3 * Function that's useful when constructing the URI string -- we frequently encounter the pattern
4 * of having to add something to the URI as we go, but only if it's present, and to include a
5 * character before or after if so.
9 * @param {string|undefined} pre To prepend
10 * @param {string} val To include
11 * @param {string} post To append
12 * @param {boolean} raw If true, val will not be encoded
13 * @return {string} Result
15 function cat( pre
, val
, post
, raw
) {
16 if ( val
=== undefined || val
=== null || val
=== '' ) {
20 return pre
+ ( raw
? val
: mw
.Uri
.encode( val
) ) + post
;
24 * Regular expressions to parse many common URIs.
26 * These are gnarly expressions. For improved readability, they have been moved to a separate
27 * file where they make use of named capture groups. That syntax isn't valid in JavaScript ES5,
28 * so the server-side strips these before delivering to the client.
32 * @property {Object} parser
35 strict
: require( './strict.regexp.js' ),
36 loose
: require( './loose.regexp.js' )
40 * The order here matches the order of captured matches in the `parser` property regexes.
44 * @property {string[]} properties
58 * A factory method to create an {@link mw.Uri} class with a default location to resolve relative URLs
59 * against (including protocol-relative URLs).
62 * @param {string|Function} documentLocation A full url, or function returning one.
63 * If passed a function, the return value may change over time and this will be honoured. (T74334)
64 * @return {mw.Uri} An mw.Uri class constructor
66 mw
.UriRelative = function ( documentLocation
) {
67 const getDefaultUri
= ( function () {
72 const hrefCur
= typeof documentLocation
=== 'string' ? documentLocation
: documentLocation();
73 if ( href
=== hrefCur
) {
77 uri
= new Uri( href
);
82 * Options for mw.Uri object.
84 * @typedef {Object} mw.Uri.UriOptions
85 * @property {boolean} [strictMode=false] Trigger strict mode parsing of the url.
86 * @property {boolean} [overrideKeys=false] Whether to let duplicate query parameters
87 * override each other (`true`) or automagically convert them to an array (`false`).
88 * @property {boolean} [arrayParams=false] Whether to parse array query parameters (e.g.
89 * `&foo[0]=a&foo[1]=b` or `&foo[]=a&foo[]=b`) or leave them alone. Currently this does not
90 * handle associative or multi-dimensional arrays, but that may be improved in the future.
91 * Implies `overrideKeys: true` (query parameters without `[...]` are not parsed as arrays).
95 * @classdesc Create and manipulate MediaWiki URIs.
97 * Intended to be minimal, but featureful; do not expect full RFC 3986 compliance. The use cases we
98 * have in mind are constructing 'next page' or 'previous page' URLs, detecting whether we need to
99 * use cross-domain proxies for an API, constructing simple URL-based API calls, etc. Parsing here
100 * is regex-based, so may not work on all URIs, but is good enough for most.
102 * You can modify the properties directly, then use the {@link mw.Uri#toString toString} method to extract the full URI
103 * string again. Example:
105 * var uri = new mw.Uri( 'http://example.com/mysite/mypage.php?quux=2' );
107 * if ( uri.host == 'example.com' ) {
108 * uri.host = 'foo.example.com';
109 * uri.extend( { bar: 1 } );
111 * $( 'a#id1' ).attr( 'href', uri );
112 * // anchor with id 'id1' now links to http://foo.example.com/mysite/mypage.php?bar=1&quux=2
114 * $( 'a#id2' ).attr( 'href', uri.clone().extend( { bar: 3, pif: 'paf' } ) );
115 * // anchor with id 'id2' now links to http://foo.example.com/mysite/mypage.php?bar=3&quux=2&pif=paf
119 * `http://usr:pwd@www.example.com:81/dir/dir.2/index.htm?q1=0&&test1&test2=&test3=value+%28escaped%29&r=1&r=2#top`
120 * the returned object will have the following properties:
125 * host 'www.example.com'
127 * path '/dir/dir.2/index.htm'
132 * test3: 'value (escaped)'
137 * Note: 'password' is technically not allowed for HTTP URIs, but it is possible with other kinds
140 * Parsing based on parseUri 1.2.2 (c) Steven Levithan <http://stevenlevithan.com>, MIT License.
141 * <http://stevenlevithan.com/demo/parseuri/js/>
147 * @description Construct a new URI object. Throws error if arguments are illegal/impossible, or
148 * otherwise don't parse.
149 * @param {Object|string} [uri] URI string, or an Object with appropriate properties (especially
150 * another URI object to clone). Object must have non-blank `protocol`, `host`, and `path`
151 * properties. If omitted (or set to `undefined`, `null` or empty string), then an object
152 * will be created for the default `uri` of this constructor (`location.href` for mw.Uri,
153 * other values for other instances -- see {@link mw.UriRelative} for details).
154 * @param {mw.Uri.UriOptions|boolean} [options] Object with options, or (backwards compatibility) a boolean
156 * @throws {Error} when the query string or fragment contains an unknown % sequence
158 function Uri( uri
, options
) {
159 const hasOptions
= ( options
!== undefined ),
160 defaultUri
= getDefaultUri();
162 options
= typeof options
=== 'object' ? options
: { strictMode
: !!options
};
163 options
= Object
.assign( {
169 this.arrayParams
= options
.arrayParams
;
171 if ( uri
!== undefined && uri
!== null && uri
!== '' ) {
172 if ( typeof uri
=== 'string' ) {
173 this.parse( uri
, options
);
174 } else if ( typeof uri
=== 'object' ) {
175 // Copy data over from existing URI object
176 for ( const prop
in uri
) {
177 // Only copy direct properties, not inherited ones
178 if ( Object
.prototype.hasOwnProperty
.call( uri
, prop
) ) {
179 // Deep copy object properties
180 if ( Array
.isArray( uri
[ prop
] ) || $.isPlainObject( uri
[ prop
] ) ) {
181 this[ prop
] = $.extend( true, {}, uri
[ prop
] );
183 this[ prop
] = uri
[ prop
];
191 } else if ( hasOptions
) {
192 // We didn't get a URI in the constructor, but we got options.
193 const hrefCur
= typeof documentLocation
=== 'string' ? documentLocation
: documentLocation();
194 this.parse( hrefCur
, options
);
196 // We didn't get a URI or options in the constructor, use the default instance.
197 return defaultUri
.clone();
200 // protocol-relative URLs
201 if ( !this.protocol
) {
202 this.protocol
= defaultUri
.protocol
;
206 this.host
= defaultUri
.host
;
209 this.port
= defaultUri
.port
;
212 if ( this.path
&& this.path
[ 0 ] !== '/' ) {
213 // A real relative URL, relative to defaultUri.path. We can't really handle that since we cannot
214 // figure out whether the last path component of defaultUri.path is a directory or a file.
215 throw new Error( 'Bad constructor arguments' );
217 if ( !( this.protocol
&& this.host
&& this.path
) ) {
218 throw new Error( 'Bad constructor arguments' );
223 * For example `http` (always present).
225 * @name mw.Uri.prototype.protocol
232 * @name mw.Uri.prototype.user
233 * @type {string|undefined}
238 * @name mw.Uri.prototype.password
239 * @type {string|undefined}
242 * For example `www.example.com` (always present).
244 * @name mw.Uri.prototype.host
250 * @name mw.Uri.prototype.port
251 * @type {string|undefined}
254 * For example `/dir/dir.2/index.htm` (always present).
256 * @name mw.Uri.prototype.path
260 * For example `{ a: '0', b: '', c: 'value' }` (always present).
262 * @name mw.Uri.prototype.query
268 * @name mw.Uri.prototype.fragment
269 * @type {string|undefined}
273 * Encode a value for inclusion in a url.
275 * Standard encodeURIComponent, with extra stuff to make all browsers work similarly and more
276 * compliant with RFC 3986. Similar to rawurlencode from PHP and our JS library
277 * {@link module:mediawiki.util.rawurlencode mw.util.rawurlencode}, except this also replaces spaces with `+`.
280 * @name mw.Uri.encode
281 * @param {string} s String to encode
282 * @return {string} Encoded string for URI
284 Uri
.encode = function ( s
) {
285 return encodeURIComponent( s
)
286 .replace( /!/g, '%21' ).replace( /'/g, '%27' ).replace( /\(/g, '%28' )
287 .replace( /\)/g, '%29' ).replace( /\*/g, '%2A
' )
288 .replace( /%20/g, '+' );
292 * Decode a url encoded value.
294 * Reversed {@link mw.Uri.encode encode}. Standard decodeURIComponent, with addition of replacing
298 * @name mw.Uri.decode
299 * @param {string} s String to decode
300 * @return {string} Decoded string
301 * @throws {Error} when the string contains an unknown % sequence
303 Uri.decode = function ( s ) {
304 return decodeURIComponent( s.replace( /\+/g, '%20' ) );
307 Uri.prototype = /** @lends mw.Uri.prototype */ {
310 * Parse a string and set our properties accordingly.
313 * @param {string} str URI, see constructor.
314 * @param {Object} options See constructor.
315 * @throws {Error} when the query string or fragment contains an unknown % sequence
317 parse: function ( str, options ) {
318 const hasOwn = Object.prototype.hasOwnProperty;
320 // Apply parser regex and set all properties based on the result
321 const matches = parser[ options.strictMode ? 'strict
' : 'loose
' ].exec( str );
322 properties.forEach( ( property, i ) => {
323 this[ property ] = matches[ i + 1 ];
326 // uri.query starts out as the query string; we will parse it into key-val pairs then make
327 // that object the "query" property.
328 // we overwrite query in uri way to make cloning easier, it can use the same list of properties.
330 // using replace to iterate over a string
333 this.query.replace( /(?:^|&)([^&=]*)(?:(=)([^&]*))?/g, ( match, k, eq, v ) => {
334 let arrayKeyMatch, i;
337 v = ( eq === '' || eq === undefined ) ? null : Uri.decode( v );
338 arrayKeyMatch = k.match( /^([^[]+)\[(\d*)\]$/ );
340 // If arrayParams and this parameter name contains an array index...
341 if ( options.arrayParams && arrayKeyMatch ) {
342 // Remove the index from parameter name
343 k = arrayKeyMatch[ 1 ];
345 // Turn the parameter value into an array (throw away anything else)
346 if ( !Array.isArray( q[ k ] ) ) {
350 i = arrayKeyMatch[ 2 ];
352 // If no explicit index, append at the end
358 // If overrideKeys, always (re)set top level value.
359 // If not overrideKeys but this key wasn't
set before
, then we
set it as well
.
360 // arrayParams implies overrideKeys (no array handling for non-array params).
361 } else if ( options
.arrayParams
|| options
.overrideKeys
|| !hasOwn
.call( q
, k
) ) {
364 // Use arrays if overrideKeys is false and key was already seen before
366 // Once before, still a string, turn into an array
367 if ( typeof q
[ k
] === 'string' ) {
371 if ( Array
.isArray( q
[ k
] ) ) {
380 // Decode uri.fragment, otherwise it gets double-encoded when serializing
381 if ( this.fragment
!== undefined ) {
382 this.fragment
= Uri
.decode( this.fragment
);
387 * Get user and password section of a URI.
391 getUserInfo: function () {
392 return cat( '', this.user
, cat( ':', this.password
, '' ) );
396 * Get host and port section of a URI.
400 getHostPort: function () {
401 return this.host
+ cat( ':', this.port
, '' );
405 * Get the userInfo, host and port section of the URI.
407 * In most real-world URLs this is simply the hostname, but the definition of 'authority' section is more general.
411 getAuthority: function () {
412 return cat( '', this.getUserInfo(), '@' ) + this.getHostPort();
416 * Get the query arguments of the URL, encoded into a string.
418 * Does not preserve the original order of arguments passed in the URI. Does handle escaping.
422 getQueryString: function () {
424 arrayParams
= this.arrayParams
;
425 Object
.keys( this.query
).forEach( ( key
) => {
426 const val
= this.query
[ key
];
427 const k
= Uri
.encode( key
),
428 isArrayParam
= Array
.isArray( val
),
429 vals
= isArrayParam
? val
: [ val
];
430 vals
.forEach( ( v
, i
) => {
432 if ( arrayParams
&& isArrayParam
) {
433 ki
+= Uri
.encode( '[' + i
+ ']' );
437 } else if ( k
=== 'title' ) {
438 args
.push( ki
+ '=' + mw
.util
.wikiUrlencode( v
) );
440 args
.push( ki
+ '=' + Uri
.encode( v
) );
444 return args
.join( '&' );
448 * Get everything after the authority section of the URI.
452 getRelativePath: function () {
453 return this.path
+ cat( '?', this.getQueryString(), '', true ) + cat( '#', this.fragment
, '' );
457 * Get the entire URI string.
459 * Note that the output may not be precisely the same as the constructor input,
460 * due to order of query arguments.
461 * Note also that the fragment is not always roundtripped as-is; some characters will
462 * become encoded, including the slash character, which can cause problems with e.g.
463 * mediawiki.router. It is recommended to use the native URL class (via
464 * web2017-polyfills, which loads a polyfill if needed) in contexts where the fragment
467 * @return {string} The URI string
469 toString: function () {
470 return this.protocol
+ '://' + this.getAuthority() + this.getRelativePath();
476 * @return {Object} New URI object with same properties
479 return new Uri( this );
483 * Extend the query section of the URI with new parameters.
485 * @param {Object} parameters Query parameters to add to ours (or to override ours with) as an
487 * @return {Object} This URI object
489 extend: function ( parameters
) {
490 for ( const name
in parameters
) {
491 const parameter
= parameters
[ name
];
492 if ( parameter
!== undefined ) {
493 this.query
[ name
] = parameter
;
504 * Default to the current browsing location (for relative URLs).
509 mw
.Uri
= mw
.UriRelative( () => location
.href
);