fgets() vs. file()

[Anon]

Hello,

Has anyone ever tested fgets() vs. file() to see which is faster? I was loading a 356 line HTML file via file(), then checking it for template data. I switched from the one line file() call to this 6 lines code:

$file_content = array();
$fp = fopen( $html_filename, "r" );
while( $buf = fgets( $fp, 1024 ) ) {
$file_content[] = $buf;
}
fclose( $fp );

Just off the top my benchmarking showed a 2-3 second improvement. Just coincidence, or is there some more "scientific" proof of this?

Travis

[Anon]

Curious, I threw together the following benchmark to get a basic idea of which method is fastest, and threw in fread as well.

Here's the benchmark:

<?
$file="test.txt";
$num=20;
$min=100;
for ($i=0;$i<$num;$i++){

unset ($lines);
unset ($body);
$start=microtime();
$fp=fopen($file,"r");
$body=fread($fp,filesize($file));
fclose($fp);
$lines=explode("\n",$body);
$stop=microtime();
$tmp1=split(" ",$start);
$tmp2=split(" ",$stop);
$time=($tmp2[1]-$tmp1[1])+$tmp2[0]-$tmp1[0];
$fread_raw+=$time;
if ($time>$max) $max=$time;
if ($time<$min) $min=$time;
}
$fread_avg=$fread_raw/$num;
print "Max time: $max<BR>";
print "Min time: $min<BR>";
print "Average of fread over $num times: $fread_avg<P>\n";
flush();

$min=100;
for ($i=0;$i<$num;$i++){

unset ($lines);
$start=microtime();
$lines=file($file);
$stop=microtime();
$tmp1=split(" ",$start);
$tmp2=split(" ",$stop);
$time=($tmp2[1]-$tmp1[1])+$tmp2[0]-$tmp1[0];
$file_raw+=$time;
if ($time>$max) $max=$time;
if ($time<$min) $min=$time;
}
$file_avg=$file_raw/$num;
print "Max time: $max<BR>";
print "Min time: $min<BR>";
print "Average of file() over $num times: $file_avg<P>";
flush();

$min=100;
for ($i=0;$i<$num;$i++){
unset ($lines);
$start=microtime();
$fp=fopen($file,"r");
while (!feof($fp)){
$lines[]=fgets($fp,1000);
}
fclose($fp);
$stop=microtime();
$tmp1=split(" ",$start);
$tmp2=split(" ",$stop);
$time=($tmp2[1]-$tmp1[1])+$tmp2[0]-$tmp1[0];
$fread_raw+=$time;
if ($time>$max) $max=$time;
if ($time<$min) $min=$time;
}
$fread_avg=$fread_raw/$num;
print "Max time: $max<BR>";
print "Min time: $min<BR>";
print "Average of fgets() over $num times: $fread_avg<P>\n";
flush();
?>

And here's the results:

Max time: 0.258759
Min time: 0.061351
Average of fread over 20 times: 0.13451405

Max time: 0.258759
Min time: 0.097068
Average of file() over 20 times: 0.14572115

Max time: 0.900884
Min time: 0.789024
Average of fgets() over 20 times: 0.9740204

I'm not sure my fgets is as optimized as it could be, but those are the numbers.

Note that fread and split("\n",$body) result in array members that do NOT have a \n on the end of each, but with file() you do. If you need your data one way or the other, then that would be the main reason to choose either one or the other.

[Anon]

This is interesting. I took the code above and changed the fgets() statement to this..

while( $lines[] = fgets( $fp, 1024 ) ) {
#nothing
}

And added this to the fread() statement

reset( $lines );
while( list( $name ) = each( $lines ) ){
$lines[$name] .= "\n";
}

That takes the advantage off of the fread over file. After this, I took and ran the same benchmark three times in a table to see how they stacked up repeatedly. Here's what I got..

1st
Max time: 0.482866
Min time: 0.034758
Average of fread over 20 times: 0.1334147

2nd
Max time: 0.212973
Min time: 0.038108
Average of fread over 20 times: 0.07108205

3rd
Max time: 0.276189
Min time: 0.034615
Average of fread over 20 times: 0.09126225

1st
Max time: 0.292695
Min time: 0.006059
Average of file() over 20 times: 0.04336025

2nd
Max time: 0.113599
Min time: 0.006252
Average of file() over 20 times: 0.01594175

3rd
Max time: 0.053411
Min time: 0.006066
Average of file() over 20 times: 0.011429

1st
Max time: 0.582056
Min time: 0.021118
Average of fgets() over 20 times: 0.20118685

2nd
Max time: 0.347907
Min time: 0.022264
Average of fgets() over 20 times: 0.14828945

3rd
Max time: 0.679353
Min time: 0.021692
Average of fgets() over 20 times: 0.2773081

I was running a 325 line 15k file through it, and file (to my amazement) does the best. file() is the only one that had a .00 as the low, and it had the lowest peaks.

One note on the code you posted, you need to add "$max=0;" on the line after "$min=100;" to get an accurate max reading..

Travis