How to programmatically determine the document roo

问题:

Here's a problem that I've been running into lately - a misconfigured apache on a webhost. This means that all scripts that rely on $_SERVER['DOCUMENT_ROOT'] break. The easiest workaround that I've found is just set the variable in some global include files that is shared, but it's a pain not to forget it. My question is, how do I determine the correct document root programatically?

For example, on one host, the setup is like this:

$_SERVER['DOCUMENT_ROOT'] == '/htdocs'

The real document roots are:

test.example.com -> /data/htdocs/example.com/test
www.example.com -> /data/htdocs/example.com/www

And I'd like a script that's run from www.example.com/blog/ (on the path /data/htdocs/example.com/www/blog) to get the correct value of /data/htdocs/example.com/www.

On another host, the setup is a bit different:

$_SERVER['DOCUMENT_ROOT'] == '/srv'
test.example.com -> /home/virtual_web/example.com/public_html/test
www.example.com -> /home/virtual_web/example.com/public_html/www

Is there any solution to this? Or is the only way simply not to ever rely on $_SERVER['DOCUMENT_ROOT'] and fix all the software that I'm running on my sites? Fixing this on the hosting's side doesn't seem to be an option, I've yet to encounter a host where this is was configured correctly. The best I got was a document root pointing to www.example.com, which was at least inside open_basedir - they used yet another naming scheme, www.example.com would point to /u2/www/example_com/data/www/.

回答1:

In PHP5 there is the magic constant __FILE__ that contains the absolute path of the file in which it appears. You can use it in combination with dirname to calculate the document root.

You can put a statement like the following one in a config file

define ('DOCUMENT_ROOT', dirname(__FILE__));

this should do the trick

回答2:

There's no need to modify all scripts.

You can run PHP file before any script is run using auto_prepend_file.

$_SERVER is just an array, you can modify it and set correct $_SERVER['DOCUMENT_ROOT'].

回答3:

This is one reason why people siphon everything through a bootstrap /index.php using htaccess and/or query strings. You can use the dirname( __FILE__ ) trick noted above and get the public base of your app that way.

If you're too far into it to switch to a single entry point, one thing I've seen people do is have a common header to their script which walks up the directory tree to find a file which is unique to the base dir:

function findAppBase( $dir ) {
    if( file_exists( "$dir/unique_file.txt" ) ) {
        return $dir;

    return findAppBase( dirname( $dir ) );
}

$base = findAppBase( dirname( __FILE__ ) );

That code hasn't been tested, and there might be a slicker way using the vars in $_ENV or $_SERVER that will do what you want...

回答4:

Based on http://www.helicron.net/php/:

$localpath=getenv("SCRIPT_NAME");
$absolutepath=getenv("SCRIPT_FILENAME");
$_SERVER['DOCUMENT_ROOT']=substr($absolutepath,0,strpos($absolutepath,$localpath));     

I had to change the basename/realpath trick because it returned an empty string on my host. Instead, I use SCRIPT_FILENAME. This probably won't work on IIS anymore (but the original scripts that used the $_SERVER variable probably wouldn't either).

回答5:

PHP should be setting the current directory to the one the script is in, so as long as that's not broken you should be able to figure out the document root using $_SERVER['SCRIPT_FILENAME'] and getcwd(). (I can't remember all the $_SERVER vars off the top of my head, there might be something in phpinfo() that's more useful.)

回答6:

Why not demand that your webhost configures his servers correctly?

these kind of things tend to linger silently in your code and never get removed (but still active) until someone finally fixes the server. Thén everything will break again.

Or, move your stuff to a host that will work. If this is broken, who knows what you will find next.