resources/src/mediawiki.Uri/Uri.js
( function () {
var parser, properties;
/**
* Function that's useful when constructing the URI string -- we frequently encounter the pattern
* of having to add something to the URI as we go, but only if it's present, and to include a
* character before or after if so.
*
* @private
* @static
* @param {string|undefined} pre To prepend
* @param {string} val To include
* @param {string} post To append
* @param {boolean} raw If true, val will not be encoded
* @return {string} Result
*/
function cat( pre, val, post, raw ) {
if ( val === undefined || val === null || val === '' ) {
return '';
}
return pre + ( raw ? val : mw.Uri.encode( val ) ) + post;
}
/**
* Regular expressions to parse many common URIs.
*
* These are gnarly expressions. For improved readability, they have been moved to a separate
* file where they make use of named capture groups. That syntax isn't valid in JavaScript ES5,
* so the server-side strips these before delivering to the client.
*
* @private
* @static
* @property {Object} parser
*/
parser = {
strict: require( './strict.regexp.js' ),
loose: require( './loose.regexp.js' )
};
/**
* The order here matches the order of captured matches in the `parser` property regexes.
*
* @private
* @static
* @property {string[]} properties
*/
properties = [
'protocol',
'user',
'password',
'host',
'port',
'path',
'query',
'fragment'
];
/**
* A factory method to create an {@link mw.Uri} class with a default location to resolve relative URLs
* against (including protocol-relative URLs).
*
* @memberof mw
* @param {string|Function} documentLocation A full url, or function returning one.
* If passed a function, the return value may change over time and this will be honoured. (T74334)
* @return {mw.Uri} An mw.Uri class constructor
*/
mw.UriRelative = function ( documentLocation ) {
var getDefaultUri = ( function () {
// Cache
var href, uri;
return function () {
var hrefCur = typeof documentLocation === 'string' ? documentLocation : documentLocation();
if ( href === hrefCur ) {
return uri;
}
href = hrefCur;
uri = new Uri( href );
return uri;
};
}() );
/**
* Options for mw.Uri object.
*
* @typedef {Object} mw.Uri.UriOptions
* @property {boolean} [strictMode=false] Trigger strict mode parsing of the url.
* @property {boolean} [overrideKeys=false] Whether to let duplicate query parameters
* override each other (`true`) or automagically convert them to an array (`false`).
* @property {boolean} [arrayParams=false] Whether to parse array query parameters (e.g.
* `&foo[0]=a&foo[1]=b` or `&foo[]=a&foo[]=b`) or leave them alone. Currently this does not
* handle associative or multi-dimensional arrays, but that may be improved in the future.
* Implies `overrideKeys: true` (query parameters without `[...]` are not parsed as arrays).
*/
/**
* @classdesc Create and manipulate MediaWiki URIs.
*
* Intended to be minimal, but featureful; do not expect full RFC 3986 compliance. The use cases we
* have in mind are constructing 'next page' or 'previous page' URLs, detecting whether we need to
* use cross-domain proxies for an API, constructing simple URL-based API calls, etc. Parsing here
* is regex-based, so may not work on all URIs, but is good enough for most.
*
* You can modify the properties directly, then use the {@link mw.Uri#toString toString} method to extract the full URI
* string again. Example:
* ```
* var uri = new mw.Uri( 'http://example.com/mysite/mypage.php?quux=2' );
*
* if ( uri.host == 'example.com' ) {
* uri.host = 'foo.example.com';
* uri.extend( { bar: 1 } );
*
* $( 'a#id1' ).attr( 'href', uri );
* // anchor with id 'id1' now links to http://foo.example.com/mysite/mypage.php?bar=1&quux=2
*
* $( 'a#id2' ).attr( 'href', uri.clone().extend( { bar: 3, pif: 'paf' } ) );
* // anchor with id 'id2' now links to http://foo.example.com/mysite/mypage.php?bar=3&quux=2&pif=paf
* }
* ```
* Given a URI like
* `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`
* the returned object will have the following properties:
* ```
* protocol 'http'
* user 'usr'
* password 'pwd'
* host 'www.example.com'
* port '81'
* path '/dir/dir.2/index.htm'
* query {
* q1: '0',
* test1: null,
* test2: '',
* test3: 'value (escaped)'
* r: ['1', '2']
* }
* fragment 'top'
* ```
* Note: 'password' is technically not allowed for HTTP URIs, but it is possible with other kinds
* of URIs.
*
* Parsing based on parseUri 1.2.2 (c) Steven Levithan <http://stevenlevithan.com>, MIT License.
* <http://stevenlevithan.com/demo/parseuri/js/>
*
* @class
* @name mw.Uri
*
* @constructor
* @description Construct a new URI object. Throws error if arguments are illegal/impossible, or
* otherwise don't parse.
* @param {Object|string} [uri] URI string, or an Object with appropriate properties (especially
* another URI object to clone). Object must have non-blank `protocol`, `host`, and `path`
* properties. If omitted (or set to `undefined`, `null` or empty string), then an object
* will be created for the default `uri` of this constructor (`location.href` for mw.Uri,
* other values for other instances -- see {@link mw.UriRelative} for details).
* @param {mw.Uri.UriOptions|boolean} [options] Object with options, or (backwards compatibility) a boolean
* for strictMode
* @throws {Error} when the query string or fragment contains an unknown % sequence
*/
function Uri( uri, options ) {
var prop, hrefCur,
hasOptions = ( options !== undefined ),
defaultUri = getDefaultUri();
options = typeof options === 'object' ? options : { strictMode: !!options };
options = $.extend( {
strictMode: false,
overrideKeys: false,
arrayParams: false
}, options );
this.arrayParams = options.arrayParams;
if ( uri !== undefined && uri !== null && uri !== '' ) {
if ( typeof uri === 'string' ) {
this.parse( uri, options );
} else if ( typeof uri === 'object' ) {
// Copy data over from existing URI object
for ( prop in uri ) {
// Only copy direct properties, not inherited ones
if ( Object.prototype.hasOwnProperty.call( uri, prop ) ) {
// Deep copy object properties
if ( Array.isArray( uri[ prop ] ) || $.isPlainObject( uri[ prop ] ) ) {
this[ prop ] = $.extend( true, {}, uri[ prop ] );
} else {
this[ prop ] = uri[ prop ];
}
}
}
if ( !this.query ) {
this.query = {};
}
}
} else if ( hasOptions ) {
// We didn't get a URI in the constructor, but we got options.
hrefCur = typeof documentLocation === 'string' ? documentLocation : documentLocation();
this.parse( hrefCur, options );
} else {
// We didn't get a URI or options in the constructor, use the default instance.
return defaultUri.clone();
}
// protocol-relative URLs
if ( !this.protocol ) {
this.protocol = defaultUri.protocol;
}
// No host given:
if ( !this.host ) {
this.host = defaultUri.host;
// port ?
if ( !this.port ) {
this.port = defaultUri.port;
}
}
if ( this.path && this.path[ 0 ] !== '/' ) {
// A real relative URL, relative to defaultUri.path. We can't really handle that since we cannot
// figure out whether the last path component of defaultUri.path is a directory or a file.
throw new Error( 'Bad constructor arguments' );
}
if ( !( this.protocol && this.host && this.path ) ) {
throw new Error( 'Bad constructor arguments' );
}
}
/**
* For example `http` (always present).
*
* @name mw.Uri.prototype.protocol
* @type {string}
*/
/**
* For example `usr`.
*
* @name mw.Uri.prototype.user
* @type {string|undefined}
*/
/**
* For example `pwd`.
*
* @name mw.Uri.prototype.password
* @type {string|undefined}
*/
/**
* For example `www.example.com` (always present).
*
* @name mw.Uri.prototype.host
* @type {string}
*/
/**
* For example `81`.
*
* @name mw.Uri.prototype.port
* @type {string|undefined}
*/
/**
* For example `/dir/dir.2/index.htm` (always present).
*
* @name mw.Uri.prototype.path
* @type {string}
*/
/**
* For example `{ a: '0', b: '', c: 'value' }` (always present).
*
* @name mw.Uri.prototype.query
* @type {Object}
*/
/**
* For example `top`.
*
* @name mw.Uri.prototype.fragment
* @type {string|undefined}
*/
/**
* Encode a value for inclusion in a url.
*
* Standard encodeURIComponent, with extra stuff to make all browsers work similarly and more
* compliant with RFC 3986. Similar to rawurlencode from PHP and our JS library
* {@link module:mediawiki.util.rawurlencode mw.util.rawurlencode}, except this also replaces spaces with `+`.
*
* @method
* @name mw.Uri.encode
* @param {string} s String to encode
* @return {string} Encoded string for URI
*/
Uri.encode = function ( s ) {
return encodeURIComponent( s )
.replace( /!/g, '%21' ).replace( /'/g, '%27' ).replace( /\(/g, '%28' )
.replace( /\)/g, '%29' ).replace( /\*/g, '%2A' )
.replace( /%20/g, '+' );
};
/**
* Decode a url encoded value.
*
* Reversed {@link mw.Uri.encode encode}. Standard decodeURIComponent, with addition of replacing
* `+` with a space.
*
* @method
* @name mw.Uri.decode
* @param {string} s String to decode
* @return {string} Decoded string
* @throws {Error} when the string contains an unknown % sequence
*/
Uri.decode = function ( s ) {
return decodeURIComponent( s.replace( /\+/g, '%20' ) );
};
Uri.prototype = /** @lends mw.Uri.prototype */ {
/**
* Parse a string and set our properties accordingly.
*
* @private
* @param {string} str URI, see constructor.
* @param {Object} options See constructor.
* @throws {Error} when the query string or fragment contains an unknown % sequence
*/
parse: function ( str, options ) {
var q, matches,
uri = this,
hasOwn = Object.prototype.hasOwnProperty;
// Apply parser regex and set all properties based on the result
matches = parser[ options.strictMode ? 'strict' : 'loose' ].exec( str );
properties.forEach( function ( property, i ) {
uri[ property ] = matches[ i + 1 ];
} );
// uri.query starts out as the query string; we will parse it into key-val pairs then make
// that object the "query" property.
// we overwrite query in uri way to make cloning easier, it can use the same list of properties.
q = {};
// using replace to iterate over a string
if ( uri.query ) {
// eslint-disable-next-line security/detect-unsafe-regex
uri.query.replace( /(?:^|&)([^&=]*)(?:(=)([^&]*))?/g, function ( match, k, eq, v ) {
var arrayKeyMatch, i;
if ( k ) {
k = Uri.decode( k );
v = ( eq === '' || eq === undefined ) ? null : Uri.decode( v );
arrayKeyMatch = k.match( /^([^[]+)\[(\d*)\]$/ );
// If arrayParams and this parameter name contains an array index...
if ( options.arrayParams && arrayKeyMatch ) {
// Remove the index from parameter name
k = arrayKeyMatch[ 1 ];
// Turn the parameter value into an array (throw away anything else)
if ( !Array.isArray( q[ k ] ) ) {
q[ k ] = [];
}
i = arrayKeyMatch[ 2 ];
if ( i === '' ) {
// If no explicit index, append at the end
i = q[ k ].length;
}
q[ k ][ i ] = v;
// If overrideKeys, always (re)set top level value.
// If not overrideKeys but this key wasn't set before, then we set it as well.
// arrayParams implies overrideKeys (no array handling for non-array params).
} else if ( options.arrayParams || options.overrideKeys || !hasOwn.call( q, k ) ) {
q[ k ] = v;
// Use arrays if overrideKeys is false and key was already seen before
} else {
// Once before, still a string, turn into an array
if ( typeof q[ k ] === 'string' ) {
q[ k ] = [ q[ k ] ];
}
// Add to the array
if ( Array.isArray( q[ k ] ) ) {
q[ k ].push( v );
}
}
}
} );
}
uri.query = q;
// Decode uri.fragment, otherwise it gets double-encoded when serializing
if ( uri.fragment !== undefined ) {
uri.fragment = Uri.decode( uri.fragment );
}
},
/**
* Get user and password section of a URI.
*
* @return {string}
*/
getUserInfo: function () {
return cat( '', this.user, cat( ':', this.password, '' ) );
},
/**
* Get host and port section of a URI.
*
* @return {string}
*/
getHostPort: function () {
return this.host + cat( ':', this.port, '' );
},
/**
* Get the userInfo, host and port section of the URI.
*
* In most real-world URLs this is simply the hostname, but the definition of 'authority' section is more general.
*
* @return {string}
*/
getAuthority: function () {
return cat( '', this.getUserInfo(), '@' ) + this.getHostPort();
},
/**
* Get the query arguments of the URL, encoded into a string.
*
* Does not preserve the original order of arguments passed in the URI. Does handle escaping.
*
* @return {string}
*/
getQueryString: function () {
var args = [],
arrayParams = this.arrayParams;
Object.keys( this.query ).forEach( function ( key ) {
var val = this.query[ key ];
var k = Uri.encode( key ),
isArrayParam = Array.isArray( val ),
vals = isArrayParam ? val : [ val ];
vals.forEach( function ( v, i ) {
var ki = k;
if ( arrayParams && isArrayParam ) {
ki += Uri.encode( '[' + i + ']' );
}
if ( v === null ) {
args.push( ki );
} else if ( k === 'title' ) {
args.push( ki + '=' + mw.util.wikiUrlencode( v ) );
} else {
args.push( ki + '=' + Uri.encode( v ) );
}
} );
}.bind( this ) );
return args.join( '&' );
},
/**
* Get everything after the authority section of the URI.
*
* @return {string}
*/
getRelativePath: function () {
return this.path + cat( '?', this.getQueryString(), '', true ) + cat( '#', this.fragment, '' );
},
/**
* Get the entire URI string.
*
* Note that the output may not be precisely the same as the constructor input,
* due to order of query arguments.
* Note also that the fragment is not always roundtripped as-is; some characters will
* become encoded, including the slash character, which can cause problems with e.g.
* mediawiki.router. It is recommended to use the native URL class (via
* web2017-polyfills, which loads a polyfill if needed) in contexts where the fragment
* is important.
*
* @return {string} The URI string
*/
toString: function () {
return this.protocol + '://' + this.getAuthority() + this.getRelativePath();
},
/**
* Clone this URI.
*
* @return {Object} New URI object with same properties
*/
clone: function () {
return new Uri( this );
},
/**
* Extend the query section of the URI with new parameters.
*
* @param {Object} parameters Query parameters to add to ours (or to override ours with) as an
* object
* @return {Object} This URI object
*/
extend: function ( parameters ) {
$.extend( this.query, parameters );
return this;
}
};
return Uri;
};
/**
* Default to the current browsing location (for relative URLs).
*
* @ignore
* @return {mw.Uri}
*/
mw.Uri = mw.UriRelative( function () {
return location.href;
} );
}() );