$val; }

I created a round down function, pretty basic but it should work :P <?php function round_down($value) { if(is_float($value)) { $exploded = explode(".",$value); return $exploded[0]; } else { die("Only float values can be entered."); } } ?> You could use the same function for rounding up: <?php function round_up($value) { if(is_float($value)) { $exploded = explode(".",$value); return $exploded[0] + 1; } else { die("Only float values can be entered."); } } ?>

Here's a function to round to an arbitary number of significant digits. Don't confuse it with rounding to a negative precision - that counts back from the decimal point, this function counts forward from the Most Significant Digit. ex: round(1241757, -3); // 1242000 RoundSigDigs(1241757, 3); // 1240000 Works on negative numbers too. $sigdigs should be >= 0 function RoundSigDigs($number, $sigdigs) { $multiplier = 1; while ($number < 0.1) { $number *= 10; $multiplier /= 10; } while ($number >= 1) { $number /= 10; $multiplier *= 10; } return round($number, $sigdigs) * $multiplier; }

Here is my little round function. Works nice on php 4.3.11 and 5.0.5 printf("0.285 - %s ",round(0.285,2)); // incorrect result 0.28 printf("1.285 - %s ",round(1.285,2)); // correct result 1.29 printf("1.255 - %s ",round(1.255,2)); // incorrect result 1.25 printf("0.285 - %s ",myRound(0.285,2)); // incorrect result 0.29 printf("1.285 - %s ",myRound(1.285,2)); // incorrect result 1.29 printf("1.255 - %s ",myRound(1.255,2)); // incorrect result 1.255 function myRound($value,$round=2) { $value *= pow(10.0,$round); $value = floor(floatval($value + 0.6)); $value /= pow(10.0,$round); return($value); }

Here is a simple version of blackhole's script. <?php function divby($val, $by) { $val -= abs($val) % $by; return $val; } ?>

Having done the following test: <?php $test_numbers = array (1.255,3.255,10.255); foreach ($test_numbers AS $number){ echo " Number: " . $number . ""; echo " Printf(): "; printf("%.2f",$number); echo " Printf() and round(): "; printf("%.2f",round($number,2)); echo " Printf() and round() *100/100: "; printf("%.2f",round($number*100)/100); echo " ----------------"; } ?> it looks that small numbers are not always rounded the way one expects, regardless of the rounding method used. The results are: Number: 1.255 Printf(): 1.25 Printf() and round(): 1.25 Printf() and round() *100/100: 1.25 ---------------- Number: 3.255 Printf(): 3.25 Printf() and round(): 3.25 Printf() and round() *100/100: 3.26 ---------------- Number: 10.255 Printf(): 10.26 Printf() and round(): 10.26 Printf() and round() *100/100: 10.26 So to get around this I came out with a small 'fuzz'-ing: <?php function myFormattedRoundedNumber($number, $fuzz = 0.00000000001){ return sprintf("%.2f", (($number>=0) ? ($number+$fuzz) : ($number-$fuzz))); } ?> As you can see, it also takes into account the negative numbers.

round

Rounds a float (PHP 4, PHP 5)
float round ( float val [, int precision] )

Example 1159. round() examples

Related Examples ( Source code ) » round

Code Examples / Notes » round

Change Language