URL 函数
在线手册：中文 英文
PHP手册

parse_url

(PHP 4, PHP 5)

parse_url — 解析 URL，返回其组成部分

说明

array parse_url ( string $url )

本函数解析一个 URL 并返回一个关联数组，包含在 URL 中出现的各种组成部分。

本函数不是用来验证给定 URL 的合法性的，只是将其分解为下面列出的部分。不完整的 URL 也被接受，parse_url() 会尝试尽量正确地将其解析。

参数

url

要解析的 URL

返回值

对严重不合格的 URL，parse_url() 可能会返回 FALSE 并发出 E_WARNING。否则会返回一个关联数组，其组成部分为（至少有一个）：

更新日志

版本 说明
5.3.3 Removed the E_WARNING that was emitted when URL parsing failed.
5.1.2 Added the component parameter.

范例

Example #1 parse_url() 例子

<?php
$url = 'http://username:password@hostname/path?arg=value#anchor';

print_r(parse_url($url));
?>

以上例程会输出：

Array
(
    [scheme] => http
    [host] => hostname
    [user] => username
    [pass] => password
    [path] => /path
    [query] => arg=value
    [fragment] => anchor
)

注释

Note:

本函数不能用于相对 URL。

Note:

parse_url() 是专门用来解析 URL 而不是 URI 的。不过为遵从 PHP 向后兼容的需要有个例外，对 file:// 协议允许三个斜线（file:///...）。其它任何协议都不能这样。

参见

URL 函数
在线手册：中文 英文
PHP手册
PHP手册 - N: 解析 URL，返回其组成部分

用户评论:

Anonymous (13-Mar-2012 09:35)

Parsing ftp urls
find port host path
allowed password char "@"

function parseFtpUrl($addres) { 
        $addres = str_replace("ftp://", "", trim($addres));
        if($addres=='') return false;
        $tmp = $addres;
        $tmp = explode("@", $tmp);
        $host_port = array_pop($tmp);
        print_r($host_port);
        if(eregi(":",$host_port)) {
            $host_port = explode(":", $host_port);
            $host = $host_port[0];
            $port = $host_port[1];
            if(eregi("/", $port)) {
                $port2 = explode("/",$port);
                $port = array_shift($port2);
                $path = implode("/", $port2);
            }
        } else {
            $host = $host_port;
            $port = 21;
        }

        $login_pass = implode("@", $tmp);
        $login_pass = explode(":", $login_pass);
        $pass  = array_pop($login_pass);
        $login = implode("", $login_pass);
       
        $result['user'] = $login;
        $result['pass'] = $pass;
        $result['host'] = $host;
        $result['port'] = $port;
        $result['path'] = @$path;
    return $result; 
} 

example

$urls[] = "ftp://login:p@@Sword@hozst.com:22/dfsdfdsfds/dfsd/index.php";
$urls[] = "ftp://login:paSSword@hozst.com";
$urls[] = "login:paSSword@hozst.com/rwerewre/ferfe/";
$urls[] = "login:paSSword@hozst.com";
foreach($urls as $addres ) {
    echo $addres."\n";
    print_r(parseFtpUrl($addres));
    echo "\n\n";
    }

RESULT
ftp://login:p@@Sword@hozst.com:22/dfsdfdsfds/dfsd/index.php
Array
(
    [user] => login
    [pass] => p@@Sword
    [host] => hozst.com
    [port] => 22
    [path] => dfsdfdsfds/dfsd/index.php
)

ftp://login:paSSword@hozst.com
Array
(
    [user] => login
    [pass] => paSSword
    [host] => hozst.com
    [port] => 21
    [path] =>
)

login:paSSword@hozst.com/rwerewre/ferfe/
Array
(
    [user] => login
    [pass] => paSSword
    [host] => hozst.com/rwerewre/ferfe/
    [port] => 21
    [path] =>
)

login:paSSword@hozst.com
Array
(
    [user] => login
    [pass] => paSSword
    [host] => hozst.com
    [port] => 21
    [path] =>
)

therselman at gmail (28-Jan-2012 09:03)

UTF-8 aware parse_url() replacement.

I've realized that even though UTF-8 characters are not allowed in URL's, I have to work with a lot of them and parse_url() will break.

Based largely on the work of "mallluhuct at gmail dot com", I added parse_url() compatible "named values" which makes the array values a lot easier to work with (instead of just numbers). I also implemented detection of port, username/password and a back-reference to better detect URL's like this: //en.wikipedia.com
... which, although is technically an invalid URL, it's used extensively on sites like wikipedia in the href of anchor tags where it's valid in browsers (one of the types of URL's you have to support when crawling pages). This will be accurately detected as the host name instead of "path" as in all other examples.

I will submit my complete function (instead of just the RegExp) which is an almost "drop-in" replacement for parse_url(). It returns a cleaned up array (or false) with values compatible with parse_url(). I could have told the preg_match() not to store the unused extra values, but it would complicate the RegExp and make it more difficult to read, understand and extend. The key to detecting UTF-8 characters is the use of the "u" parameter in preg_match().

function parse_utf8_url($url)
{
    static $keys = array('scheme'=>0,'user'=>0,'pass'=>0,'host'=>0,'port'=>0,'path'=>0,'query'=>0,'fragment'=>0);
    if (is_string($url) && preg_match(
            '~^((?P<scheme>[^:/?#]+):(//))?((\3|//)?(?:(?P<user>[^:]+):(?P<pass>[^@]+)@)?(?P<host>[^/?:#]*))(:(?P<port>\d+))?' .
            '(?P<path>[^?#]*)(\?(?P<query>[^#]*))?(#(?P<fragment>.*))?~u', $url, $matches))
    {
        foreach ($matches as $key => $value)
            if (!isset($keys[$key]) || empty($value))
                unset($matches[$key]);
        return $matches;
    }
    return false;
}

UTF-8 URL's can/should be "normalized" after extraction with this function.

Further work on this is welcomed and thanks to all the other people who submitted their work!

PS: I had to break the RegExp up because the site complained about the length and wouldn't allow me to post it!

thomas at gielfeldt dot com (02-Dec-2011 10:50)

[If you haven't yet] been able to find a simple conversion back to string from a parsed url, here's an example:

<?php

$url = 'http://usr:pss@example.com:81/mypath/myfile.html?a=b&b[]=2&b[]=3#myfragment';
if ($url === unparse_url(parse_url($url))) {
  print "YES, they match!\n";
}

function unparse_url($parsed_url) {
  $scheme   = isset($parsed_url['scheme']) ? $parsed_url['scheme'] . '://' : '';
  $host     = isset($parsed_url['host']) ? $parsed_url['host'] : '';
  $port     = isset($parsed_url['port']) ? ':' . $parsed_url['port'] : '';
  $user     = isset($parsed_url['user']) ? $parsed_url['user'] : '';
  $pass     = isset($parsed_url['pass']) ? ':' . $parsed_url['pass']  : '';
  $pass     = ($user || $pass) ? "$pass@" : '';
  $path     = isset($parsed_url['path']) ? $parsed_url['path'] : '';
  $query    = isset($parsed_url['query']) ? '?' . $parsed_url['query'] : '';
  $fragment = isset($parsed_url['fragment']) ? '#' . $parsed_url['fragment'] : '';
  return "$scheme$user$pass$host$port$path$query$fragment";
}

?>

mallluhuct at gmail dot com (19-Jul-2011 03:38)

From rfc3986:

   As the "first-match-wins" algorithm is identical to the "greedy"
   disambiguation method used by POSIX regular expressions, it is
   natural and commonplace to use a regular expression for parsing the
   potential five components of a URI reference.

   The following line is the regular expression for breaking-down a
   well-formed URI reference into its components.

      ^(([^:/?#]+):)?(//([^/?#]*))?([^?#]*)(\?([^#]*))?(#(.*))?
       12            3  4          5       6  7        8 9

   The numbers in the second line above are only to assist readability;
   they indicate the reference points for each subexpression (i.e., each
   paired parenthesis).  We refer to the value matched for subexpression
   <n> as $<n>.  For example, matching the above expression to

      http://www.ics.uci.edu/pub/ietf/uri/#Related

   results in the following subexpression matches:

      $1 = http:
      $2 = http
      $3 = //www.ics.uci.edu
      $4 = www.ics.uci.edu
      $5 = /pub/ietf/uri/
      $6 = <undefined>
      $7 = <undefined>
      $8 = #Related
      $9 = Related

   where <undefined> indicates that the component is not present, as is
   the case for the query component in the above example.  Therefore, we
   can determine the value of the five components as

      scheme    = $2
      authority = $4
      path      = $5
      query     = $7
      fragment  = $9

   Going in the opposite direction, we can recreate a URI reference from
   its components by using the algorithm of Section 5.3.

gustavo dot andriuolo at vulcabras dot com dot ar (13-Jul-2011 02:31)

Here's a method to get the REAL name of a domain. This return just the domain name, not the rest. First check if is not an IP, then return the name:

<?php
   function esip($ip_addr)
{
  //first of all the format of the ip address is matched
  if(preg_match("/^(\d{1,3})\.(\d{1,3})\.(\d{1,3})\.(\d{1,3})$/",$ip_addr))
  {
    //now all the intger values are separated
    $parts=explode(".",$ip_addr);
    //now we need to check each part can range from 0-255
    foreach($parts as $ip_parts)
    {
      if(intval($ip_parts)>255 || intval($ip_parts)<0)
      return FALSE; //if number is not within range of 0-255
    }
    return TRUE;
  }
  else
    return FALSE; //if format of ip address doesn't matches
}

   
   
    function domain($domainb)
    {
    $bits = explode('/', $domainb);
    if ($bits[0]=='http:' || $bits[0]=='https:')
    {
    $domainb= $bits[2];
    } else {
    $domainb= $bits[0];
    }
    unset($bits);
    $bits = explode('.', $domainb);
    $idz=count($bits);
    $idz-=3;
    if (strlen($bits[($idz+2)])==2) {
    $url=$bits[$idz].'.'.$bits[($idz+1)].'.'.$bits[($idz+2)];
    } else if (strlen($bits[($idz+2)])==0) {
    $url=$bits[($idz)].'.'.$bits[($idz+1)];
    } else {
    $url=$bits[($idz+1)].'.'.$bits[($idz+2)];
    }
    return $url;
    }
   
    $address='clients1.sub3.google.co.uk';
       $parsed_url = parse_url($address);
      $check = esip($parsed_url['host']);
    $host = $parsed_url['host'];
    if ($check == FALSE){
        if ($host != ""){
        $host = domain($host);
        }else{
        $host = domain($address);   
            }
    }
    echo $host;
    ?>
This return: google.co.uk
'http://sub1.sub2.sub3.example.com:443';
return: example.com
'example.com';
return: example.com

Many times  parse_url return nothing when domain is google.com only for example.

Now, google.com, or google.co.uk will return the same.

Maybe is a little dirty, but works well for me, i use it to group internet access log from squid.

Regards.

David Beck (04-Jul-2011 06:54)

Lengthy function to correct (mainly user-submitted) URLs.

<?php

function correctURL($address)
{
    if (!empty($address) AND $address{0} != '#' AND
    strpos(strtolower($address), 'mailto:') === FALSE AND
    strpos(strtolower($address), 'javascript:') === FALSE)
    {
        $address = explode('/', $address);
        $keys = array_keys($address, '..');

        foreach($keys AS $keypos => $key)
            array_splice($address, $key - ($keypos * 2 + 1), 2);

        $address = implode('/', $address);
        $address = str_replace('./', '', $address);
       
        $scheme = parse_url($address);
       
        if (empty($scheme['scheme']))
            $address = 'http://' . $address;

        $parts = parse_url($address);
        $address = strtolower($parts['scheme']) . '://';

        if (!empty($parts['user']))
        {
            $address .= $parts['user'];

            if (!empty($parts['pass']))
                $address .= ':' . $parts['pass'];

            $address .= '@';
        }

        if (!empty($parts['host']))
        {
            $host = str_replace(',', '.', strtolower($parts['host']));

            if (strpos(ltrim($host, 'www.'), '.') === FALSE)
                $host .= '.com';

            $address .= $host;
        }

        if (!empty($parts['port']))
            $address .= ':' . $parts['port'];

        $address .= '/';

        if (!empty($parts['path']))
        {
            $path = trim($parts['path'], ' /\\');

            if (!empty($path) AND strpos($path, '.') === FALSE)
                $path .= '/';
               
            $address .= $path;
        }

        if (!empty($parts['query']))
            $address .= '?' . $parts['query'];

        return $address;
    }

    else
        return FALSE;
}

?>

Simon D (21-Jun-2011 02:45)

To get the params (url query) as Associative array, use this function:

<?php
/**
 * Returns the url query as associative array
 *
 * @param    string    query
 * @return    array    params
 */
function convertUrlQuery($query) {
    $queryParts = explode('&', $query);
   
    $params = array();
    foreach ($queryParts as $param) {
        $item = explode('=', $param);
        $params[$item[0]] = $item[1];
    }
   
    return $params;
}
?>

Egor Chernodarov (06-Apr-2011 09:34)

Noticed the following differences in error handling:

<?php
    print_r(parse_url('ftp://user:password@host:port'));
?>

In PHP 5.2.6 returns:
Array
(
    [scheme] => ftp
    [host] => host
    [user] => user
    [pass] => password
)

port is just skipped.

But in PHP 5.3.6 returns NULL without any warnings.

bahtiar at gadimov dot de (16-Dec-2010 01:38)

Hi,

if you have problems with UTF8 encoded urls please see http://bugs.php.net/bug.php?id=52923 . parse_url breaks the utf8. :( You have to implement it yourself.

Mark Dobrinic (10-Dec-2010 10:55)

It seems the host-part strips off the last [:port] off the end of the hostname

When something is wrong in the actual request, this proves to be the wrong way to do things.

It would be better to not strip off the [:port], but to keep the string *before* the first [:port] as the hostname.

Problem with (maybe malformed) provided HTTP_HOST
hostname:443:443
that resolved in
'host' => 'hostname:443'

Which gave me problems.

Solution would be to enforce this yourself, explicitly:

<?php
$p = parse_url( $url );
$host = explode(':', $p['host']);
$hostname = $host[0];
?>

jesse at example dot com (26-Sep-2010 03:48)

@ solenoid: Your code was very helpful, but it fails when the current URL has no query string (it appends '&' instead of '?' before the query).  Below is a fixed version that catches this edge case and corrects it.

<?php
function modify_url($mod)
{
    $url = "http://".$_SERVER['HTTP_HOST'].$_SERVER['REQUEST_URI'];
    $query = explode("&", $_SERVER['QUERY_STRING']);
    if (!$_SERVER['QUERY_STRING']) {$queryStart = "?";} else {$queryStart = "&";}
    // modify/delete data
    foreach($query as $q)
    {
        list($key, $value) = explode("=", $q);
        if(array_key_exists($key, $mod))
        {
            if($mod[$key])
            {
                $url = preg_replace('/'.$key.'='.$value.'/', $key.'='.$mod[$key], $url);
            }
            else
            {
                $url = preg_replace('/&?'.$key.'='.$value.'/', '', $url);
            }
        }
    }
    // add new data
    foreach($mod as $key => $value)
    {
        if($value && !preg_match('/'.$key.'=/', $url))
        {
            $url .= $queryStart.$key.'='.$value;
        }
    }
    return $url;
}
?>

martin at planio dot com (17-Sep-2010 03:02)

For those of you sending URLs in HTML emails with a redirect address in the query string, note that Hotmail unescapes / and : characters in the query string. So that breaks the parse_url() function call. Take this as an example:
href="http://example.com/redirect?url=http%3A%2F%2Fplanio.com"
Hotmail will transform it to this:
href="http://example.com/redirect?url=http://planio.com"

The solution is to be preventive before the call to parse_url():
<?php
$q_index = strpos($uri, '?');
if ($q_index !== FALSE &&
  (strpos($uri, ':', $q_index) !== FALSE || strpos($uri, '/', $q_index) !== FALSE)) {
  $begin = substr($uri, 0, $q_index);
  $end = substr($uri, $q_index, strlen($uri)-$q_index);
  $end = str_replace('/', '%2F', $end);
  $end = str_replace(':', '%3A', $end);
  $uri = $begin.$end;
}
?>

przemek at sobstel dot org (19-Aug-2010 09:44)

If you want to get host, function will return NULL if you pass only host.

Example:

<?php
parse_url($url, PHP_URL_HOST);
?>

$url => value returned

http://example.com => string 'example.com' (length=11)
http://www.example.com =>string 'www.example.com' (length=15)
http://www.example.com:8080 => string 'www.example.com' (length=15)
example.com => null
www.example.com => null
example.com:8080 => string 'example.com' (length=11)
www.example.com:8080 => string 'www.example.com' (length=15)

solenoid at example dot com (23-Apr-2010 12:05)

Here's a piece of code that modifies, replaces or removes the url query. This can typically used in paging situations where there are more parameters than the page.

<?php
function modify_url($mod)
{
    $url = "http://".$_SERVER['HTTP_HOST'].$_SERVER['REQUEST_URI'];
    $query = explode("&", $_SERVER['QUERY_STRING']);
    // modify/delete data
    foreach($query as $q)
    {
        list($key, $value) = explode("=", $q);
        if(array_key_exists($key, $mod))
        {
            if($mod[$key])
            {
                $url = preg_replace('/'.$key.'='.$value.'/', $key.'='.$mod[$key], $url);
            }
            else
            {
                $url = preg_replace('/&?'.$key.'='.$value.'/', '', $url);
            }
        }
    }
    // add new data
    foreach($mod as $key => $value)
    {
        if($value && !preg_match('/'.$key.'=/', $url))
        {
            $url .= '&'.$key.'='.$value;
        }
    }
    return $url;
}

// page url: "http://www.example.com/page.php?p=5&show=list&style=23"

$url = modify_url(array('p' => 4, 'show' => 'column'));

// $url = "http://www.example.com/page.php?p=4&show=column&style=23"
?>

james at roundeights dot com (26-Feb-2010 07:24)

I was writing unit tests and needed to cause this function to kick out an error and return FALSE in order to test a specific execution path. If anyone else needs to force a failure, the following inputs will work:

<?php
parse_url("http:///example.com");
parse_url("http://:80");
parse_url("http://user@:80");
?>

need_sunny at yahoo dot com (25-Dec-2009 01:57)

Thanks to xellisx for his parse_query function. I used it in one of my projects and it works well. But it has an error. I fixed the error and improved it a little bit. Here is my version of it:

<?php
// Originally written by xellisx
function parse_query($var)
 {
  /**
   *  Use this function to parse out the query array element from
   *  the output of parse_url().
   */
  $var  = parse_url($var, PHP_URL_QUERY);
  $var  = html_entity_decode($var);
  $var  = explode('&', $var);
  $arr  = array();

  foreach($var as $val)
   {
    $x          = explode('=', $val);
    $arr[$x[0]] = $x[1];
   }
  unset($val, $x, $var);
  return $arr;
 }
?>

At the first line there was parse_query($val), I made it $var. It used to return a null array before this fix.

I have added the parse_url line. So now the function will only focus in the query part, not the whole URL. This is useful if something like below is done:
<?php
$my_GET = parse_query($_SERVER['REQUEST_URI']);
?>

nirazuelos at gmail dot com (09-Oct-2009 10:45)

Hello, for some odd reason, parse_url returns the host (ex. example.com) as the path when no scheme is provided in the input url. So I've written a quick function to get the real host:

<?php
function getHost($Address) {
   $parseUrl = parse_url(trim($Address));
   return trim($parseUrl[host] ? $parseUrl[host] : array_shift(explode('/', $parseUrl[path], 2)));
}

getHost("example.com"); // Gives example.com
getHost("http://example.com"); // Gives example.com
getHost("www.example.com"); // Gives www.example.com
getHost("http://example.com/xyz"); // Gives example.com
?>

You could try anything! It gives the host (including the subdomain if exists).

Hope it helped you.

ap dot public1 at gmail dot com (15-Jul-2009 04:36)

Simple static library that allows easy manipulation of url parameters:

<?php
    /**
     * File provides easy way to manipulate url parameters
     * @author Alexander Podgorny
     */

    class Url {
        /**
         * Splits url into array of it's pieces as follows:
         * [scheme]://[user]:[pass]@[host]/[path]?[query]#[fragment]
         * In addition it adds 'query_params' key which contains array of
         * url-decoded key-value pairs
         *
         * @param String $sUrl Url
         * @return Array Parsed url pieces
         */
        public static function explode($sUrl) {
            $aUrl = parse_url($sUrl);
            $aUrl['query_params'] = array();
            $aPairs = explode('&', $aUrl['query']);
            DU::show($aPairs);
            foreach($aPairs as $sPair) {
                if (trim($sPair) == '') { continue; }
                list($sKey, $sValue) = explode('=', $sPair);
                $aUrl['query_params'][$sKey] = urldecode($sValue);
            }
            return $aUrl;
        }
        /**
         * Compiles url out of array of it's pieces (returned by explodeUrl)
         * 'query' is ignored if 'query_params' is present
         *
         * @param Array $aUrl Array of url pieces
         */
        public static function implode($aUrl) {
            //[scheme]://[user]:[pass]@[host]/[path]?[query]#[fragment]
           
            $sQuery = '';
           
            // Compile query
            if (isset($aUrl['query_params']) && is_array($aUrl['query_params'])) {
                $aPairs = array();
                foreach ($aUrl['query_params'] as $sKey=>$sValue) {
                    $aPairs[] = $sKey.'='.urlencode($sValue);              
                }
                $sQuery = implode('&', $aPairs);   
            } else {
                $sQuery = $aUrl['query'];
            }
           
            // Compile url
            $sUrl =
                $aUrl['scheme'] . '://' . (
                    isset($aUrl['user']) && $aUrl['user'] != '' && isset($aUrl['pass'])
                       ? $aUrl['user'] . ':' . $aUrl['pass'] . '@'
                       : ''
                ) .
                $aUrl['host'] . (
                    isset($aUrl['path']) && $aUrl['path'] != ''
                       ? $aUrl['path']
                       : ''
                ) . (
                   $sQuery != ''
                       ? '?' . $sQuery
                       : ''
                ) . (
                   isset($aUrl['fragment']) && $aUrl['fragment'] != ''
                       ? '#' . $aUrl['fragment']
                       : ''
                );
            return $sUrl;
        }
        /**
         * Parses url and returns array of key-value pairs of url params
         *
         * @param String $sUrl
         * @return Array
         */
        public static function getParams($sUrl) {
            $aUrl = self::explode($sUrl);
            return $aUrl['query_params'];
        }
        /**
         * Removes existing url params and sets them to those specified in $aParams
         *
         * @param String $sUrl Url
         * @param Array $aParams Array of Key-Value pairs to set url params to
         * @return  String Newly compiled url
         */
        public static function setParams($sUrl, $aParams) {
            $aUrl = self::explode($sUrl);
            $aUrl['query'] = '';
            $aUrl['query_params'] = $aParams;
            return self::implode($aUrl);
        }
        /**
         * Updates values of existing url params and/or adds (if not set) those specified in $aParams
         *
         * @param String $sUrl Url
         * @param Array $aParams Array of Key-Value pairs to set url params to
         * @return  String Newly compiled url
         */
        public static function updateParams($sUrl, $aParams) {
            $aUrl = self::explode($sUrl);
            $aUrl['query'] = '';
            $aUrl['query_params'] = array_merge($aUrl['query_params'], $aParams);
            return self::implode($aUrl);
        }
    }

?>

vuatintac at yahoo dot com (23-Apr-2009 03:54)

careful while use parse_str() function with full url.
see:

<?php
$sLink = "http://localhost/khm/search.php?act=result&q=a";

parse_str($sLink, $vars);

print_r($vars);

echo $sLink = rawurldecode(http_build_query($vars));

// will output:
// Array
// (
//     [http://localhost/khm/search_php?act] => result
//     [q] => a
// )

// http://localhost/khm/search_php?act=result&q=a

// search.php become search_php
?>

theoriginalmarksimpson at gmail dot com (18-Apr-2009 01:08)

An update to the function by FredLudd at gmail dot com.  I added IPv6 functionality as well.

<?php
function j_parseUrl($url) {
  $r  = "(?:([a-z0-9+-._]+)://)?";
  $r .= "(?:";
  $r .=   "(?:((?:[a-z0-9-._~!$&'()*+,;=:]|%[0-9a-f]{2})*)@)?";
  $r .=   "(?:\[((?:[a-z0-9:])*)\])?";
  $r .=   "((?:[a-z0-9-._~!$&'()*+,;=]|%[0-9a-f]{2})*)";
  $r .=   "(?::(\d*))?";
  $r .=   "(/(?:[a-z0-9-._~!$&'()*+,;=:@/]|%[0-9a-f]{2})*)?";
  $r .=   "|";
  $r .=   "(/?";
  $r .=     "(?:[a-z0-9-._~!$&'()*+,;=:@]|%[0-9a-f]{2})+";
  $r .=     "(?:[a-z0-9-._~!$&'()*+,;=:@\/]|%[0-9a-f]{2})*";
  $r .=    ")?";
  $r .= ")";
  $r .= "(?:\?((?:[a-z0-9-._~!$&'()*+,;=:\/?@]|%[0-9a-f]{2})*))?";
  $r .= "(?:#((?:[a-z0-9-._~!$&'()*+,;=:\/?@]|%[0-9a-f]{2})*))?";
  preg_match("`$r`i", $url, $match);
  $parts = array(
            "scheme"=>'',
            "userinfo"=>'',
            "authority"=>'',
            "host"=> '',
            "port"=>'',
            "path"=>'',
            "query"=>'',
            "fragment"=>'');
  switch (count ($match)) {
    case 10: $parts['fragment'] = $match[9];
    case 9: $parts['query'] = $match[8];
    case 8: $parts['path'] =  $match[7];
    case 7: $parts['path'] =  $match[6] . $parts['path'];
    case 6: $parts['port'] =  $match[5];
    case 5: $parts['host'] =  $match[3]?"[".$match[3]."]":$match[4];
    case 4: $parts['userinfo'] =  $match[2];
    case 3: $parts['scheme'] =  $match[1];
  }
  $parts['authority'] = ($parts['userinfo']?$parts['userinfo']."@":"").
                         $parts['host'].
                        ($parts['port']?":".$parts['port']:"");
  return $parts;
}
?>

When using the url

  /* line too long for this site's comment handler */
  "foo://username:password@[2001:4860:0:2001::68]:8042".
      "/over/there/index.dtb;type=animal?name=ferret#nose"

The original would return

Array
(
    [scheme] => foo
    [userinfo] => username:password
    [authority] => username:password@
    [host] =>
    [port] =>
    [path] =>
    [query] =>
    [fragment] =>
)

The new one returns

Array
(
    [scheme] => foo
    [userinfo] => username:password
    [authority] => username:password@[2001:4860:0:2001::68]:8042
    [host] => [2001:4860:0:2001::68]
    [port] => 8042
    [path] => /over/there/index.dtb;type=animal
    [query] => name=ferret
    [fragment] => nose
)

All of the other examples FredLudd used below still work exactly the same.

usrflo at gmx dot de (12-Apr-2009 01:05)

Additional information:

if you like to compute a registered domain name of the parsed host name you can use the PHP library at http://www.dkim-reputation.org/regdom-libs/
It relies on the effective TLD list of the Mozilla Foundation.

pbcomm at gmail dot com (26-Oct-2008 05:55)

Modification to the code from:
theoriginalmarksimpson at gmail dot com

Change:
$r .= "(?:(?P<login>\w+):(?P<pass>\w+)@)?";

Replace with:
$r .= "(?:(?P<login>\w+):?(?P<pass>\w+)?@)?";

This will cover the case the only username is present in the url:

http://username@subdomain.domain.com/index.php?arg1=test#anchor

vdklah at hotmail dot com (17-Oct-2008 10:53)

Some example that determines the URL port.
When port not specified, it derives it from the scheme.

<?php
function getUrlPort( $urlInfo )
{
    if( isset($urlInfo['port']) ) {
        $port = $urlInfo['port'];
    } else { // no port specified; get default port
        if (isset($urlInfo['scheme']) ) {
            switch( $urlInfo['scheme'] ) {
                case 'http':
                    $port = 80; // default for http
                    break;
                case 'https':
                    $port = 443; // default for https
                    break;
                case 'ftp':
                    $port = 21; // default for ftp
                    break;
                case 'ftps':
                    $port = 990; // default for ftps
                    break;
                default:
                    $port = 0; // error; unsupported scheme
                    break;
            }
        } else {
            $port = 0; // error; unknown scheme
        }
    }
    return $port;
}

$url = "http://nl3.php.net/manual/en/function.parse-url.php";
$urlInfo = parse_url( $url );
$urlPort = getUrlPort( $urlInfo );
if( $urlPort !== 0 ) {
    print 'Found URL port: '.$urlPort;
} else {
    print 'ERROR: Could not find port at URL: '.$url;
}
?>

gautam at rogers dot com (15-Oct-2008 07:08)

What about using something like this to safely encoding all the values that are passed in the query portion?

Example input:
http://www.example.com/?first=john&last=smith&email=john@smith.com

Result:
http://www.example.com/?first=john&last=smith&email=john%40smith.com

<?php
function safe_url($url) {
  // Make sure we have a string to work with
  if(!empty($url)) {
    // Explode into URL keys
    $urllist=parse_url($url);

    // Make sure we have a valid result set and a query field
    if(is_array($urllist) && isset($urllist["query"])) {
      // Explode into key/value array
      $keyvalue_list=explode("&",($urllist["query"]));

      // Store resulting key/value pairs
      $keyvalue_result=array();

      foreach($keyvalue_list as $key=>$value) {
        // Explode each individual key/value into an array
        $keyvalue=explode("=",$value);

        // Make sure we have a "key=value" array
        if(count($keyvalue)==2) {
          // Encode the value portion
          $keyvalue[1]=urlencode($keyvalue[1]);

          // Add our key and encoded value into the result
          array_push($keyvalue_result,implode("=",$keyvalue));
        }
      }

      // Repopulate our query key with encoded results
      $urllist["query"]=implode("&",$keyvalue_result);

      // Build the the final output URL
      $url=(isset($urllist["scheme"])?$urllist["scheme"]."://":"").
           (isset($urllist["user"])?$urllist["user"].":":"").
           (isset($urllist["pass"])?$urllist["pass"]."@":"").
           (isset($urllist["host"])?$urllist["host"]:"").
           (isset($urllist["port"])?":".$urllist["port"]:"").
           (isset($urllist["path"])?$urllist["path"]:"").
           (isset($urllist["query"])?"?".$urllist["query"]:"").
           (isset($urllist["fragment"])?"#".$urllist["fragment"]:"");
    }
  }
 
  return $url;
}
?>

FredLudd at gmail dot com (03-Oct-2008 07:24)

Another shot at trying to find a better parser. I noticed that the laulibrius/theoriginalmarksimpson functions didn't quite handle the URL for the page they were displayed on. For my mirror, ca3, this is
  http://ca3.php.net/manual/en/function.parse-url.php

Run it through the function and it parses to
  scheme => http
  login =>
  pass =>
  host => ca3.php.net
  ip =>
  subdomain => ca3
  domain => php.
  extension => net
  port =>
  path => /manual/en/function.parse
  file => function.parse
that is, the file name gets a bit mangled

Rather than tweak the function's regular expression yet again, I opted to adapt a RegExp that served me well in Javascript:

function j_parseUrl($url) {
  $r  = "(?:([a-z0-9+-._]+)://)?";
  $r .= "(?:";
  $r .=   "(?:((?:[a-z0-9-._~!$&'()*+,;=:]|%[0-9a-f]{2})*)@)?";
  $r .=   "((?:[a-z0-9-._~!$&'()*+,;=]|%[0-9a-f]{2})*)";
  $r .=   "(?::(\d*))?";
  $r .=   "(/(?:[a-z0-9-._~!$&'()*+,;=:@/]|%[0-9a-f]{2})*)?";
  $r .=   "|";
  $r .=   "(/?";
  $r .=     "(?:[a-z0-9-._~!$&'()*+,;=:@]|%[0-9a-f]{2})+";
  $r .=     "(?:[a-z0-9-._~!$&'()*+,;=:@\/]|%[0-9a-f]{2})*";
  $r .=    ")?";
  $r .= ")";
  $r .= "(?:\?((?:[a-z0-9-._~!$&'()*+,;=:\/?@]|%[0-9a-f]{2})*))?";
  $r .= "(?:#((?:[a-z0-9-._~!$&'()*+,;=:\/?@]|%[0-9a-f]{2})*))?";
  preg_match("`$r`i", $url, $match);
  $parts = array(
            "scheme"=>'',
            "userinfo"=>'',
            "authority"=>'',
            "host"=> '',
            "port"=>'',
            "path"=>'',
            "query"=>'',
            "fragment"=>'');
  switch (count ($match)) {
    case 9: $parts['fragment'] = $match[8];
    case 8: $parts['query'] = $match[7];
    case 7: $parts['path'] =  $match[6];
    case 6: $parts['path'] =  $match[5] . $parts['path'];
    case 5: $parts['port'] =  $match[4];
    case 4: $parts['host'] =  $match[3];
    case 3: $parts['userinfo'] =  $match[2];
    case 2: $parts['scheme'] =  $match[1];
  }
  $parts['authority'] = ($parts['userinfo']?$parts['userinfo']."@":"").
                         $parts['host'].
                        ($parts['port']?":".$parts['port']:"");
  return $parts;
}

This function, when fed "http://ca3.php.net/manual/en/function.parse-url.php", returns
  scheme => http
  userinfo =>
  authority => ca3.php.net
  host => ca3.php.net
  port =>
  path => /manual/en/function.parse-url.php
  query =>
  fragment =>
which is somewhat closer to my needs.

But everything should be tested against the two examples provided by RFC3986,

  /* line too long for this site's commnet handler */
  "foo://username:password@example.com:8042".
      "/over/there/index.dtb;type=animal?name=ferret#nose"
and
  "urn:example:animal:ferret:nose"

Here the native function parse_url() performs admirably on that "urn:" example. Mine fails to pick out the path ("example:animal:ferret:nose") and the laulibrius/theoriginalmarksimpson function can't decipher anything there. On the "foo:" example, both my function and parse_url() get it right, while the other examples on this page don't.

The laulibrius/theoriginalmarksimpson function delivers
  scheme => foo
  login => username
  pass => password
  host => example.com
  ip =>
  subdomain =>
  domain => example.
  extension => com
  port => 8042
  path => /over/there/index.dtb
  file => index.dtb

As you can see, the query string ("name=ferret") and fragment ("nose") have dropped off, as well as the parameter ("type=animal").

xellisx (01-Oct-2008 04:37)

I need to parse out the query string from the referrer, so I created this function.

<?php
function parse_query($val)
 {
  /**
   *  Use this function to parse out the query array element from
   *  the output of parse_url().
   */
  $var  = html_entity_decode($var);
  $var  = explode('&', $var);
  $arr  = array();

  foreach($var as $val)
   {
    $x          = explode('=', $val);
    $arr[$x[0]] = $x[1];
   }
  unset($val, $x, $var);
  return $arr;
 }
?>

ilja at radusch dot com (26-Sep-2008 12:26)

Here is an update to the glue_url() function.

It can now handle relative URLs if only 'path' is provided.

<?php
function glue_url($parsed) {
    if (!is_array($parsed)) {
        return false;
    }

    $uri = isset($parsed['scheme']) ? $parsed['scheme'].':'.((strtolower($parsed['scheme']) == 'mailto') ? '' : '//') : '';
    $uri .= isset($parsed['user']) ? $parsed['user'].(isset($parsed['pass']) ? ':'.$parsed['pass'] : '').'@' : '';
    $uri .= isset($parsed['host']) ? $parsed['host'] : '';
    $uri .= isset($parsed['port']) ? ':'.$parsed['port'] : '';

    if (isset($parsed['path'])) {
        $uri .= (substr($parsed['path'], 0, 1) == '/') ?
            $parsed['path'] : ((!empty($uri) ? '/' : '' ) . $parsed['path']);
    }

    $uri .= isset($parsed['query']) ? '?'.$parsed['query'] : '';
    $uri .= isset($parsed['fragment']) ? '#'.$parsed['fragment'] : '';

    return $uri;
}
?>

nospam at spellingcow dot com (08-Sep-2008 10:03)

URL's in the query string of a relative URL will cause a problem

fails:
/page.php?foo=bar&url=http://www.example.com

parses:
http://www.foo.com/page.php?foo=bar&url=http://www.example.com

theoriginalmarksimpson at gmail dot com (05-Sep-2008 08:12)

A rehash of code modified by "laulibrius at hotmail dot com".

This also parses urls for hosts that don't have a domain name and just use an IP as the hostname.

The old code would assume that the IP octets were a subdomain.
So the url "http://255.255.255.255/" would return 255.255 as a subdomain of 255.255.

<?php
parseUrl($url)
   {
   $r  = "^(?:(?P<scheme>\w+)://)?";
   $r .= "(?:(?P<login>\w+):(?P<pass>\w+)@)?";

   $ip="(?:[0-9]{1,3}+\.){3}+[0-9]{1,3}";//ip check
   $s="(?P<subdomain>[-\w\.]+)\.)?";//subdomain
   $d="(?P<domain>[-\w]+\.)";//domain
   $e="(?P<extension>\w+)";//extension

   $r.="(?P<host>(?(?=".$ip.")(?P<ip>".$ip.")|(?:".$s.$d.$e."))";

   $r .= "(?::(?P<port>\d+))?";
   $r .= "(?P<path>[\w/]*/(?P<file>\w+(?:\.\w+)?)?)?";
   $r .= "(?:\?(?P<arg>[\w=&]+))?";
   $r .= "(?:#(?P<anchor>\w+))?";
   $r = "!$r!";   // Delimiters
   preg_match($r, $url,$out);
   }
?>

If you need to validate the host IP this is easier than using regex.

<?php

$parsed=parseUrl($url);

if($parsed['ip']) {
   if(long2ip(ip2long($parsed['ip']))==$parsed['ip']){//validates IP
   echo $parsed['ip']." is a valid host";
   }
   else {
   echo $parsed['ip']." is not a valid host";
   }
}
?>

marco panichi (23-Aug-2008 09:47)

my function catch the url written on the browser by the user and does the same thing of parse_url. but better, I think. I don't like parse_url because it says nothing about elements that it doesn't find in the url. my function instead return an empty string.

<?php
function get_url()
{
    $arr = array();
    $uri = $_SERVER['REQUEST_URI'];

    // query
    $x = array_pad( explode( '?', $uri ), 2, false );
    $arr['query'] = ( $x[1] )? $x[1] : '' ;

    // resource
    $x         = array_pad( explode( '/', $x[0] ), 2, false );
    $x_last = array_pop( $x );
    if( strpos( $x_last, '.' ) === false )
    {
        $arr['resource'] = '';
        $x[] = $x_last;
    }
    else
    {
        $arr['resource'] = $x_last;
    }

    // path   
    $arr['path'] = implode( '/', $x );
    if( substr( $arr['path'], -1 ) !== '/' ) $arr['path'] .= '/';

    // domain
    $arr['domain']    = $_SERVER['SERVER_NAME'];

    // scheme
    $server_prt        = explode( '/', $_SERVER['SERVER_PROTOCOL'] );
    $arr['scheme']    = strtolower( $server_prt[0] );

    // url
    $arr['url'] = $arr['scheme'].'://'.$arr['domain'].$uri;

    return $arr;
}
?>

PS: I found working with explode is faster than using preg_match (I tryed with getmicrotime function and 'for' cycles).

PPS: I used array_pad to prevent any notice.

andrewtheartist at hotmail dot com (29-Jun-2008 01:28)

Here's the easiest way to get the URL to the path that your script is in (so not the actual script name itself, just the complete URL to the folder it's in)

echo "http://".$_SERVER['HTTP_HOST'].dirname($_SERVER['PHP_SELF']);

Cool Coyote (23-Jun-2008 01:35)

based on the "laulibrius at hotmail dot com" function, this work for relatives url only:

<?php
function parseUrl($url) {
$r = "^(?:(?P<path>[\.\w/]*/)?";
$r .= "(?P<file>\w+(?:\.\w+)?)?)\.(?P<extension>\w+)?";
$r .= "(?:\?(?P<arg>[\w=&]+))?";
$r .= "(?:#(?P<anchor>\w+))?";
$r = "!$r!";

preg_match ( $r, $url, $out );
return $out;
}

print_r(parseUrl("../test/f.aq.php?p=1&v=blabla#X1"));

?>

returns:
Array
(
    [0] => ../test/faq.php?p=1&v=blabla#X1
    [path] => ../test/
    [1] => ../test/
    [file] => faq
    [2] => faq
    [extension] => php
    [3] => php
    [arg] => p=1&v=blabla
    [4] => p=1&v=blabla
    [anchor] => X1
    [5] => X1
)

laulibrius at hotmail dot com (16-Jun-2008 05:31)

There was one thing missing in the function dropped by "to1ne at hotmail dot com" when i tried it : domain and subdomain couldn't have a dash "-". So i add it in the regexp and the function looks like this now :

<?php
    function parseUrl($url) {
        $r  = "^(?:(?P<scheme>\w+)://)?";
        $r .= "(?:(?P<login>\w+):(?P<pass>\w+)@)?";
        $r .= "(?P<host>(?:(?P<subdomain>[-\w\.]+)\.)?" . "(?P<domain>[-\w]+\.(?P<extension>\w+)))";
        $r .= "(?::(?P<port>\d+))?";
        $r .= "(?P<path>[\w/]*/(?P<file>\w+(?:\.\w+)?)?)?";
        $r .= "(?:\?(?P<arg>[\w=&]+))?";
        $r .= "(?:#(?P<anchor>\w+))?";
        $r = "!$r!";                                                // Delimiters
      
        preg_match ( $r, $url, $out );
      
        return $out;
    }
?>

Btw, thanks for the function, it helps me a lot.

to1ne at hotmail dot com (13-Jun-2008 07:01)

Based on the idea of "jbr at ya-right dot com" have I been working on a new function to parse the url:

<?php
function parseUrl($url) {
    $r  = "^(?:(?P<scheme>\w+)://)?";
    $r .= "(?:(?P<login>\w+):(?P<pass>\w+)@)?";
    $r .= "(?P<host>(?:(?P<subdomain>[\w\.]+)\.)?" . "(?P<domain>\w+\.(?P<extension>\w+)))";
    $r .= "(?::(?P<port>\d+))?";
    $r .= "(?P<path>[\w/]*/(?P<file>\w+(?:\.\w+)?)?)?";
    $r .= "(?:\?(?P<arg>[\w=&]+))?";
    $r .= "(?:#(?P<anchor>\w+))?";
    $r = "!$r!";                                                // Delimiters
   
    preg_match ( $r, $url, $out );
   
    return $out;
}
print_r ( parseUrl ( 'me:you@sub.site.org:29000/pear/validate.html?happy=me&sad=you#url' ) );
?>

This returns:
Array
(
    [0] => me:you@sub.site.org:29000/pear/validate.html?happy=me&sad=you#url
    [scheme] =>
    [1] =>
    [login] => me
    [2] => me
    [pass] => you
    [3] => you
    [host] => sub.site.org
    [4] => sub.site.org
    [subdomain] => sub
    [5] => sub
    [domain] => site.org
    [6] => site.org
    [extension] => org
    [7] => org
    [port] => 29000
    [8] => 29000
    [path] => /pear/validate.html
    [9] => /pear/validate.html
    [file] => validate.html
    [10] => validate.html
    [arg] => happy=me&sad=you
    [11] => happy=me&sad=you
    [anchor] => url
    [12] => url
)

So both named and numbered array keys are possible.

It's quite advanced, but I think it works in any case... Let me know if it doesn't...

jbr at ya-right dot com (03-May-2008 04:24)

This function never works the way you think it should...

Example....

<?php

print_r ( parse_url ( 'me:you@sub.site.org/pear/validate.html?happy=me&sad=you#url' ) );

?>

Returns...

Array
(
    [scheme] => me
    [path] => you@sub.site.org/pear/validate.html
    [query] => happy=me&sad=you
    [fragment] => url
)

Here my way of doing parse_url

<?php

function parseUrl ( $url )
{
    $r  = '!(?:(\w+)://)?(?:(\w+)\:(\w+)@)?([^/:]+)?';
    $r .= '(?:\:(\d*))?([^#?]+)?(?:\?([^#]+))?(?:#(.+$))?!i';

    preg_match ( $r, $url, $out );

    return $out;
}

print_r ( parseUrl ( 'me:you@sub.site.org/pear/validate.html?happy=me&sad=you#url' ) );

?>

Returns...

Array
(
    [0] => me:you@sub.site.org/pear/validate.html?happy=me&sad=you#url
    [1] =>
    [2] => me
    [3] => you
    [4] => sub.site.org
    [5] =>
    [6] => /pear/validate.html
    [7] => happy=me&sad=you
    [8] => url
)

Where as...

out[0] = full url
out[1] = scheme or '' if no scheme was found
out[2] = username or '' if no auth username was found
out[3] = password or '' if no auth password was found
out[4] = domain name or '' if no domain name was found
out[5] = port number or '' if no port number was found
out[6] = path or '' if no path was found
out[7] = query or '' if no query was found
out[8] = fragment or '' if no fragment was found

Nicolas Merlet - admin(at)merletn.org (14-Mar-2008 03:05)

Please note that parse_url seems not to produce always the same results when passing non-standard urls.

Eg. I was using this code since 2005 (both under PHP 4.3.10 and PHP 5.2.3) :

<?php

  $p = parse_url ( 'http://domain.tld/tcp://domain2.tld/dir/file' ) ;

  $d2 = parse_url ( $p['path'] ) ;

  echo $d2 ;  // returns '/dir/file'

?>

Of course my example is very specific, as URL is not really correct. But using parse_url was a great trick to split URL easily (without using regular expressions).

Unfortunately under PHP 5.2.0-8 (+etch10), parse_url will fail as it does not accept the slash (/) at the beginning of URL.

Here is a possible patch :

<?php

  $p = parse_url ( 'http://domain.tld/tcp://domain2.tld/dir/file' ) ;

  $d2 = parse_url ( substr ( $p['path'] , 1 ) ) ;

  echo $d2 ;  // returns '/dir/file'

?>

However this last code is not optimized at all, and should be replaced by a regular expression to split URL (so that parse_url would be no longer used).

So you should use parse_url very carefully, and verify that you pass only standard URLs...

Nick Smith (05-Sep-2007 10:32)

Note that older versions of PHP (e.g., 4.1) returned an blank string as the path for URLs without any path, such as http://www.php.net

However more recent versions of PHP (e.g., 4.4.7) don't set the path element in the array, so old code will get a PHP warning about an undefined index.

Michael Muryn (27-Aug-2007 04:51)

Another update to the glue_url function: applied the "isset" treatment to $parsed['pass'].

<?php
function glue_url($parsed)
{
    if (!is_array($parsed)) return false;
    $uri = isset($parsed['scheme']) ? $parsed['scheme'].':'.((strtolower($parsed['scheme']) == 'mailto') ? '' : '//') : '';
    $uri .= isset($parsed['user']) ? $parsed['user'].(isset($parsed['pass']) ? ':'.$parsed['pass'] : '').'@' : '';
    $uri .= isset($parsed['host']) ? $parsed['host'] : '';
    $uri .= isset($parsed['port']) ? ':'.$parsed['port'] : '';
    if(isset($parsed['path']))
    {
        $uri .= (substr($parsed['path'], 0, 1) == '/') ? $parsed['path'] : ('/'.$parsed['path']);
    }
    $uri .= isset($parsed['query']) ? '?'.$parsed['query'] : '';
    $uri .= isset($parsed['fragment']) ? '#'.$parsed['fragment'] : '';
    return $uri;
}
?>

stevenlewis at hotmail dot com (13-Aug-2007 11:08)

an update to the glue url function.

you are able to put a host and a path without a slash at the beginning of the path

<?php
function glue_url($parsed)
    {
    if (! is_array($parsed)) return false;
    $uri = isset($parsed['scheme']) ? $parsed['scheme'].':'.((strtolower($parsed['scheme']) == 'mailto') ? '':'//'): '';
    $uri .= isset($parsed['user']) ? $parsed['user'].($parsed['pass']? ':'.$parsed['pass']:'').'@':'';
    $uri .= isset($parsed['host']) ? $parsed['host'] : '';
    $uri .= isset($parsed['port']) ? ':'.$parsed['port'] : '';
    if(isset($parsed['path']))
        {
        $uri .= (substr($parsed['path'],0,1) == '/')?$parsed['path']:'/'.$parsed['path'];
        }
    $uri .= isset($parsed['query']) ? '?'.$parsed['query'] : '';
    $uri .= isset($parsed['fragment']) ? '#'.$parsed['fragment'] : '';
    return $uri;
    }
?>

spam at paulisageek dot com (08-Aug-2007 08:05)

In reply to adrian,

Thank you very much for your function. There is a small issue with your relative protocol function. You need to remove the // when making the url the path. Here is the new function.

function resolve_url($base, $url) {
        if (!strlen($base)) return $url;
        // Step 2
        if (!strlen($url)) return $base;
        // Step 3
        if (preg_match('!^[a-z]+:!i', $url)) return $url;
        $base = parse_url($base);
        if ($url{0} == "#") {
                // Step 2 (fragment)
                $base['fragment'] = substr($url, 1);
                return unparse_url($base);
        }
        unset($base['fragment']);
        unset($base['query']);
        if (substr($url, 0, 2) == "//") {
                // Step 4
                return unparse_url(array(
                        'scheme'=>$base['scheme'],
                        'path'=>substr($url,2),
                ));
        } else if ($url{0} == "/") {
                // Step 5
                $base['path'] = $url;
        } else {
                // Step 6
                $path = explode('/', $base['path']);
                $url_path = explode('/', $url);
                // Step 6a: drop file from base
                array_pop($path);
                // Step 6b, 6c, 6e: append url while removing "." and ".." from
                // the directory portion
                $end = array_pop($url_path);
                foreach ($url_path as $segment) {
                        if ($segment == '.') {
                                // skip
                        } else if ($segment == '..' && $path && $path[sizeof($path)-1] != '..') {
                                array_pop($path);
                        } else {
                                $path[] = $segment;
                        }
                }
                // Step 6d, 6f: remove "." and ".." from file portion
                if ($end == '.') {
                        $path[] = '';
                } else if ($end == '..' && $path && $path[sizeof($path)-1] != '..') {
                        $path[sizeof($path)-1] = '';
                } else {
                        $path[] = $end;
                }
                // Step 6h
                $base['path'] = join('/', $path);

        }
        // Step 7
        return unparse_url($base);
}

christian at resource-it dot dk (03-Aug-2007 08:57)

I searched for an implementation of rfc3986, which is a newer version of rfc 2392. I may find it here : <http://www.chrsen.dk/fundanemt/files/scripter/php/misc/rfc3986.php> - read the rfc at <http://rfc.net/rfc3986.html>

adrian-php at sixfingeredman dot net (25-Jul-2007 10:58)

Here's a function which implements resolving a relative URL according to RFC 2396 section 5.2. No doubt there are more efficient implementations, but this one tries to remain close to the standard for clarity. It relies on a function called "unparse_url" to implement section 7, left as an exercise for the reader (or you can substitute the "glue_url" function posted earlier).

<?php
/**
 * Resolve a URL relative to a base path. This happens to work with POSIX
 * filenames as well. This is based on RFC 2396 section 5.2.
 */
function resolve_url($base, $url) {
        if (!strlen($base)) return $url;
        // Step 2
        if (!strlen($url)) return $base;
        // Step 3
        if (preg_match('!^[a-z]+:!i', $url)) return $url;
        $base = parse_url($base);
        if ($url{0} == "#") {
                // Step 2 (fragment)
                $base['fragment'] = substr($url, 1);
                return unparse_url($base);
        }
        unset($base['fragment']);
        unset($base['query']);
        if (substr($url, 0, 2) == "//") {
                // Step 4
                return unparse_url(array(
                        'scheme'=>$base['scheme'],
                        'path'=>$url,
                ));
        } else if ($url{0} == "/") {
                // Step 5
                $base['path'] = $url;
        } else {
                // Step 6
                $path = explode('/', $base['path']);
                $url_path = explode('/', $url);
                // Step 6a: drop file from base
                array_pop($path);
                // Step 6b, 6c, 6e: append url while removing "." and ".." from
                // the directory portion
                $end = array_pop($url_path);
                foreach ($url_path as $segment) {
                        if ($segment == '.') {
                                // skip
                        } else if ($segment == '..' && $path && $path[sizeof($path)-1] != '..') {
                                array_pop($path);
                        } else {
                                $path[] = $segment;
                        }
                }
                // Step 6d, 6f: remove "." and ".." from file portion
                if ($end == '.') {
                        $path[] = '';
                } else if ($end == '..' && $path && $path[sizeof($path)-1] != '..') {
                        $path[sizeof($path)-1] = '';
                } else {
                        $path[] = $end;
                }
                // Step 6h
                $base['path'] = join('/', $path);

        }
        // Step 7
        return unparse_url($base);
}
?>

Antti Haapala (17-Jul-2007 09:42)

Actually the behaviour noticed by the previous poster is quite correct. When the URI scheme is not present, it is plain wrong to assume that something starting with www. is a domain name, and that the scheme is HTTP. Internet Explorer does it that way, sure, but it does not make it any more correct. The documentation says that the function tries to decode the URL as well as it can, and the only sensible and standards-compliant way to decode such URL is to expect it to be a relative URI.

Elliott Brueggeman (03-Jun-2007 11:59)

Note that if you pass this function a url without a scheme (www.php.net, as opposed to http://www.php.net), the function will incorrectly parse the results. In my test case it returned the domain under the ['path'] element and nothing in the ['host'] element.

Marc-Antoine Ross (14-Mar-2007 04:10)

Do not look for the fragment in $_SERVER['QUERY_STRING'], you will not find it. You should read the fragment in JavaScript for example.

alistair at 21degrees dot com dot au (24-Oct-2006 03:21)

Heres a simple function to add the $component option in for PHP4. Haven't done exhaustive testing, but should work ok.

<?php

    ## Defines only available in PHP 5, created for PHP4
    if(!defined('PHP_URL_SCHEME')) define('PHP_URL_SCHEME', 1);
    if(!defined('PHP_URL_HOST')) define('PHP_URL_HOST', 2);
    if(!defined('PHP_URL_PORT')) define('PHP_URL_PORT', 3);
    if(!defined('PHP_URL_USER')) define('PHP_URL_USER', 4);
    if(!defined('PHP_URL_PASS')) define('PHP_URL_PASS', 5);
    if(!defined('PHP_URL_PATH')) define('PHP_URL_PATH', 6);
    if(!defined('PHP_URL_QUERY')) define('PHP_URL_QUERY', 7);                       
    if(!defined('PHP_URL_FRAGMENT')) define('PHP_URL_FRAGMENT', 8);   
   
    function parse_url_compat($url, $component=NULL){
       
        if(!$component) return parse_url($url);
       
        ## PHP 5
        if(phpversion() >= 5)
            return parse_url($url, $component);

        ## PHP 4
        $bits = parse_url($url);
       
        switch($component){
            case PHP_URL_SCHEME: return $bits['scheme'];
            case PHP_URL_HOST: return $bits['host'];
            case PHP_URL_PORT: return $bits['port'];
            case PHP_URL_USER: return $bits['user'];
            case PHP_URL_PASS: return $bits['pass'];
            case PHP_URL_PATH: return $bits['path'];
            case PHP_URL_QUERY: return $bits['query'];
            case PHP_URL_FRAGMENT: return $bits['fragment'];
        }
       
    }

?>

TheShadow (30-Dec-2004 08:36)

You may want to check out the PEAR NET_URL class. It provides easy means to manipulate URL strings.

http://pear.php.net/package/Net_URL