GlobalFunctions.php

<?php
/**
 * Global functions used everywhere.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
 * http://www.gnu.org/copyleft/gpl.html
 *
 * @file
 */

if ( !defined( 'MEDIAWIKI' ) ) {
	die( "This file is part of MediaWiki, it is not a valid entry point" );
}

use Liuggio\StatsdClient\Sender\SocketSender;
use MediaWiki\Logger\LoggerFactory;

// Hide compatibility functions from Doxygen
/// @cond

/**
 * Compatibility functions
 *
 * We support PHP 5.3.3 and up.
 * Re-implementations of newer functions or functions in non-standard
 * PHP extensions may be included here.
 */

if ( !function_exists( 'mb_substr' ) ) {
	/**
	 * @codeCoverageIgnore
	 * @see Fallback::mb_substr
	 * @return string
	 */
	function mb_substr( $str, $start, $count = 'end' ) {
		return Fallback::mb_substr( $str, $start, $count );
	}

	/**
	 * @codeCoverageIgnore
	 * @see Fallback::mb_substr_split_unicode
	 * @return int
	 */
	function mb_substr_split_unicode( $str, $splitPos ) {
		return Fallback::mb_substr_split_unicode( $str, $splitPos );
	}
}

if ( !function_exists( 'mb_strlen' ) ) {
	/**
	 * @codeCoverageIgnore
	 * @see Fallback::mb_strlen
	 * @return int
	 */
	function mb_strlen( $str, $enc = '' ) {
		return Fallback::mb_strlen( $str, $enc );
	}
}

if ( !function_exists( 'mb_strpos' ) ) {
	/**
	 * @codeCoverageIgnore
	 * @see Fallback::mb_strpos
	 * @return int
	 */
	function mb_strpos( $haystack, $needle, $offset = 0, $encoding = '' ) {
		return Fallback::mb_strpos( $haystack, $needle, $offset, $encoding );
	}
}

if ( !function_exists( 'mb_strrpos' ) ) {
	/**
	 * @codeCoverageIgnore
	 * @see Fallback::mb_strrpos
	 * @return int
	 */
	function mb_strrpos( $haystack, $needle, $offset = 0, $encoding = '' ) {
		return Fallback::mb_strrpos( $haystack, $needle, $offset, $encoding );
	}
}

// gzdecode function only exists in PHP >= 5.4.0
// http://php.net/gzdecode
if ( !function_exists( 'gzdecode' ) ) {
	/**
	 * @codeCoverageIgnore
	 * @param string $data
	 * @return string
	 */
	function gzdecode( $data ) {
		return gzinflate( substr( $data, 10, -8 ) );
	}
}

// hash_equals function only exists in PHP >= 5.6.0
// http://php.net/hash_equals
if ( !function_exists( 'hash_equals' ) ) {
	/**
	 * Check whether a user-provided string is equal to a fixed-length secret string
	 * without revealing bytes of the secret string through timing differences.
	 *
	 * The usual way to compare strings (PHP's === operator or the underlying memcmp()
	 * function in C) is to compare corresponding bytes and stop at the first difference,
	 * which would take longer for a partial match than for a complete mismatch. This
	 * is not secure when one of the strings (e.g. an HMAC or token) must remain secret
	 * and the other may come from an attacker. Statistical analysis of timing measurements
	 * over many requests may allow the attacker to guess the string's bytes one at a time
	 * (and check his guesses) even if the timing differences are extremely small.
	 *
	 * When making such a security-sensitive comparison, it is essential that the sequence
	 * in which instructions are executed and memory locations are accessed not depend on
	 * the secret string's value. HOWEVER, for simplicity, we do not attempt to minimize
	 * the inevitable leakage of the string's length. That is generally known anyway as
	 * a chararacteristic of the hash function used to compute the secret value.
	 *
	 * Longer explanation: http://www.emerose.com/timing-attacks-explained
	 *
	 * @codeCoverageIgnore
	 * @param string $known_string Fixed-length secret string to compare against
	 * @param string $user_string User-provided string
	 * @return bool True if the strings are the same, false otherwise
	 */
	function hash_equals( $known_string, $user_string ) {
		// Strict type checking as in PHP's native implementation
		if ( !is_string( $known_string ) ) {
			trigger_error( 'hash_equals(): Expected known_string to be a string, ' .
				gettype( $known_string ) . ' given', E_USER_WARNING );

			return false;
		}

		if ( !is_string( $user_string ) ) {
			trigger_error( 'hash_equals(): Expected user_string to be a string, ' .
				gettype( $user_string ) . ' given', E_USER_WARNING );

			return false;
		}

		$known_string_len = strlen( $known_string );
		if ( $known_string_len !== strlen( $user_string ) ) {
			return false;
		}

		$result = 0;
		for ( $i = 0; $i < $known_string_len; $i++ ) {
			$result |= ord( $known_string[$i] ) ^ ord( $user_string[$i] );
		}

		return ( $result === 0 );
	}
}
/// @endcond

/**
 * Load an extension
 *
 * This queues an extension to be loaded through
 * the ExtensionRegistry system.
 *
 * @param string $ext Name of the extension to load
 * @param string|null $path Absolute path of where to find the extension.json file
 * @since 1.25
 */
function wfLoadExtension( $ext, $path = null ) {
	if ( !$path ) {
		global $wgExtensionDirectory;
		$path = "$wgExtensionDirectory/$ext/extension.json";
	}
	ExtensionRegistry::getInstance()->queue( $path );
}

/**
 * Load multiple extensions at once
 *
 * Same as wfLoadExtension, but more efficient if you
 * are loading multiple extensions.
 *
 * If you want to specify custom paths, you should interact with
 * ExtensionRegistry directly.
 *
 * @see wfLoadExtension
 * @param string[] $exts Array of extension names to load
 * @since 1.25
 */
function wfLoadExtensions( array $exts ) {
	global $wgExtensionDirectory;
	$registry = ExtensionRegistry::getInstance();
	foreach ( $exts as $ext ) {
		$registry->queue( "$wgExtensionDirectory/$ext/extension.json" );
	}
}

/**
 * Load a skin
 *
 * @see wfLoadExtension
 * @param string $skin Name of the extension to load
 * @param string|null $path Absolute path of where to find the skin.json file
 * @since 1.25
 */
function wfLoadSkin( $skin, $path = null ) {
	if ( !$path ) {
		global $wgStyleDirectory;
		$path = "$wgStyleDirectory/$skin/skin.json";
	}
	ExtensionRegistry::getInstance()->queue( $path );
}

/**
 * Load multiple skins at once
 *
 * @see wfLoadExtensions
 * @param string[] $skins Array of extension names to load
 * @since 1.25
 */
function wfLoadSkins( array $skins ) {
	global $wgStyleDirectory;
	$registry = ExtensionRegistry::getInstance();
	foreach ( $skins as $skin ) {
		$registry->queue( "$wgStyleDirectory/$skin/skin.json" );
	}
}

/**
 * Like array_diff( $a, $b ) except that it works with two-dimensional arrays.
 * @param array $a
 * @param array $b
 * @return array
 */
function wfArrayDiff2( $a, $b ) {
	return array_udiff( $a, $b, 'wfArrayDiff2_cmp' );
}

/**
 * @param array|string $a
 * @param array|string $b
 * @return int
 */
function wfArrayDiff2_cmp( $a, $b ) {
	if ( is_string( $a ) && is_string( $b ) ) {
		return strcmp( $a, $b );
	} elseif ( count( $a ) !== count( $b ) ) {
		return count( $a ) < count( $b ) ? -1 : 1;
	} else {
		reset( $a );
		reset( $b );
		while ( ( list( , $valueA ) = each( $a ) ) && ( list( , $valueB ) = each( $b ) ) ) {
			$cmp = strcmp( $valueA, $valueB );
			if ( $cmp !== 0 ) {
				return $cmp;
			}
		}
		return 0;
	}
}

/**
 * Appends to second array if $value differs from that in $default
 *
 * @param string|int $key
 * @param mixed $value
 * @param mixed $default
 * @param array $changed Array to alter
 * @throws MWException
 */
function wfAppendToArrayIfNotDefault( $key, $value, $default, &$changed ) {
	if ( is_null( $changed ) ) {
		throw new MWException( 'GlobalFunctions::wfAppendToArrayIfNotDefault got null' );
	}
	if ( $default[$key] !== $value ) {
		$changed[$key] = $value;
	}
}

/**
 * Merge arrays in the style of getUserPermissionsErrors, with duplicate removal
 * e.g.
 *	wfMergeErrorArrays(
 *		array( array( 'x' ) ),
 *		array( array( 'x', '2' ) ),
 *		array( array( 'x' ) ),
 *		array( array( 'y' ) )
 *	);
 * returns:
 * 		array(
 *   		array( 'x', '2' ),
 *   		array( 'x' ),
 *   		array( 'y' )
 *   	)
 *
 * @param array $array1,...
 * @return array
 */
function wfMergeErrorArrays( /*...*/ ) {
	$args = func_get_args();
	$out = array();
	foreach ( $args as $errors ) {
		foreach ( $errors as $params ) {
			# @todo FIXME: Sometimes get nested arrays for $params,
			# which leads to E_NOTICEs
			$spec = implode( "\t", $params );
			$out[$spec] = $params;
		}
	}
	return array_values( $out );
}

/**
 * Insert array into another array after the specified *KEY*
 *
 * @param array $array The array.
 * @param array $insert The array to insert.
 * @param mixed $after The key to insert after
 * @return array
 */
function wfArrayInsertAfter( array $array, array $insert, $after ) {
	// Find the offset of the element to insert after.
	$keys = array_keys( $array );
	$offsetByKey = array_flip( $keys );

	$offset = $offsetByKey[$after];

	// Insert at the specified offset
	$before = array_slice( $array, 0, $offset + 1, true );
	$after = array_slice( $array, $offset + 1, count( $array ) - $offset, true );

	$output = $before + $insert + $after;

	return $output;
}

/**
 * Recursively converts the parameter (an object) to an array with the same data
 *
 * @param object|array $objOrArray
 * @param bool $recursive
 * @return array
 */
function wfObjectToArray( $objOrArray, $recursive = true ) {
	$array = array();
	if ( is_object( $objOrArray ) ) {
		$objOrArray = get_object_vars( $objOrArray );
	}
	foreach ( $objOrArray as $key => $value ) {
		if ( $recursive && ( is_object( $value ) || is_array( $value ) ) ) {
			$value = wfObjectToArray( $value );
		}

		$array[$key] = $value;
	}

	return $array;
}

/**
 * Get a random decimal value between 0 and 1, in a way
 * not likely to give duplicate values for any realistic
 * number of articles.
 *
 * @return string
 */
function wfRandom() {
	# The maximum random value is "only" 2^31-1, so get two random
	# values to reduce the chance of dupes
	$max = mt_getrandmax() + 1;
	$rand = number_format( ( mt_rand() * $max + mt_rand() ) / $max / $max, 12, '.', '' );

	return $rand;
}

/**
 * Get a random string containing a number of pseudo-random hex
 * characters.
 * @note This is not secure, if you are trying to generate some sort
 *       of token please use MWCryptRand instead.
 *
 * @param int $length The length of the string to generate
 * @return string
 * @since 1.20
 */
function wfRandomString( $length = 32 ) {
	$str = '';
	for ( $n = 0; $n < $length; $n += 7 ) {
		$str .= sprintf( '%07x', mt_rand() & 0xfffffff );
	}
	return substr( $str, 0, $length );
}

/**
 * We want some things to be included as literal characters in our title URLs
 * for prettiness, which urlencode encodes by default.  According to RFC 1738,
 * all of the following should be safe:
 *
 * ;:@&=$-_.+!*'(),
 *
 * RFC 1738 says ~ is unsafe, however RFC 3986 considers it an unreserved
 * character which should not be encoded. More importantly, google chrome
 * always converts %7E back to ~, and converting it in this function can
 * cause a redirect loop (T105265).
 *
 * But + is not safe because it's used to indicate a space; &= are only safe in
 * paths and not in queries (and we don't distinguish here); ' seems kind of
 * scary; and urlencode() doesn't touch -_. to begin with.  Plus, although /
 * is reserved, we don't care.  So the list we unescape is:
 *
 * ;:@$!*(),/~
 *
 * However, IIS7 redirects fail when the url contains a colon (Bug 22709),
 * so no fancy : for IIS7.
 *
 * %2F in the page titles seems to fatally break for some reason.
 *
 * @param string $s
 * @return string
 */
function wfUrlencode( $s ) {
	static $needle;

	if ( is_null( $s ) ) {
		$needle = null;
		return '';
	}

	if ( is_null( $needle ) ) {
		$needle = array( '%3B', '%40', '%24', '%21', '%2A', '%28', '%29', '%2C', '%2F', '%7E' );
		if ( !isset( $_SERVER['SERVER_SOFTWARE'] ) ||
			( strpos( $_SERVER['SERVER_SOFTWARE'], 'Microsoft-IIS/7' ) === false )
		) {
			$needle[] = '%3A';
		}
	}

	$s = urlencode( $s );
	$s = str_ireplace(
		$needle,
		array( ';', '@', '$', '!', '*', '(', ')', ',', '/', '~', ':' ),
		$s
	);

	return $s;
}

/**
 * This function takes two arrays as input, and returns a CGI-style string, e.g.
 * "days=7&limit=100". Options in the first array override options in the second.
 * Options set to null or false will not be output.
 *
 * @param array $array1 ( String|Array )
 * @param array $array2 ( String|Array )
 * @param string $prefix
 * @return string
 */
function wfArrayToCgi( $array1, $array2 = null, $prefix = '' ) {
	if ( !is_null( $array2 ) ) {
		$array1 = $array1 + $array2;
	}

	$cgi = '';
	foreach ( $array1 as $key => $value ) {
		if ( !is_null( $value ) && $value !== false ) {
			if ( $cgi != '' ) {
				$cgi .= '&';
			}
			if ( $prefix !== '' ) {
				$key = $prefix . "[$key]";
			}
			if ( is_array( $value ) ) {
				$firstTime = true;
				foreach ( $value as $k => $v ) {
					$cgi .= $firstTime ? '' : '&';
					if ( is_array( $v ) ) {
						$cgi .= wfArrayToCgi( $v, null, $key . "[$k]" );
					} else {
						$cgi .= urlencode( $key . "[$k]" ) . '=' . urlencode( $v );
					}
					$firstTime = false;
				}
			} else {
				if ( is_object( $value ) ) {
					$value = $value->__toString();
				}
				$cgi .= urlencode( $key ) . '=' . urlencode( $value );
			}
		}
	}
	return $cgi;
}

/**
 * This is the logical opposite of wfArrayToCgi(): it accepts a query string as
 * its argument and returns the same string in array form.  This allows compatibility
 * with legacy functions that accept raw query strings instead of nice
 * arrays.  Of course, keys and values are urldecode()d.
 *
 * @param string $query Query string
 * @return string[] Array version of input
 */
function wfCgiToArray( $query ) {
	if ( isset( $query[0] ) && $query[0] == '?' ) {
		$query = substr( $query, 1 );
	}
	$bits = explode( '&', $query );
	$ret = array();
	foreach ( $bits as $bit ) {
		if ( $bit === '' ) {
			continue;
		}
		if ( strpos( $bit, '=' ) === false ) {
			// Pieces like &qwerty become 'qwerty' => '' (at least this is what php does)
			$key = $bit;
			$value = '';
		} else {
			list( $key, $value ) = explode( '=', $bit );
		}
		$key = urldecode( $key );
		$value = urldecode( $value );
		if ( strpos( $key, '[' ) !== false ) {
			$keys = array_reverse( explode( '[', $key ) );
			$key = array_pop( $keys );
			$temp = $value;
			foreach ( $keys as $k ) {
				$k = substr( $k, 0, -1 );
				$temp = array( $k => $temp );
			}
			if ( isset( $ret[$key] ) ) {
				$ret[$key] = array_merge( $ret[$key], $temp );
			} else {
				$ret[$key] = $temp;
			}
		} else {
			$ret[$key] = $value;
		}
	}
	return $ret;
}

/**
 * Append a query string to an existing URL, which may or may not already
 * have query string parameters already. If so, they will be combined.
 *
 * @param string $url
 * @param string|string[] $query String or associative array
 * @return string
 */
function wfAppendQuery( $url, $query ) {
	if ( is_array( $query ) ) {
		$query = wfArrayToCgi( $query );
	}
	if ( $query != '' ) {
		if ( false === strpos( $url, '?' ) ) {
			$url .= '?';
		} else {
			$url .= '&';
		}
		$url .= $query;
	}
	return $url;
}

/**
 * Expand a potentially local URL to a fully-qualified URL. Assumes $wgServer
 * is correct.
 *
 * The meaning of the PROTO_* constants is as follows:
 * PROTO_HTTP: Output a URL starting with http://
 * PROTO_HTTPS: Output a URL starting with https://
 * PROTO_RELATIVE: Output a URL starting with // (protocol-relative URL)
 * PROTO_CURRENT: Output a URL starting with either http:// or https:// , depending
 *    on which protocol was used for the current incoming request
 * PROTO_CANONICAL: For URLs without a domain, like /w/index.php , use $wgCanonicalServer.
 *    For protocol-relative URLs, use the protocol of $wgCanonicalServer
 * PROTO_INTERNAL: Like PROTO_CANONICAL, but uses $wgInternalServer instead of $wgCanonicalServer
 *
 * @todo this won't work with current-path-relative URLs
 * like "subdir/foo.html", etc.
 *
 * @param string $url Either fully-qualified or a local path + query
 * @param string $defaultProto One of the PROTO_* constants. Determines the
 *    protocol to use if $url or $wgServer is protocol-relative
 * @return string Fully-qualified URL, current-path-relative URL or false if
 *    no valid URL can be constructed
 */
function wfExpandUrl( $url, $defaultProto = PROTO_CURRENT ) {
	global $wgServer, $wgCanonicalServer, $wgInternalServer, $wgRequest,
		$wgHttpsPort;
	if ( $defaultProto === PROTO_CANONICAL ) {
		$serverUrl = $wgCanonicalServer;
	} elseif ( $defaultProto === PROTO_INTERNAL && $wgInternalServer !== false ) {
		// Make $wgInternalServer fall back to $wgServer if not set
		$serverUrl = $wgInternalServer;
	} else {
		$serverUrl = $wgServer;
		if ( $defaultProto === PROTO_CURRENT ) {
			$defaultProto = $wgRequest->getProtocol() . '://';
		}
	}

	// Analyze $serverUrl to obtain its protocol
	$bits = wfParseUrl( $serverUrl );
	$serverHasProto = $bits && $bits['scheme'] != '';

	if ( $defaultProto === PROTO_CANONICAL || $defaultProto === PROTO_INTERNAL ) {
		if ( $serverHasProto ) {
			$defaultProto = $bits['scheme'] . '://';
		} else {
			// $wgCanonicalServer or $wgInternalServer doesn't have a protocol.
			// This really isn't supposed to happen. Fall back to HTTP in this
			// ridiculous case.
			$defaultProto = PROTO_HTTP;
		}
	}

	$defaultProtoWithoutSlashes = substr( $defaultProto, 0, -2 );

	if ( substr( $url, 0, 2 ) == '//' ) {
		$url = $defaultProtoWithoutSlashes . $url;
	} elseif ( substr( $url, 0, 1 ) == '/' ) {
		// If $serverUrl is protocol-relative, prepend $defaultProtoWithoutSlashes,
		// otherwise leave it alone.
		$url = ( $serverHasProto ? '' : $defaultProtoWithoutSlashes ) . $serverUrl . $url;
	}

	$bits = wfParseUrl( $url );

	// ensure proper port for HTTPS arrives in URL
	// https://bugzilla.wikimedia.org/show_bug.cgi?id=65184
	if ( $defaultProto === PROTO_HTTPS && $wgHttpsPort != 443 ) {
		$bits['port'] = $wgHttpsPort;
	}

	if ( $bits && isset( $bits['path'] ) ) {
		$bits['path'] = wfRemoveDotSegments( $bits['path'] );
		return wfAssembleUrl( $bits );
	} elseif ( $bits ) {
		# No path to expand
		return $url;
	} elseif ( substr( $url, 0, 1 ) != '/' ) {
		# URL is a relative path
		return wfRemoveDotSegments( $url );
	}

	# Expanded URL is not valid.
	return false;
}

/**
 * This function will reassemble a URL parsed with wfParseURL.  This is useful
 * if you need to edit part of a URL and put it back together.
 *
 * This is the basic structure used (brackets contain keys for $urlParts):
 * [scheme][delimiter][user]:[pass]@[host]:[port][path]?[query]#[fragment]
 *
 * @todo Need to integrate this into wfExpandUrl (bug 32168)
 *
 * @since 1.19
 * @param array $urlParts URL parts, as output from wfParseUrl
 * @return string URL assembled from its component parts
 */
function wfAssembleUrl( $urlParts ) {
	$result = '';

	if ( isset( $urlParts['delimiter'] ) ) {
		if ( isset( $urlParts['scheme'] ) ) {
			$result .= $urlParts['scheme'];
		}

		$result .= $urlParts['delimiter'];
	}

	if ( isset( $urlParts['host'] ) ) {
		if ( isset( $urlParts['user'] ) ) {
			$result .= $urlParts['user'];
			if ( isset( $urlParts['pass'] ) ) {
				$result .= ':' . $urlParts['pass'];
			}
			$result .= '@';
		}

		$result .= $urlParts['host'];

		if ( isset( $urlParts['port'] ) ) {
			$result .= ':' . $urlParts['port'];
		}
	}

	if ( isset( $urlParts['path'] ) ) {
		$result .= $urlParts['path'];
	}

	if ( isset( $urlParts['query'] ) ) {
		$result .= '?' . $urlParts['query'];
	}

	if ( isset( $urlParts['fragment'] ) ) {
		$result .= '#' . $urlParts['fragment'];
	}

	return $result;
}

/**
 * Remove all dot-segments in the provided URL path.  For example,
 * '/a/./b/../c/' becomes '/a/c/'.  For details on the algorithm, please see
 * RFC3986 section 5.2.4.
 *
 * @todo Need to integrate this into wfExpandUrl (bug 32168)
 *
 * @param string $urlPath URL path, potentially containing dot-segments
 * @return string URL path with all dot-segments removed
 */
function wfRemoveDotSegments( $urlPath ) {
	$output = '';
	$inputOffset = 0;
	$inputLength = strlen( $urlPath );

	while ( $inputOffset < $inputLength ) {
		$prefixLengthOne = substr( $urlPath, $inputOffset, 1 );
		$prefixLengthTwo = substr( $urlPath, $inputOffset, 2 );
		$prefixLengthThree = substr( $urlPath, $inputOffset, 3 );
		$prefixLengthFour = substr( $urlPath, $inputOffset, 4 );
		$trimOutput = false;

		if ( $prefixLengthTwo == './' ) {
			# Step A, remove leading "./"
			$inputOffset += 2;
		} elseif ( $prefixLengthThree == '../' ) {
			# Step A, remove leading "../"
			$inputOffset += 3;
		} elseif ( ( $prefixLengthTwo == '/.' ) && ( $inputOffset + 2 == $inputLength ) ) {
			# Step B, replace leading "/.$" with "/"
			$inputOffset += 1;
			$urlPath[$inputOffset] = '/';
		} elseif ( $prefixLengthThree == '/./' ) {
			# Step B, replace leading "/./" with "/"
			$inputOffset += 2;
		} elseif ( $prefixLengthThree == '/..' && ( $inputOffset + 3 == $inputLength ) ) {
			# Step C, replace leading "/..$" with "/" and
			# remove last path component in output
			$inputOffset += 2;
			$urlPath[$inputOffset] = '/';
			$trimOutput = true;
		} elseif ( $prefixLengthFour == '/../' ) {
			# Step C, replace leading "/../" with "/" and
			# remove last path component in output
			$inputOffset += 3;
			$trimOutput = true;
		} elseif ( ( $prefixLengthOne == '.' ) && ( $inputOffset + 1 == $inputLength ) ) {
			# Step D, remove "^.$"
			$inputOffset += 1;
		} elseif ( ( $prefixLengthTwo == '..' ) && ( $inputOffset + 2 == $inputLength ) ) {
			# Step D, remove "^..$"
			$inputOffset += 2;
		} else {
			# Step E, move leading path segment to output
			if ( $prefixLengthOne == '/' ) {
				$slashPos = strpos( $urlPath, '/', $inputOffset + 1 );
			} else {
				$slashPos = strpos( $urlPath, '/', $inputOffset );
			}
			if ( $slashPos === false ) {
				$output .= substr( $urlPath, $inputOffset );
				$inputOffset = $inputLength;
			} else {
				$output .= substr( $urlPath, $inputOffset, $slashPos - $inputOffset );
				$inputOffset += $slashPos - $inputOffset;
			}
		}

		if ( $trimOutput ) {
			$slashPos = strrpos( $output, '/' );
			if ( $slashPos === false ) {
				$output = '';
			} else {
				$output = substr( $output, 0, $slashPos );
			}
		}
	}

	return $output;
}

/**
 * Returns a regular expression of url protocols
 *
 * @param bool $includeProtocolRelative If false, remove '//' from the returned protocol list.
 *        DO NOT USE this directly, use wfUrlProtocolsWithoutProtRel() instead
 * @return string
 */
function wfUrlProtocols( $includeProtocolRelative = true ) {
	global $wgUrlProtocols;

	// Cache return values separately based on $includeProtocolRelative
	static $withProtRel = null, $withoutProtRel = null;
	$cachedValue = $includeProtocolRelative ? $withProtRel : $withoutProtRel;
	if ( !is_null( $cachedValue ) ) {
		return $cachedValue;
	}

	// Support old-style $wgUrlProtocols strings, for backwards compatibility
	// with LocalSettings files from 1.5
	if ( is_array( $wgUrlProtocols ) ) {
		$protocols = array();
		foreach ( $wgUrlProtocols as $protocol ) {
			// Filter out '//' if !$includeProtocolRelative
			if ( $includeProtocolRelative || $protocol !== '//' ) {
				$protocols[] = preg_quote( $protocol, '/' );
			}
		}

		$retval = implode( '|', $protocols );
	} else {
		// Ignore $includeProtocolRelative in this case
		// This case exists for pre-1.6 compatibility, and we can safely assume
		// that '//' won't appear in a pre-1.6 config because protocol-relative
		// URLs weren't supported until 1.18
		$retval = $wgUrlProtocols;
	}

	// Cache return value
	if ( $includeProtocolRelative ) {
		$withProtRel = $retval;
	} else {
		$withoutProtRel = $retval;
	}
	return $retval;
}

/**
 * Like wfUrlProtocols(), but excludes '//' from the protocol list. Use this if
 * you need a regex that matches all URL protocols but does not match protocol-
 * relative URLs
 * @return string
 */
function wfUrlProtocolsWithoutProtRel() {
	return wfUrlProtocols( false );
}

/**
 * parse_url() work-alike, but non-broken.  Differences:
 *
 * 1) Does not raise warnings on bad URLs (just returns false).
 * 2) Handles protocols that don't use :// (e.g., mailto: and news:, as well as
 *    protocol-relative URLs) correctly.
 * 3) Adds a "delimiter" element to the array, either '://', ':' or '//' (see (2)).
 *
 * @param string $url A URL to parse
 * @return string[] Bits of the URL in an associative array, per PHP docs
 */
function wfParseUrl( $url ) {
	global $wgUrlProtocols; // Allow all protocols defined in DefaultSettings/LocalSettings.php

	// Protocol-relative URLs are handled really badly by parse_url(). It's so
	// bad that the easiest way to handle them is to just prepend 'http:' and
	// strip the protocol out later.
	$wasRelative = substr( $url, 0, 2 ) == '//';
	if ( $wasRelative ) {
		$url = "http:$url";
	}
	MediaWiki\suppressWarnings();
	$bits = parse_url( $url );
	MediaWiki\restoreWarnings();
	// parse_url() returns an array without scheme for some invalid URLs, e.g.
	// parse_url("%0Ahttp://example.com") == array( 'host' => '%0Ahttp', 'path' => 'example.com' )
	if ( !$bits || !isset( $bits['scheme'] ) ) {
		return false;
	}

	// parse_url() incorrectly handles schemes case-sensitively. Convert it to lowercase.
	$bits['scheme'] = strtolower( $bits['scheme'] );

	// most of the protocols are followed by ://, but mailto: and sometimes news: not, check for it
	if ( in_array( $bits['scheme'] . '://', $wgUrlProtocols ) ) {
		$bits['delimiter'] = '://';
	} elseif ( in_array( $bits['scheme'] . ':', $wgUrlProtocols ) ) {
		$bits['delimiter'] = ':';
		// parse_url detects for news: and mailto: the host part of an url as path
		// We have to correct this wrong detection
		if ( isset( $bits['path'] ) ) {
			$bits['host'] = $bits['path'];
			$bits['path'] = '';
		}
	} else {
		return false;
	}

	/* Provide an empty host for eg. file:/// urls (see bug 28627) */
	if ( !isset( $bits['host'] ) ) {
		$bits['host'] = '';

		// bug 45069
		if ( isset( $bits['path'] ) ) {
			/* parse_url loses the third / for file:///c:/ urls (but not on variants) */
			if ( substr( $bits['path'], 0, 1 ) !== '/' ) {
				$bits['path'] = '/' . $bits['path'];
			}
		} else {
			$bits['path'] = '';
		}
	}

	// If the URL was protocol-relative, fix scheme and delimiter
	if ( $wasRelative ) {
		$bits['scheme'] = '';
		$bits['delimiter'] = '//';
	}
	return $bits;
}

/**
 * Take a URL, make sure it's expanded to fully qualified, and replace any
 * encoded non-ASCII Unicode characters with their UTF-8 original forms
 * for more compact display and legibility for local audiences.
 *
 * @todo handle punycode domains too
 *
 * @param string $url
 * @return string
 */
function wfExpandIRI( $url ) {
	return preg_replace_callback(
		'/((?:%[89A-F][0-9A-F])+)/i',
		'wfExpandIRI_callback',
		wfExpandUrl( $url )
	);
}

/**
 * Private callback for wfExpandIRI
 * @param array $matches
 * @return string
 */
function wfExpandIRI_callback( $matches ) {
	return urldecode( $matches[1] );
}

/**
 * Make URL indexes, appropriate for the el_index field of externallinks.
 *
 * @param string $url
 * @return array
 */
function wfMakeUrlIndexes( $url ) {
	$bits = wfParseUrl( $url );

	// Reverse the labels in the hostname, convert to lower case
	// For emails reverse domainpart only
	if ( $bits['scheme'] == 'mailto' ) {
		$mailparts = explode( '@', $bits['host'], 2 );
		if ( count( $mailparts ) === 2 ) {
			$domainpart = strtolower( implode( '.', array_reverse( explode( '.', $mailparts[1] ) ) ) );
		} else {
			// No domain specified, don't mangle it
			$domainpart = '';
		}
		$reversedHost = $domainpart . '@' . $mailparts[0];
	} else {
		$reversedHost = strtolower( implode( '.', array_reverse( explode( '.', $bits['host'] ) ) ) );
	}
	// Add an extra dot to the end
	// Why? Is it in wrong place in mailto links?
	if ( substr( $reversedHost, -1, 1 ) !== '.' ) {
		$reversedHost .= '.';
	}
	// Reconstruct the pseudo-URL
	$prot = $bits['scheme'];
	$index = $prot . $bits['delimiter'] . $reversedHost;
	// Leave out user and password. Add the port, path, query and fragment
	if ( isset( $bits['port'] ) ) {
		$index .= ':' . $bits['port'];
	}
	if ( isset( $bits['path'] ) ) {
		$index .= $bits['path'];
	} else {
		$index .= '/';
	}
	if ( isset( $bits['query'] ) ) {
		$index .= '?' . $bits['query'];
	}
	if ( isset( $bits['fragment'] ) ) {
		$index .= '#' . $bits['fragment'];
	}

	if ( $prot == '' ) {
		return array( "http:$index", "https:$index" );
	} else {
		return array( $index );
	}
}