Chris@0: <?php
Chris@0: namespace GuzzleHttp\Psr7;
Chris@0: 
Chris@0: use Psr\Http\Message\UriInterface;
Chris@0: 
Chris@0: /**
Chris@0:  * Provides methods to normalize and compare URIs.
Chris@0:  *
Chris@0:  * @author Tobias Schultze
Chris@0:  *
Chris@0:  * @link https://tools.ietf.org/html/rfc3986#section-6
Chris@0:  */
Chris@0: final class UriNormalizer
Chris@0: {
Chris@0:     /**
Chris@0:      * Default normalizations which only include the ones that preserve semantics.
Chris@0:      *
Chris@0:      * self::CAPITALIZE_PERCENT_ENCODING | self::DECODE_UNRESERVED_CHARACTERS | self::CONVERT_EMPTY_PATH |
Chris@0:      * self::REMOVE_DEFAULT_HOST | self::REMOVE_DEFAULT_PORT | self::REMOVE_DOT_SEGMENTS
Chris@0:      */
Chris@0:     const PRESERVING_NORMALIZATIONS = 63;
Chris@0: 
Chris@0:     /**
Chris@0:      * All letters within a percent-encoding triplet (e.g., "%3A") are case-insensitive, and should be capitalized.
Chris@0:      *
Chris@0:      * Example: http://example.org/a%c2%b1b → http://example.org/a%C2%B1b
Chris@0:      */
Chris@0:     const CAPITALIZE_PERCENT_ENCODING = 1;
Chris@0: 
Chris@0:     /**
Chris@0:      * Decodes percent-encoded octets of unreserved characters.
Chris@0:      *
Chris@0:      * For consistency, percent-encoded octets in the ranges of ALPHA (%41–%5A and %61–%7A), DIGIT (%30–%39),
Chris@0:      * hyphen (%2D), period (%2E), underscore (%5F), or tilde (%7E) should not be created by URI producers and,
Chris@0:      * when found in a URI, should be decoded to their corresponding unreserved characters by URI normalizers.
Chris@0:      *
Chris@0:      * Example: http://example.org/%7Eusern%61me/ → http://example.org/~username/
Chris@0:      */
Chris@0:     const DECODE_UNRESERVED_CHARACTERS = 2;
Chris@0: 
Chris@0:     /**
Chris@0:      * Converts the empty path to "/" for http and https URIs.
Chris@0:      *
Chris@0:      * Example: http://example.org → http://example.org/
Chris@0:      */
Chris@0:     const CONVERT_EMPTY_PATH = 4;
Chris@0: 
Chris@0:     /**
Chris@0:      * Removes the default host of the given URI scheme from the URI.
Chris@0:      *
Chris@0:      * Only the "file" scheme defines the default host "localhost".
Chris@0:      * All of `file:/myfile`, `file:///myfile`, and `file://localhost/myfile`
Chris@0:      * are equivalent according to RFC 3986. The first format is not accepted
Chris@0:      * by PHPs stream functions and thus already normalized implicitly to the
Chris@0:      * second format in the Uri class. See `GuzzleHttp\Psr7\Uri::composeComponents`.
Chris@0:      *
Chris@0:      * Example: file://localhost/myfile → file:///myfile
Chris@0:      */
Chris@0:     const REMOVE_DEFAULT_HOST = 8;
Chris@0: 
Chris@0:     /**
Chris@0:      * Removes the default port of the given URI scheme from the URI.
Chris@0:      *
Chris@0:      * Example: http://example.org:80/ → http://example.org/
Chris@0:      */
Chris@0:     const REMOVE_DEFAULT_PORT = 16;
Chris@0: 
Chris@0:     /**
Chris@0:      * Removes unnecessary dot-segments.
Chris@0:      *
Chris@0:      * Dot-segments in relative-path references are not removed as it would
Chris@0:      * change the semantics of the URI reference.
Chris@0:      *
Chris@0:      * Example: http://example.org/../a/b/../c/./d.html → http://example.org/a/c/d.html
Chris@0:      */
Chris@0:     const REMOVE_DOT_SEGMENTS = 32;
Chris@0: 
Chris@0:     /**
Chris@0:      * Paths which include two or more adjacent slashes are converted to one.
Chris@0:      *
Chris@0:      * Webservers usually ignore duplicate slashes and treat those URIs equivalent.
Chris@0:      * But in theory those URIs do not need to be equivalent. So this normalization
Chris@0:      * may change the semantics. Encoded slashes (%2F) are not removed.
Chris@0:      *
Chris@0:      * Example: http://example.org//foo///bar.html → http://example.org/foo/bar.html
Chris@0:      */
Chris@0:     const REMOVE_DUPLICATE_SLASHES = 64;
Chris@0: 
Chris@0:     /**
Chris@0:      * Sort query parameters with their values in alphabetical order.
Chris@0:      *
Chris@0:      * However, the order of parameters in a URI may be significant (this is not defined by the standard).
Chris@0:      * So this normalization is not safe and may change the semantics of the URI.
Chris@0:      *
Chris@0:      * Example: ?lang=en&article=fred → ?article=fred&lang=en
Chris@0:      *
Chris@0:      * Note: The sorting is neither locale nor Unicode aware (the URI query does not get decoded at all) as the
Chris@0:      * purpose is to be able to compare URIs in a reproducible way, not to have the params sorted perfectly.
Chris@0:      */
Chris@0:     const SORT_QUERY_PARAMETERS = 128;
Chris@0: 
Chris@0:     /**
Chris@0:      * Returns a normalized URI.
Chris@0:      *
Chris@0:      * The scheme and host component are already normalized to lowercase per PSR-7 UriInterface.
Chris@0:      * This methods adds additional normalizations that can be configured with the $flags parameter.
Chris@0:      *
Chris@0:      * PSR-7 UriInterface cannot distinguish between an empty component and a missing component as
Chris@0:      * getQuery(), getFragment() etc. always return a string. This means the URIs "/?#" and "/" are
Chris@0:      * treated equivalent which is not necessarily true according to RFC 3986. But that difference
Chris@0:      * is highly uncommon in reality. So this potential normalization is implied in PSR-7 as well.
Chris@0:      *
Chris@0:      * @param UriInterface $uri   The URI to normalize
Chris@0:      * @param int          $flags A bitmask of normalizations to apply, see constants
Chris@0:      *
Chris@0:      * @return UriInterface The normalized URI
Chris@0:      * @link https://tools.ietf.org/html/rfc3986#section-6.2
Chris@0:      */
Chris@0:     public static function normalize(UriInterface $uri, $flags = self::PRESERVING_NORMALIZATIONS)
Chris@0:     {
Chris@0:         if ($flags & self::CAPITALIZE_PERCENT_ENCODING) {
Chris@0:             $uri = self::capitalizePercentEncoding($uri);
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::DECODE_UNRESERVED_CHARACTERS) {
Chris@0:             $uri = self::decodeUnreservedCharacters($uri);
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::CONVERT_EMPTY_PATH && $uri->getPath() === '' &&
Chris@0:             ($uri->getScheme() === 'http' || $uri->getScheme() === 'https')
Chris@0:         ) {
Chris@0:             $uri = $uri->withPath('/');
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::REMOVE_DEFAULT_HOST && $uri->getScheme() === 'file' && $uri->getHost() === 'localhost') {
Chris@0:             $uri = $uri->withHost('');
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::REMOVE_DEFAULT_PORT && $uri->getPort() !== null && Uri::isDefaultPort($uri)) {
Chris@0:             $uri = $uri->withPort(null);
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::REMOVE_DOT_SEGMENTS && !Uri::isRelativePathReference($uri)) {
Chris@0:             $uri = $uri->withPath(UriResolver::removeDotSegments($uri->getPath()));
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::REMOVE_DUPLICATE_SLASHES) {
Chris@0:             $uri = $uri->withPath(preg_replace('#//++#', '/', $uri->getPath()));
Chris@0:         }
Chris@0: 
Chris@0:         if ($flags & self::SORT_QUERY_PARAMETERS && $uri->getQuery() !== '') {
Chris@0:             $queryKeyValues = explode('&', $uri->getQuery());
Chris@0:             sort($queryKeyValues);
Chris@0:             $uri = $uri->withQuery(implode('&', $queryKeyValues));
Chris@0:         }
Chris@0: 
Chris@0:         return $uri;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Whether two URIs can be considered equivalent.
Chris@0:      *
Chris@0:      * Both URIs are normalized automatically before comparison with the given $normalizations bitmask. The method also
Chris@0:      * accepts relative URI references and returns true when they are equivalent. This of course assumes they will be
Chris@0:      * resolved against the same base URI. If this is not the case, determination of equivalence or difference of
Chris@0:      * relative references does not mean anything.
Chris@0:      *
Chris@0:      * @param UriInterface $uri1           An URI to compare
Chris@0:      * @param UriInterface $uri2           An URI to compare
Chris@0:      * @param int          $normalizations A bitmask of normalizations to apply, see constants
Chris@0:      *
Chris@0:      * @return bool
Chris@0:      * @link https://tools.ietf.org/html/rfc3986#section-6.1
Chris@0:      */
Chris@0:     public static function isEquivalent(UriInterface $uri1, UriInterface $uri2, $normalizations = self::PRESERVING_NORMALIZATIONS)
Chris@0:     {
Chris@0:         return (string) self::normalize($uri1, $normalizations) === (string) self::normalize($uri2, $normalizations);
Chris@0:     }
Chris@0: 
Chris@0:     private static function capitalizePercentEncoding(UriInterface $uri)
Chris@0:     {
Chris@0:         $regex = '/(?:%[A-Fa-f0-9]{2})++/';
Chris@0: 
Chris@0:         $callback = function (array $match) {
Chris@0:             return strtoupper($match[0]);
Chris@0:         };
Chris@0: 
Chris@0:         return
Chris@0:             $uri->withPath(
Chris@0:                 preg_replace_callback($regex, $callback, $uri->getPath())
Chris@0:             )->withQuery(
Chris@0:                 preg_replace_callback($regex, $callback, $uri->getQuery())
Chris@0:             );
Chris@0:     }
Chris@0: 
Chris@0:     private static function decodeUnreservedCharacters(UriInterface $uri)
Chris@0:     {
Chris@0:         $regex = '/%(?:2D|2E|5F|7E|3[0-9]|[46][1-9A-F]|[57][0-9A])/i';
Chris@0: 
Chris@0:         $callback = function (array $match) {
Chris@0:             return rawurldecode($match[0]);
Chris@0:         };
Chris@0: 
Chris@0:         return
Chris@0:             $uri->withPath(
Chris@0:                 preg_replace_callback($regex, $callback, $uri->getPath())
Chris@0:             )->withQuery(
Chris@0:                 preg_replace_callback($regex, $callback, $uri->getQuery())
Chris@0:             );
Chris@0:     }
Chris@0: 
Chris@0:     private function __construct()
Chris@0:     {
Chris@0:         // cannot be instantiated
Chris@0:     }
Chris@0: }