Question about how fopen works.

habbardone

Hi,

I have been reading up on fsockopen, fopen file() and file_get_contents() and am not sure if I understand it properly.

When we use these commands to download a web page, I assume that in reality PHP is ultimately using the same procedures (code) to connect to the remote server get the page. The different commands allow us access to different aspects of the procedure, correct ?

So this means that when we use the fopen() to open a file stream on a web page, PHP looks after the sockets for us. So PHP opens a socket on a default port by itself. After all a connection via a socket will be needed. So which port does PHP use for this - is it 80 ?

Then when we use file_get_contents(), again PHP must need to use both a socket and a handle, so presumably it uses fopensock() code (as above), then it also uses the fopen() code and creates a default handle where the second parameter is set to "rb" and it uses the fread() code to grab the contents by setting the bytes parameter to the filesize.

So the file_get_contents() is a nice short cut because PHP is handling the necessary steps for us, but behind the scenes the same code is being used to set up a connection on a port, create a file stream and read the file.

Is that about right ?

Would appreciate it if anyone who knows about how this works would correct me or fill me in on the bits I am missing

Thanks

Weedpacket

$buffer = file_get_contents($stream) is roughly short for

$buffer = '';
$handle_foo_unique_id = fopen($stream, 'rb');
while(!feof($handle_foo_unique_id))
{
    $buffer .= fread($handle, 8192);
}
fclose($handle_foo_unique_id);
unset($handle_foo_unique_id);

except that since it's implemented natively in C it has access to lower levels of the operating system and hence lower-level optimisations can be deployed. If you really want to see how it works the source code is freely available.

Both wrap around PHP's stream mechanism; as well as the existing stream types defined (file, http, ftp, php ...) you can define your own.

If you do define your own stream it's quite possible it will involve working with sockets, because you'll be having to do all the protocol stuff yourself (if it wasn't already available, you'd have to implement quite a big chunk of RFC 2616 to create a stream handler for HTTP).

habbardone

Thanks for the answer , that seems to be saying roughly what I was thinking.

What I would like to know is when we use fopen file() and file_get_contents()
does PHP handle the socket functions for us and if so what port does it use ?

Thanks

Installer

By "what port" I assume you mean "what client port" (the server port will, of course, be the well-known IANA port for the particular service, e.g. 80 for http). The client port will be an "ephemeral port" chosen by the operating system from a range of higher-numbered ports. The range used varies among operating systems. For most Linux systems you can see the port range with "sudo cat /proc/sys/net/ipv4/ip_local_port_range". Windows uses 1025-5000 (through Server 2003) and 49152-65535 (the IANA range), I believe, for Vista, 7, and Server 2008.

A packet sniffer, such as Wireshark (for Windows) is a great tool for finding out a lot about individual connections.

habbardone

Is cUrl just another "window" on the same bunch of c code.

That is not meant to be derogatory as I think I will probably chose curl to work with in the main.

Looks like I can access the port number, access method, and everything else with curl, so I think I'll use it for web downloads and uploads etc.

But I assume it uses that same code the fsockopen() and fopen() uses to access the specified url ? Or is it all completely rewritten methods ?

Installer

No, curl uses the libcurl library (that's planly stated in the [URL="http://us:php.net/manual/en/book.curl.php"]curl manual section[/URL]; in the first sentence of the introduction, in fact).

If you want to specify the local port, you can do that with both fopen() and file_get_contents():

$url  = 'http://www.example.com';
$opts = array('socket' => array('bindto' => '0:1234'));
$context = stream_context_create($opts);

// Then either:

$fp = fopen($url, 'rb', false, $context); 
$txt = fread($fp, 8192);
fclose($fp);

// Or:

$txt = file_get_contents('http://www.example.com', false, $context);

You can also specify the method, local interface, and other options in the context.

fsockopen() also allows specifying the port as an optional argument.

Although the libcurl C API has SETOPT_LOCALPORT, it's not implemented in PHP.

habbardone

So if you can do practically everything with
fopen($url, 'rb', false, $context);

and fread($fp, 8192);

Or:

file_get_contents()

Why bother writing the curl library and also why have the fsockopen() ?

Seems a lot of duplicated effort - no ?

Installer

I wouldn't say practically anything. Comparing the http context options to the curl options shows, I think, that curl offers a whole lot more; just for starters the callback function options. There are also the (very useful and underused) curl_multi_* functions. The fopen() and fsockopen() functions provide handles which can be used with different functions (such as stream_copy_to_stream()). The file_get_contents() function was added as a higher-level abstraction of the underlying OS functionality.

Edit: "function" phrase count: 6 🙂

habbardone

So for most of the time, when you just want to get the data from a file,
one can chose whether to use:

1) the curl library or

2) the fopen() with fread() and fwrite() or

3) the file_get_contents() and file_put_contents()

Is that bit right ?

I did notice that (2) has good precision handling with

rewind() and fseek() which maybe you can not do with 1 or 3.

( But I guess you are going to tell me that you can !! )

I also read somewhere that curl is a faster in its operatiopn compared to the others - not sure if it is true though.

even the apart from the

Installer

Most of the time, when retrieving one or just a few files, the difference in speed would be a small concern. For speed on a large number of files, and/or for its wealth of options, I prefer curl.

For precision handling, I don't know, I guess it depends on the type of file (RSS vs. plaintext vs. whatever), how you parse it, etc.

Weedpacket

habbardone wrote:
Seems a lot of duplicated effort - no ?

No. fsockopen is a lower-level function than fopen, cURL is a much higher-level function; they're directed at different parts of the protocol stack. You can't use cURL for BEEP communcations, for instance - you'd have to write your own handler starting with fsockopen(). On the other hand, you wouldn't want to waste time writing an entire HTTP client from fsockopen() up when cURL already exists (which, as you'll note, is not part of PHP but an third-party library developed separately, which is why it duplicates the lower-level stuff internally).