reading a file on same server

#native_company# #native_desc#

#native_cta#

This site is best viewed in a modern browser with JavaScript enabled.

phantom007

Suppose I have a php file "read_file.php" that is used to read .txt files, the structure of the folder/files are:

/www/read_file.php
/www/public/upload/

I was wondering which will be faster and why in PHP..

Method 1:
read file using relative path

Eg: get_file_contents("public/upload/test.txt");

Method 2:
read file using URL (note that the server which is making the request and the server where the file is being read is same)

Eg: get_file_contents("http://mydomain.com/public/upload/test.txt");

Can someone explain me which is faster and why?

Thanks

NoEffinWay

Relative will always be faster.

When using a URL, the script has to open a http connection to the file, download the file (in a sense) and then read the file.

Relative simply opens the file.

Here are some benchmarks I just ran.
Time to open 40MB file with URL path (Using localhost): 4.9771611690521ms
Time to open 40MB file with URL path (Using localhost): 4.1154129505157ms
Time to open 40MB file with URL path (Using localhost): 4.2288191318512ms
An extreme of: 14.318599939346ms

Time to open 40MB file with relative path: 2.2637801170349ms
Time to open 40MB file with relative path: 2.113431930542ms
Time to open 40MB file with relative path: 2.2605271339417ms

phantom007

oh ok got what I was looking for...

thanks

root

It would be interesting to know what script was used to obtain those figures. Not all servers are equal because your server could be a dedicated rack, a virtual server which will be shared space and that impacts on delivery speed or your server may be one that is a shares server that hibernates until something other than the index page is used.

NoEffinWay

Yeah, fair enough. It was ran on my computer that I develop on, it's a Windows machine. Apache2.2 and PHP5.3.8

Note that this script was ran from my IDE (Removing Apache from the equation all together), I did this just to show an idea of the time it would take.
Code used for Testing:

<?php
echo file_get_contents('demo.txt');
$end = microtime();
echo "Time to open 40MB file with relative path: " . ($end - $_SERVER["REQUEST_TIME_FLOAT"]) . "ms";
?>

Code used to generate file being read:

<?php
$content = "";
for($x=0;$x<1000000;$x++){
    $content .= sha1($x) . "\n";
}

$f = fopen('demo.txt','w+');

fputs($f, $content);

fclose($f);

echo "Done";
?>

root

I was getting strange negative figures so I modified the script and got the following

Time to open 40MB file with relative path: 0.15625ms

Time to open 40MB file with fixed path: 0.140625ms

the script was modified to :-

<?php
$start = microtime();
$var = file_get_contents('H:\Program Files\Abyss Web Server\htdocs\demo.txt');
$end = microtime();
echo "Time to open 40MB file with fixed path: " . ($end-$start) . "ms";
?>

I had the output dumped in to a variable because it is more likely that this would be the scenario on a production server, the 40MB file was produced with the supplied script unchanged.

Unless your script is going to offer up a service that millions of people will be relying on, the minute difference of 1/100th of a second is not going to make that much difference but it will if your server is going to be serving millions of users, the 1/100th of a second difference will make a big difference.

I noted a small variance in both methods, this was approximately 3/100ths of a second over a short period of time.

That was a pretty interesting exercise and experiment and would be interesting to find others who are willing to try both scripts and see what results are gained.

My local machine used to run XAMPP but because developers have basically crippled it with silly security, I dumped it and installed Abyss (which is made by former Apache developers if you look in to the history of the departure!) and PHP 5.4

NoEffinWay

Yeah, that is interesting. I wonder why the time is so different. This is very interesting to me. I ran it again (Note: the {hidden} is work related directories):

Time to open 40MB file with path 'http://localhost:81/{hidden}/demo.txt': 2.5665011405945ms
Time to open 40MB file with path 'demo.txt': 0.096905946731567ms
Time to open 40MB file with path 'C:{hidden}\htdocs{hidden}\demo.txt': 0.055022001266479ms

Time to open 40MB file with path 'http://localhost:81/{hidden}/demo.txt': 2.5337021350861ms
Time to open 40MB file with path 'demo.txt': 0.16361808776855ms
Time to open 40MB file with path 'C:{hidden}\htdocs{hidden}\demo.txt': 0.06325888633728ms

Time to open 40MB file with path 'http://localhost:81/{hidden}/demo.txt': 2.4332671165466ms
Time to open 40MB file with path 'demo.txt': 0.15351891517639ms
Time to open 40MB file with path 'C:{hidden}\htdocs{hidden}\demo.txt': 0.075954914093018ms

I think my times are so much lower this time because on the previous tests, I was running a rather heavy data calculator while running the script.
Here is the script (and again, I ran the script from the IDE):

<?php
$path[] = 'http://localhost:81/{hidden}/demo.txt';
$path[] = 'demo.txt';
$path[] = 'C:\OCPOS\htdocs\{hidden}\demo.txt';

foreach($path as $x => $p){
    $start = microtime(true);
    $c =  file_get_contents($p);
    $end = microtime(true);
    echo "Time to open 40MB file with path '{$p}': " . ($end - $start) . "ms\n";

unset($start,$end,$c);
}
?>

NogDog

Just to be sure we're not talking apples vs. oranges here, the original question was really about the difference between using a file path versus a HTTP request via a URI. The difference between a relative file path and absolute file path is a different issue, and likely to be very minimal.

root

The difference I feel is minimal appart from the fixed path is fixed so any movement in the server will break the script and path therefore an absolute path vs relative will be obvious, the advantage being that with a relative path the server works out the path to the file.

NoEffinWay

Im not sure about \.\ here but my tests include a URI accessed file. (As shown in my second set of tests)

root

Yes both methods were used.

Time depends on the production server and its load.

If the host is shared or dedicated will also play a part in speed.