php, sprintf and Memory Usage

Published: 01/22/2009

Programming, Code

I was going through some legacy code for a client and ran across a whole bunch of calls to sprintf(). The calls were all similiar to the below:

$query = sprintf('SELECT * FROM table WHERE field1 = "%s" AND field2 = "%s" LIMIT 1',$var1, $var2);

It seemed an odd use to call the above instead of the straight forward, and low impact, way of declaring the variable.

$query = "SELECT * FROM table WHERE field1 = '$var1' AND field2 = '$var2' LIMIT 1";

I’ve never really had the need to use sprintf() so I’m not familiar with it enough to know much so I asked a colleague of mine, Devin, “What process would require the use of sprintf?”. According to Devin, sprintf is used for complicated and dynamic formatting. Here’s an example from the php manual on the subject:

<?php
$n =  43951789;
$u = -43951789;
$c = 65; // ASCII 65 is 'A'
 
// notice the double %%, this prints a literal '%' character
printf("%%b = '%b'\n", $n); // binary representation
printf("%%c = '%c'\n", $c); // print the ascii character, same as chr() function
printf("%%d = '%d'\n", $n); // standard integer representation
printf("%%e = '%e'\n", $n); // scientific notation
printf("%%u = '%u'\n", $n); // unsigned integer representation of a positive integer
printf("%%u = '%u'\n", $u); // unsigned integer representation of a negative integer
printf("%%f = '%f'\n", $n); // floating point representation
printf("%%o = '%o'\n", $n); // octal representation
printf("%%s = '%s'\n", $n); // string representation
printf("%%x = '%x'\n", $n); // hexadecimal representation (lower-case)
printf("%%X = '%X'\n", $n); // hexadecimal representation (upper-case)
 
printf("%%+d = '%+d'\n", $n); // sign specifier on a positive integer
printf("%%+d = '%+d'\n", $u); // sign specifier on a negative integer
?>

The above outputs:

%b = '10100111101010011010101101'
%c = 'A'
%d = '43951789'
%e = '4.39518e+7'
%u = '43951789'
%u = '4251015507'
%f = '43951789.000000'
%o = '247523255'
%s = '43951789'
%x = '29ea6ad'
%X = '29EA6AD'
%+d = '+43951789'
%+d = '-43951789'

Pretty cool, but I’m not still not sure what the purpose of using sprintf() like the first example demonstrated would accomplish. Surely, there must be some sort of performance penalty, right? According to a programmer Devin used to work, there’s a HUGE penalty for using sprintf. To find out I wrote a simple test script, that’s the farthest thing from scientific, to just get a generic baseline on memory usage and sprintf().

<?php
 
for($i=0;$i<100000;$i++){
	//Call sprintf
}
 
echo filesize_format(memory_get_usage());
echo '<br>';
echo filesize_format(memory_get_peak_usage());
 
function filesize_format($bytes, $format = '%01.5lf %s', $force = '')
{
	$force = strtoupper($force);
	$defaultFormat = '%01d %s';
	if (strlen($format) == 0)
			$format = $defaultFormat;
 
	$bytes = max(0, (int) $bytes);
 
	$units = array('B', 'KB', 'MB', 'GB', 'TB', 'PB');
 
	$power = array_search($force, $units);
 
	if ($power === false) {
		$power = $bytes > 0 ? floor(log($bytes, 1024)) : 0;
	}
 
	return sprintf($format, $bytes / pow(1024, $power), $units);
}
?>

First the baseline. This is without running the loop.

/*
Baseline of 0 loops just php start stop:
Total: 71.39062 KB
Peak: 76.84375 KB
*/

Next, I did a run of sprintf() calls like the below:

/*
Total: 73.76562 KB
Peak: 77.22656 KB
*/
for($i=0;$i<100000;$i++){
	$t = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2" LIMIT 1';
}

Followed by a straight variable declaration:

/*
Total: 72.82812 KB
Peak: 77.17969 KB
*/
for($i=0;$i<100000;$i++){
	$t = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2"  LIMIT 1';
}

Huh… the numbers don’t really add up to all that much. I really thought it’d be more of a difference than that. Well, just to be sure I ran two more tests like the above, except with 4 variables being created instead of 1.

/*
Peak: 76.69531 KB
Peak: 78.28125 KB
*/
for($i=0;$i<100000;$i++){
	$t = sprintf('SELECT col1,col2,col3 FROM table WHERE col1="%s" AND col2="%s" LIMIT 1','1', '2');
	$r = sprintf('SELECT col1,col2,col3 FROM table WHERE col1="%s" AND col2="%s" LIMIT 1','1', '2');
	$k = sprintf('SELECT col1,col2,col3 FROM table WHERE col1="%s" AND col2="%s" LIMIT 1','1', '2');
	$o = sprintf('SELECT col1,col2,col3 FROM table WHERE col1="%s" AND col2="%s" LIMIT 1','1', '2');
 
}

And 

/*
Peak: 73.95312 KB
Peak: 77.48438 KB
*/
for($i=0;$i<100000;$i++){
	$t = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2" LIMIT 1';
	$r = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2" LIMIT 1';
	$k = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2" LIMIT 1';
	$o = 'SELECT col1,col2,col3 FROM table WHERE col1="1" AND col2="2" LIMIT 1';
}

Look at that; nothing. There doesn’t appear to be any significant increase in memory in relation to usage. For a loop that runs 100,000 times compared to a memory increase of only about 1 KB this really isn’t bad. Use sprintf if you want to.

Oh yeah, Devin you were lied to 😊