2 * Experimental advanced wikitext parser-emitter.
3 * See: https://www.mediawiki.org/wiki/Extension:UploadWizard/MessageParser for docs
5 * @author neilk@wikimedia.org
6 * @author mflaschen@wikimedia.org
15 slice
= Array
.prototype.slice
,
18 'SITENAME': mw
.config
.get( 'wgSiteName' )
20 // This is a whitelist based on, but simpler than, Sanitizer.php.
21 // Self-closing tags are not currently supported.
22 allowedHtmlElements
: [
26 // Key tag name, value allowed attributes for that tag.
27 // See Sanitizer::setupAttributeWhitelist
28 allowedHtmlCommonAttributes
: [
41 // Attributes allowed for specific elements.
42 // Key is element name in lower case
43 // Value is array of allowed attributes for that element
44 allowedHtmlAttributesByElement
: {},
45 messages
: mw
.messages
,
46 language
: mw
.language
,
48 // Same meaning as in mediawiki.js.
50 // Only 'text', 'parse', and 'escaped' are supported, and the
51 // actual escaping for 'escaped' is done by other code (generally
52 // through mediawiki.js).
54 // However, note that this default only
55 // applies to direct calls to jqueryMsg. The default for mediawiki.js itself
56 // is 'text', including when it uses jqueryMsg.
62 * Wrapper around jQuery append that converts all non-objects to TextNode so append will not
63 * convert what it detects as an htmlString to an element.
65 * Object elements of children (jQuery, HTMLElement, TextNode, etc.) will be left as is.
68 * @param {jQuery} $parent Parent node wrapped by jQuery
69 * @param {Object|string|Array} children What to append, with the same possible types as jQuery
70 * @return {jQuery} $parent
72 function appendWithoutParsing( $parent
, children
) {
75 if ( !$.isArray( children
) ) {
76 children
= [children
];
79 for ( i
= 0, len
= children
.length
; i
< len
; i
++ ) {
80 if ( typeof children
[i
] !== 'object' ) {
81 children
[i
] = document
.createTextNode( children
[i
] );
85 return $parent
.append( children
);
89 * Decodes the main HTML entities, those encoded by mw.html.escape.
92 * @param {string} encoded Encoded string
93 * @return {string} String with those entities decoded
95 function decodePrimaryHtmlEntities( encoded
) {
97 .replace( /'/g, '\'' )
98 .replace( /"/g, '"' )
99 .replace( /</g, '<' )
100 .replace( />/g, '>' )
101 .replace( /&/g, '&' );
105 * Given parser options, return a function that parses a key and replacements, returning jQuery object
107 * Try to parse a key and optional replacements, returning a jQuery object that may be a tree of jQuery nodes.
108 * If there was an error parsing, return the key and the error message (wrapped in jQuery). This should put the error right into
109 * the interface, without causing the page to halt script execution, and it hopefully should be clearer how to fix it.
111 * @param {Object} options Parser options
113 * @return {Array} return.args First element is the key, replacements may be in array in 2nd element, or remaining elements.
114 * @return {jQuery} return.return
116 function getFailableParserFn( options
) {
117 var parser
= new mw
.jqueryMsg
.parser( options
);
119 return function ( args
) {
122 argsArray
= $.isArray( args
[1] ) ? args
[1] : slice
.call( args
, 1 );
124 return parser
.parse( key
, argsArray
);
126 fallback
= parser
.settings
.messages
.get( key
);
127 mw
.log
.warn( 'mediawiki.jqueryMsg: ' + key
+ ': ' + e
.message
);
128 return $( '<span>' ).text( fallback
);
136 * Returns a function suitable for use as a global, to construct strings from the message key (and optional replacements).
139 * window.gM = mediaWiki.parser.getMessageFunction( options );
140 * $( 'p#headline' ).html( gM( 'hello-user', username ) );
142 * Like the old gM() function this returns only strings, so it destroys any bindings. If you want to preserve bindings use the
143 * jQuery plugin version instead. This is only included for backwards compatibility with gM().
145 * N.B. replacements are variadic arguments or an array in second parameter. In other words:
146 * somefunction( a, b, c, d )
148 * somefunction( a, [b, c, d] )
150 * @param {Object} options parser options
151 * @return {Function} Function suitable for assigning to window.gM
152 * @return {string} return.key Message key.
153 * @return {Array|Mixed} return.replacements Optional variable replacements (variadically or an array).
154 * @return {string} return.return Rendered HTML.
156 mw
.jqueryMsg
.getMessageFunction = function ( options
) {
157 var failableParserFn
= getFailableParserFn( options
),
160 if ( options
&& options
.format
!== undefined ) {
161 format
= options
.format
;
163 format
= parserDefaults
.format
;
167 var failableResult
= failableParserFn( arguments
);
168 if ( format
=== 'text' || format
=== 'escaped' ) {
169 return failableResult
.text();
171 return failableResult
.html();
177 * Returns a jQuery plugin which parses the message in the message key, doing replacements optionally, and appends the nodes to
178 * the current selector. Bindings to passed-in jquery elements are preserved. Functions become click handlers for [$1 linktext] links.
181 * $.fn.msg = mediaWiki.parser.getJqueryPlugin( options );
182 * var userlink = $( '<a>' ).click( function () { alert( "hello!!" ) } );
183 * $( 'p#headline' ).msg( 'hello-user', userlink );
185 * N.B. replacements are variadic arguments or an array in second parameter. In other words:
186 * somefunction( a, b, c, d )
188 * somefunction( a, [b, c, d] )
190 * We append to 'this', which in a jQuery plugin context will be the selected elements.
192 * @param {Object} options Parser options
193 * @return {Function} Function suitable for assigning to jQuery plugin, such as jQuery#msg
194 * @return {string} return.key Message key.
195 * @return {Array|Mixed} return.replacements Optional variable replacements (variadically or an array).
196 * @return {jQuery} return.return
198 mw
.jqueryMsg
.getPlugin = function ( options
) {
199 var failableParserFn
= getFailableParserFn( options
);
202 var $target
= this.empty();
203 // TODO: Simply appendWithoutParsing( $target, failableParserFn( arguments ).contents() )
204 // or Simply appendWithoutParsing( $target, failableParserFn( arguments ) )
205 $.each( failableParserFn( arguments
).contents(), function ( i
, node
) {
206 appendWithoutParsing( $target
, node
);
214 * Describes an object, whose primary duty is to .parse() message keys.
218 * @param {Object} options
220 mw
.jqueryMsg
.parser = function ( options
) {
221 this.settings
= $.extend( {}, parserDefaults
, options
);
222 this.settings
.onlyCurlyBraceTransform
= ( this.settings
.format
=== 'text' || this.settings
.format
=== 'escaped' );
224 this.emitter
= new mw
.jqueryMsg
.htmlEmitter( this.settings
.language
, this.settings
.magic
);
227 mw
.jqueryMsg
.parser
.prototype = {
229 * Cache mapping MediaWiki message keys and the value onlyCurlyBraceTransform, to the AST of the message.
231 * In most cases, the message is a string so this is identical.
232 * (This is why we would like to move this functionality server-side).
234 * The two parts of the key are separated by colon. For example:
236 * "message-key:true": ast
238 * if they key is "message-key" and onlyCurlyBraceTransform is true.
240 * This cache is shared by all instances of mw.jqueryMsg.parser.
242 * NOTE: We promise, it's static - when you create this empty object
243 * in the prototype, each new instance of the class gets a reference
244 * to the same object.
252 * Where the magic happens.
253 * Parses a message from the key, and swaps in replacements as necessary, wraps in jQuery
254 * If an error is thrown, returns original key, and logs the error
255 * @param {string} key Message key.
256 * @param {Array} replacements Variable replacements for $1, $2... $n
259 parse: function ( key
, replacements
) {
260 return this.emitter
.emit( this.getAst( key
), replacements
);
264 * Fetch the message string associated with a key, return parsed structure. Memoized.
265 * Note that we pass '[' + key + ']' back for a missing message here.
266 * @param {string} key
267 * @return {string|Array} string of '[key]' if message missing, simple string if possible, array of arrays if needs parsing
269 getAst: function ( key
) {
270 var cacheKey
= [key
, this.settings
.onlyCurlyBraceTransform
].join( ':' ), wikiText
;
272 if ( this.astCache
[ cacheKey
] === undefined ) {
273 wikiText
= this.settings
.messages
.get( key
);
274 if ( typeof wikiText
!== 'string' ) {
275 wikiText
= '\\[' + key
+ '\\]';
277 this.astCache
[ cacheKey
] = this.wikiTextToAst( wikiText
);
279 return this.astCache
[ cacheKey
];
283 * Parses the input wikiText into an abstract syntax tree, essentially an s-expression.
285 * CAVEAT: This does not parse all wikitext. It could be more efficient, but it's pretty good already.
286 * n.b. We want to move this functionality to the server. Nothing here is required to be on the client.
288 * @param {string} input Message string wikitext
290 * @return {Mixed} abstract syntax tree
292 wikiTextToAst: function ( input
) {
293 var pos
, settings
= this.settings
, concat
= Array
.prototype.concat
,
294 regularLiteral
, regularLiteralWithoutBar
, regularLiteralWithoutSpace
, regularLiteralWithSquareBrackets
,
295 doubleQuote
, singleQuote
, backslash
, anyCharacter
, asciiAlphabetLiteral
,
296 escapedOrLiteralWithoutSpace
, escapedOrLiteralWithoutBar
, escapedOrRegularLiteral
,
297 whitespace
, dollar
, digits
, htmlDoubleQuoteAttributeValue
, htmlSingleQuoteAttributeValue
,
298 htmlAttributeEquals
, openHtmlStartTag
, optionalForwardSlash
, openHtmlEndTag
, closeHtmlTag
,
299 openExtlink
, closeExtlink
, wikilinkPage
, wikilinkContents
, openWikilink
, closeWikilink
, templateName
, pipe
, colon
,
300 templateContents
, openTemplate
, closeTemplate
,
301 nonWhitespaceExpression
, paramExpression
, expression
, curlyBraceTransformExpression
, result
;
303 // Indicates current position in input as we parse through it.
304 // Shared among all parsing functions below.
307 // =========================================================
308 // parsing combinators - could be a library on its own
309 // =========================================================
312 * Try parsers until one works, if none work return null
314 * @param {Function[]} ps
315 * @return {string|null}
317 function choice( ps
) {
320 for ( i
= 0; i
< ps
.length
; i
++ ) {
322 if ( result
!== null ) {
331 * Try several ps in a row, all must succeed or return null.
332 * This is the only eager one.
334 * @param {Function[]} ps
335 * @return {string|null}
337 function sequence( ps
) {
341 for ( i
= 0; i
< ps
.length
; i
++ ) {
343 if ( res
=== null ) {
353 * Run the same parser over and over until it fails.
354 * Must succeed a minimum of n times or return null.
357 * @param {Function} p
358 * @return {string|null}
360 function nOrMore( n
, p
) {
362 var originalPos
= pos
,
365 while ( parsed
!== null ) {
366 result
.push( parsed
);
369 if ( result
.length
< n
) {
378 * There is a general pattern -- parse a thing, if that worked, apply transform, otherwise return null.
380 * TODO: But using this as a combinator seems to cause problems when combined with #nOrMore().
381 * May be some scoping issue
384 * @param {Function} p
385 * @param {Function} fn
386 * @return {string|null}
388 function transform( p
, fn
) {
391 return result
=== null ? null : fn( result
);
396 * Just make parsers out of simpler JS builtin types
400 * @return {string} return.return
402 function makeStringParser( s
) {
406 if ( input
.substr( pos
, len
) === s
) {
415 * Makes a regex parser, given a RegExp object.
416 * The regex being passed in should start with a ^ to anchor it to the start
420 * @param {RegExp} regex anchored regex
421 * @return {Function} function to parse input based on the regex
423 function makeRegexParser( regex
) {
425 var matches
= input
.slice( pos
).match( regex
);
426 if ( matches
=== null ) {
429 pos
+= matches
[0].length
;
434 // ===================================================================
435 // General patterns above this line -- wikitext specific parsers below
436 // ===================================================================
438 // Parsing functions follow. All parsing functions work like this:
439 // They don't accept any arguments.
440 // Instead, they just operate non destructively on the string 'input'
441 // As they can consume parts of the string, they advance the shared variable pos,
442 // and return tokens (or whatever else they want to return).
443 // some things are defined as closures and other things as ordinary functions
444 // converting everything to a closure makes it a lot harder to debug... errors pop up
445 // but some debuggers can't tell you exactly where they come from. Also the mutually
446 // recursive functions seem not to work in all browsers then. (Tested IE6-7, Opera, Safari, FF)
447 // This may be because, to save code, memoization was removed
449 regularLiteral
= makeRegexParser( /^[^{}\[\]$<\\]/ );
450 regularLiteralWithoutBar
= makeRegexParser( /^[^{}\[\]$\\|]/ );
451 regularLiteralWithoutSpace
= makeRegexParser( /^[^{}\[\]$\s]/ );
452 regularLiteralWithSquareBrackets
= makeRegexParser( /^[^{}$\\]/ );
454 backslash
= makeStringParser( '\\' );
455 doubleQuote
= makeStringParser( '"' );
456 singleQuote
= makeStringParser( '\'' );
457 anyCharacter
= makeRegexParser( /^./ );
459 openHtmlStartTag
= makeStringParser( '<' );
460 optionalForwardSlash
= makeRegexParser( /^\/?/ );
461 openHtmlEndTag
= makeStringParser( '</' );
462 htmlAttributeEquals
= makeRegexParser( /^\s*=\s*/ );
463 closeHtmlTag
= makeRegexParser( /^\s*>/ );
465 function escapedLiteral() {
466 var result
= sequence( [
470 return result
=== null ? null : result
[1];
472 escapedOrLiteralWithoutSpace
= choice( [
474 regularLiteralWithoutSpace
476 escapedOrLiteralWithoutBar
= choice( [
478 regularLiteralWithoutBar
480 escapedOrRegularLiteral
= choice( [
484 // Used to define "literals" without spaces, in space-delimited situations
485 function literalWithoutSpace() {
486 var result
= nOrMore( 1, escapedOrLiteralWithoutSpace
)();
487 return result
=== null ? null : result
.join( '' );
489 // Used to define "literals" within template parameters. The pipe character is the parameter delimeter, so by default
490 // it is not a literal in the parameter
491 function literalWithoutBar() {
492 var result
= nOrMore( 1, escapedOrLiteralWithoutBar
)();
493 return result
=== null ? null : result
.join( '' );
496 // Used for wikilink page names. Like literalWithoutBar, but
497 // without allowing escapes.
498 function unescapedLiteralWithoutBar() {
499 var result
= nOrMore( 1, regularLiteralWithoutBar
)();
500 return result
=== null ? null : result
.join( '' );
504 var result
= nOrMore( 1, escapedOrRegularLiteral
)();
505 return result
=== null ? null : result
.join( '' );
508 function curlyBraceTransformExpressionLiteral() {
509 var result
= nOrMore( 1, regularLiteralWithSquareBrackets
)();
510 return result
=== null ? null : result
.join( '' );
513 asciiAlphabetLiteral
= makeRegexParser( /[A-Za-z]+/ );
514 htmlDoubleQuoteAttributeValue
= makeRegexParser( /^[^"]*/ );
515 htmlSingleQuoteAttributeValue
= makeRegexParser( /^[^']*/ );
517 whitespace
= makeRegexParser( /^\s+/ );
518 dollar
= makeStringParser( '$' );
519 digits
= makeRegexParser( /^\d+/ );
521 function replacement() {
522 var result
= sequence( [
526 if ( result
=== null ) {
529 return [ 'REPLACE', parseInt( result
[1], 10 ) - 1 ];
531 openExtlink
= makeStringParser( '[' );
532 closeExtlink
= makeStringParser( ']' );
533 // this extlink MUST have inner contents, e.g. [foo] not allowed; [foo bar] [foo <i>bar</i>], etc. are allowed
535 var result
, parsedResult
;
537 parsedResult
= sequence( [
539 nonWhitespaceExpression
,
541 nOrMore( 1, expression
),
544 if ( parsedResult
!== null ) {
545 result
= [ 'EXTLINK', parsedResult
[1] ];
546 // TODO (mattflaschen, 2013-03-22): Clean this up if possible.
547 // It's avoiding CONCAT for single nodes, so they at least doesn't get the htmlEmitter span.
548 if ( parsedResult
[3].length
=== 1 ) {
549 result
.push( parsedResult
[3][0] );
551 result
.push( ['CONCAT'].concat( parsedResult
[3] ) );
556 // this is the same as the above extlink, except that the url is being passed on as a parameter
557 function extLinkParam() {
558 var result
= sequence( [
566 if ( result
=== null ) {
569 return [ 'EXTLINKPARAM', parseInt( result
[2], 10 ) - 1, result
[4] ];
571 openWikilink
= makeStringParser( '[[' );
572 closeWikilink
= makeStringParser( ']]' );
573 pipe
= makeStringParser( '|' );
575 function template() {
576 var result
= sequence( [
581 return result
=== null ? null : result
[1];
584 wikilinkPage
= choice( [
585 unescapedLiteralWithoutBar
,
589 function pipedWikilink() {
590 var result
= sequence( [
595 return result
=== null ? null : [ result
[0], result
[2] ];
598 wikilinkContents
= choice( [
600 wikilinkPage
// unpiped link
603 function wikilink() {
604 var result
, parsedResult
, parsedLinkContents
;
607 parsedResult
= sequence( [
612 if ( parsedResult
!== null ) {
613 parsedLinkContents
= parsedResult
[1];
614 result
= [ 'WIKILINK' ].concat( parsedLinkContents
);
619 // TODO: Support data- if appropriate
620 function doubleQuotedHtmlAttributeValue() {
621 var parsedResult
= sequence( [
623 htmlDoubleQuoteAttributeValue
,
626 return parsedResult
=== null ? null : parsedResult
[1];
629 function singleQuotedHtmlAttributeValue() {
630 var parsedResult
= sequence( [
632 htmlSingleQuoteAttributeValue
,
635 return parsedResult
=== null ? null : parsedResult
[1];
638 function htmlAttribute() {
639 var parsedResult
= sequence( [
641 asciiAlphabetLiteral
,
644 doubleQuotedHtmlAttributeValue
,
645 singleQuotedHtmlAttributeValue
648 return parsedResult
=== null ? null : [parsedResult
[1], parsedResult
[3]];
652 * Checks if HTML is allowed
654 * @param {string} startTagName HTML start tag name
655 * @param {string} endTagName HTML start tag name
656 * @param {Object} attributes array of consecutive key value pairs,
657 * with index 2 * n being a name and 2 * n + 1 the associated value
658 * @return {boolean} true if this is HTML is allowed, false otherwise
660 function isAllowedHtml( startTagName
, endTagName
, attributes
) {
661 var i
, len
, attributeName
;
663 startTagName
= startTagName
.toLowerCase();
664 endTagName
= endTagName
.toLowerCase();
665 if ( startTagName
!== endTagName
|| $.inArray( startTagName
, settings
.allowedHtmlElements
) === -1 ) {
669 for ( i
= 0, len
= attributes
.length
; i
< len
; i
+= 2 ) {
670 attributeName
= attributes
[i
];
671 if ( $.inArray( attributeName
, settings
.allowedHtmlCommonAttributes
) === -1 &&
672 $.inArray( attributeName
, settings
.allowedHtmlAttributesByElement
[startTagName
] || [] ) === -1 ) {
680 function htmlAttributes() {
681 var parsedResult
= nOrMore( 0, htmlAttribute
)();
682 // Un-nest attributes array due to structure of jQueryMsg operations (see emit).
683 return concat
.apply( ['HTMLATTRIBUTES'], parsedResult
);
686 // Subset of allowed HTML markup.
687 // Most elements and many attributes allowed on the server are not supported yet.
689 var result
= null, parsedOpenTagResult
, parsedHtmlContents
,
690 parsedCloseTagResult
, wrappedAttributes
, attributes
,
691 startTagName
, endTagName
, startOpenTagPos
, startCloseTagPos
,
692 endOpenTagPos
, endCloseTagPos
;
694 // Break into three sequence calls. That should allow accurate reconstruction of the original HTML, and requiring an exact tag name match.
695 // 1. open through closeHtmlTag
697 // 3. openHtmlEnd through close
698 // This will allow recording the positions to reconstruct if HTML is to be treated as text.
700 startOpenTagPos
= pos
;
701 parsedOpenTagResult
= sequence( [
703 asciiAlphabetLiteral
,
705 optionalForwardSlash
,
709 if ( parsedOpenTagResult
=== null ) {
714 startTagName
= parsedOpenTagResult
[1];
716 parsedHtmlContents
= nOrMore( 0, expression
)();
718 startCloseTagPos
= pos
;
719 parsedCloseTagResult
= sequence( [
721 asciiAlphabetLiteral
,
725 if ( parsedCloseTagResult
=== null ) {
726 // Closing tag failed. Return the start tag and contents.
727 return [ 'CONCAT', input
.slice( startOpenTagPos
, endOpenTagPos
) ]
728 .concat( parsedHtmlContents
);
731 endCloseTagPos
= pos
;
732 endTagName
= parsedCloseTagResult
[1];
733 wrappedAttributes
= parsedOpenTagResult
[2];
734 attributes
= wrappedAttributes
.slice( 1 );
735 if ( isAllowedHtml( startTagName
, endTagName
, attributes
) ) {
736 result
= [ 'HTMLELEMENT', startTagName
, wrappedAttributes
]
737 .concat( parsedHtmlContents
);
739 // HTML is not allowed, so contents will remain how
740 // it was, while HTML markup at this level will be
742 // E.g. assuming script tags are not allowed:
744 // <script>[[Foo|bar]]</script>
746 // results in '<script>' and '</script>'
747 // (not treated as an HTML tag), surrounding a fully
750 // Concatenate everything from the tag, flattening the contents.
751 result
= [ 'CONCAT', input
.slice( startOpenTagPos
, endOpenTagPos
) ]
752 .concat( parsedHtmlContents
, input
.slice( startCloseTagPos
, endCloseTagPos
) );
758 templateName
= transform(
759 // see $wgLegalTitleChars
760 // not allowing : due to the need to catch "PLURAL:$1"
761 makeRegexParser( /^[ !"$&'()*,.\/0-9;=?@A-Z\^_`a-z~\x80-\xFF+\-]+/ ),
762 function ( result
) { return result
.toString(); }
764 function templateParam() {
768 nOrMore( 0, paramExpression
)
770 if ( result
=== null ) {
774 // use a CONCAT operator if there are multiple nodes, otherwise return the first node, raw.
775 return expr
.length
> 1 ? [ 'CONCAT' ].concat( expr
) : expr
[0];
778 function templateWithReplacement() {
779 var result
= sequence( [
784 return result
=== null ? null : [ result
[0], result
[2] ];
786 function templateWithOutReplacement() {
787 var result
= sequence( [
792 return result
=== null ? null : [ result
[0], result
[2] ];
794 function templateWithOutFirstParameter() {
795 var result
= sequence( [
799 return result
=== null ? null : [ result
[0], '' ];
801 colon
= makeStringParser( ':' );
802 templateContents
= choice( [
804 var res
= sequence( [
805 // templates can have placeholders for dynamic replacement eg: {{PLURAL:$1|one car|$1 cars}}
806 // or no placeholders eg: {{GRAMMAR:genitive|{{SITENAME}}}
807 choice( [ templateWithReplacement
, templateWithOutReplacement
, templateWithOutFirstParameter
] ),
808 nOrMore( 0, templateParam
)
810 return res
=== null ? null : res
[0].concat( res
[1] );
813 var res
= sequence( [
815 nOrMore( 0, templateParam
)
817 if ( res
=== null ) {
820 return [ res
[0] ].concat( res
[1] );
823 openTemplate
= makeStringParser( '{{' );
824 closeTemplate
= makeStringParser( '}}' );
825 nonWhitespaceExpression
= choice( [
833 paramExpression
= choice( [
842 expression
= choice( [
852 // Used when only {{-transformation is wanted, for 'text'
853 // or 'escaped' formats
854 curlyBraceTransformExpression
= choice( [
857 curlyBraceTransformExpressionLiteral
863 * @param {Function} rootExpression root parse function
865 function start( rootExpression
) {
866 var result
= nOrMore( 0, rootExpression
)();
867 if ( result
=== null ) {
870 return [ 'CONCAT' ].concat( result
);
872 // everything above this point is supposed to be stateless/static, but
873 // I am deferring the work of turning it into prototypes & objects. It's quite fast enough
874 // finally let's do some actual work...
876 // If you add another possible rootExpression, you must update the astCache key scheme.
877 result
= start( this.settings
.onlyCurlyBraceTransform
? curlyBraceTransformExpression
: expression
);
880 * For success, the p must have gotten to the end of the input
881 * and returned a non-null.
882 * n.b. This is part of language infrastructure, so we do not throw an internationalizable message.
884 if ( result
=== null || pos
!== input
.length
) {
885 throw new Error( 'Parse error at position ' + pos
.toString() + ' in input: ' + input
);
893 * htmlEmitter - object which primarily exists to emit HTML from parser ASTs
895 mw
.jqueryMsg
.htmlEmitter = function ( language
, magic
) {
896 this.language
= language
;
898 $.each( magic
, function ( key
, val
) {
899 jmsg
[ key
.toLowerCase() ] = function () {
905 * (We put this method definition here, and not in prototype, to make sure it's not overwritten by any magic.)
906 * Walk entire node structure, applying replacements and template functions when appropriate
907 * @param {Mixed} node Abstract syntax tree (top node or subnode)
908 * @param {Array} replacements for $1, $2, ... $n
909 * @return {Mixed} single-string node or array of nodes suitable for jQuery appending
911 this.emit = function ( node
, replacements
) {
912 var ret
, subnodes
, operation
,
914 switch ( typeof node
) {
919 // typeof returns object for arrays
921 // node is an array of nodes
922 subnodes
= $.map( node
.slice( 1 ), function ( n
) {
923 return jmsg
.emit( n
, replacements
);
925 operation
= node
[0].toLowerCase();
926 if ( typeof jmsg
[operation
] === 'function' ) {
927 ret
= jmsg
[ operation
]( subnodes
, replacements
);
929 throw new Error( 'Unknown operation "' + operation
+ '"' );
933 // Parsing the empty string (as an entire expression, or as a paramExpression in a template) results in undefined
934 // Perhaps a more clever parser can detect this, and return the empty string? Or is that useful information?
935 // The logical thing is probably to return the empty string here when we encounter undefined.
939 throw new Error( 'Unexpected type in AST: ' + typeof node
);
945 // For everything in input that follows double-open-curly braces, there should be an equivalent parser
946 // function. For instance {{PLURAL ... }} will be processed by 'plural'.
947 // If you have 'magic words' then configure the parser to have them upon creation.
949 // An emitter method takes the parent node, the array of subnodes and the array of replacements (the values that $1, $2... should translate to).
950 // Note: all such functions must be pure, with the exception of referring to other pure functions via this.language (convertPlural and so on)
951 mw
.jqueryMsg
.htmlEmitter
.prototype = {
953 * Parsing has been applied depth-first we can assume that all nodes here are single nodes
954 * Must return a single node to parents -- a jQuery with synthetic span
955 * However, unwrap any other synthetic spans in our children and pass them upwards
956 * @param {Mixed[]} nodes Some single nodes, some arrays of nodes
959 concat: function ( nodes
) {
960 var $span
= $( '<span>' ).addClass( 'mediaWiki_htmlEmitter' );
961 $.each( nodes
, function ( i
, node
) {
962 if ( node
instanceof jQuery
&& node
.hasClass( 'mediaWiki_htmlEmitter' ) ) {
963 $.each( node
.contents(), function ( j
, childNode
) {
964 appendWithoutParsing( $span
, childNode
);
967 // Let jQuery append nodes, arrays of nodes and jQuery objects
968 // other things (strings, numbers, ..) are appended as text nodes (not as HTML strings)
969 appendWithoutParsing( $span
, node
);
976 * Return escaped replacement of correct index, or string if unavailable.
977 * Note that we expect the parsed parameter to be zero-based. i.e. $1 should have become [ 0 ].
978 * if the specified parameter is not found return the same string
979 * (e.g. "$99" -> parameter 98 -> not found -> return "$99" )
981 * TODO: Throw error if nodes.length > 1 ?
983 * @param {Array} nodes List of one element, integer, n >= 0
984 * @param {Array} replacements List of at least n strings
985 * @return {String} replacement
987 replace: function ( nodes
, replacements
) {
988 var index
= parseInt( nodes
[0], 10 );
990 if ( index
< replacements
.length
) {
991 return replacements
[index
];
993 // index not found, fallback to displaying variable
994 return '$' + ( index
+ 1 );
999 * Transform wiki-link
1002 * It only handles basic cases, either no pipe, or a pipe with an explicit
1005 * It does not attempt to handle features like the pipe trick.
1006 * However, the pipe trick should usually not be present in wikitext retrieved
1007 * from the server, since the replacement is done at save time.
1008 * It may, though, if the wikitext appears in extension-controlled content.
1012 wikilink: function ( nodes
) {
1013 var page
, anchor
, url
;
1016 url
= mw
.util
.getUrl( page
);
1018 // [[Some Page]] or [[Namespace:Some Page]]
1019 if ( nodes
.length
=== 1 ) {
1024 * [[Some Page|anchor text]] or
1025 * [[Namespace:Some Page|anchor]
1031 return $( '<a>' ).attr( {
1038 * Converts array of HTML element key value pairs to object
1040 * @param {Array} nodes Array of consecutive key value pairs, with index 2 * n being a
1041 * name and 2 * n + 1 the associated value
1042 * @return {Object} Object mapping attribute name to attribute value
1044 htmlattributes: function ( nodes
) {
1045 var i
, len
, mapping
= {};
1046 for ( i
= 0, len
= nodes
.length
; i
< len
; i
+= 2 ) {
1047 mapping
[nodes
[i
]] = decodePrimaryHtmlEntities( nodes
[i
+ 1] );
1053 * Handles an (already-validated) HTML element.
1055 * @param {Array} nodes Nodes to process when creating element
1056 * @return {jQuery|Array} jQuery node for valid HTML or array for disallowed element
1058 htmlelement: function ( nodes
) {
1059 var tagName
, attributes
, contents
, $element
;
1061 tagName
= nodes
.shift();
1062 attributes
= nodes
.shift();
1064 $element
= $( document
.createElement( tagName
) ).attr( attributes
);
1065 return appendWithoutParsing( $element
, contents
);
1069 * Transform parsed structure into external link
1070 * If the href is a jQuery object, treat it as "enclosing" the link text.
1072 * - ... function, treat it as the click handler.
1073 * - ... string, treat it as a URI.
1075 * TODO: throw an error if nodes.length > 2 ?
1077 * @param {Array} nodes List of two elements, {jQuery|Function|String} and {String}
1080 extlink: function ( nodes
) {
1083 contents
= nodes
[1];
1084 if ( arg
instanceof jQuery
) {
1088 if ( typeof arg
=== 'function' ) {
1089 $el
.attr( 'href', '#' )
1090 .click( function ( e
) {
1095 $el
.attr( 'href', arg
.toString() );
1098 return appendWithoutParsing( $el
, contents
);
1102 * This is basically use a combination of replace + external link (link with parameter
1103 * as url), but we don't want to run the regular replace here-on: inserting a
1104 * url as href-attribute of a link will automatically escape it already, so
1105 * we don't want replace to (manually) escape it as well.
1107 * TODO: throw error if nodes.length > 1 ?
1109 * @param {Array} nodes List of one element, integer, n >= 0
1110 * @param {Array} replacements List of at least n strings
1111 * @return {string} replacement
1113 extlinkparam: function ( nodes
, replacements
) {
1115 index
= parseInt( nodes
[0], 10 );
1116 if ( index
< replacements
.length
) {
1117 replacement
= replacements
[index
];
1119 replacement
= '$' + ( index
+ 1 );
1121 return this.extlink( [ replacement
, nodes
[1] ] );
1125 * Transform parsed structure into pluralization
1126 * n.b. The first node may be a non-integer (for instance, a string representing an Arabic number).
1127 * So convert it back with the current language's convertNumber.
1128 * @param {Array} nodes List of nodes, [ {string|number}, {string}, {string} ... ]
1129 * @return {string} selected pluralized form according to current language
1131 plural: function ( nodes
) {
1132 var forms
, firstChild
, firstChildText
,
1133 explicitPluralForms
= {}, explicitPluralFormNumber
, formIndex
, form
, count
;
1135 count
= parseFloat( this.language
.convertNumber( nodes
[0], true ) );
1136 forms
= nodes
.slice( 1 );
1137 for ( formIndex
= 0; formIndex
< forms
.length
; formIndex
++ ) {
1138 form
= forms
[formIndex
];
1140 if ( form
.jquery
&& form
.hasClass( 'mediaWiki_htmlEmitter' ) ) {
1141 // This is a nested node, may be an explicit plural form like 5=[$2 linktext]
1142 firstChild
= form
.contents().get( 0 );
1143 if ( firstChild
&& firstChild
.nodeType
=== Node
.TEXT_NODE
) {
1144 firstChildText
= firstChild
.textContent
;
1145 if ( /^\d+=/.test( firstChildText
) ) {
1146 explicitPluralFormNumber
= parseInt( firstChildText
.split( /=/ )[0], 10 );
1147 // Use the digit part as key and rest of first text node and
1148 // rest of child nodes as value.
1149 firstChild
.textContent
= firstChildText
.slice( firstChildText
.indexOf( '=' ) + 1 );
1150 explicitPluralForms
[explicitPluralFormNumber
] = form
;
1151 forms
[formIndex
] = undefined;
1154 } else if ( /^\d+=/.test( form
) ) {
1155 // Simple explicit plural forms like 12=a dozen
1156 explicitPluralFormNumber
= parseInt( form
.split( /=/ )[0], 10 );
1157 explicitPluralForms
[explicitPluralFormNumber
] = form
.slice( form
.indexOf( '=' ) + 1 );
1158 forms
[formIndex
] = undefined;
1162 // Remove explicit plural forms from the forms. They were set undefined in the above loop.
1163 forms
= $.map( forms
, function ( form
) {
1167 return this.language
.convertPlural( count
, forms
, explicitPluralForms
);
1171 * Transform parsed structure according to gender.
1173 * Usage: {{gender:[ mw.user object | '' | 'male' | 'female' | 'unknown' ] | masculine form | feminine form | neutral form}}.
1175 * The first node must be one of:
1176 * - the mw.user object (or a compatible one)
1177 * - an empty string - indicating the current user, same effect as passing the mw.user object
1178 * - a gender string ('male', 'female' or 'unknown')
1180 * @param {Array} nodes List of nodes, [ {string|mw.user}, {string}, {string}, {string} ]
1181 * @return {string} Selected gender form according to current language
1183 gender: function ( nodes
) {
1185 maybeUser
= nodes
[0],
1186 forms
= nodes
.slice( 1 );
1188 if ( maybeUser
=== '' ) {
1189 maybeUser
= mw
.user
;
1192 // If we are passed a mw.user-like object, check their gender.
1193 // Otherwise, assume the gender string itself was passed .
1194 if ( maybeUser
&& maybeUser
.options
instanceof mw
.Map
) {
1195 gender
= maybeUser
.options
.get( 'gender' );
1200 return this.language
.gender( gender
, forms
);
1204 * Transform parsed structure into grammar conversion.
1205 * Invoked by putting `{{grammar:form|word}}` in a message
1206 * @param {Array} nodes List of nodes [{Grammar case eg: genitive}, {string word}]
1207 * @return {string} selected grammatical form according to current language
1209 grammar: function ( nodes
) {
1210 var form
= nodes
[0],
1212 return word
&& form
&& this.language
.convertGrammar( word
, form
);
1216 * Tranform parsed structure into a int: (interface language) message include
1217 * Invoked by putting `{{int:othermessage}}` into a message
1218 * @param {Array} nodes List of nodes
1219 * @return {string} Other message
1221 'int': function ( nodes
) {
1222 return mw
.jqueryMsg
.getMessageFunction()( nodes
[0].toLowerCase() );
1226 * Takes an unformatted number (arab, no group separators and . as decimal separator)
1227 * and outputs it in the localized digit script and formatted with decimal
1228 * separator, according to the current language.
1229 * @param {Array} nodes List of nodes
1230 * @return {number|string} Formatted number
1232 formatnum: function ( nodes
) {
1233 var isInteger
= ( nodes
[1] && nodes
[1] === 'R' ) ? true : false,
1236 return this.language
.convertNumber( number
, isInteger
);
1240 // Deprecated! don't rely on gM existing.
1241 // The window.gM ought not to be required - or if required, not required here.
1242 // But moving it to extensions breaks it (?!)
1243 // Need to fix plugin so it could do attributes as well, then will be okay to remove this.
1244 // @deprecated since 1.23
1245 mw
.log
.deprecate( window
, 'gM', mw
.jqueryMsg
.getMessageFunction(), 'Use mw.message( ... ).parse() instead.' );
1250 * @see mw.jqueryMsg#getPlugin
1252 $.fn
.msg
= mw
.jqueryMsg
.getPlugin();
1254 // Replace the default message parser with jqueryMsg
1255 oldParser
= mw
.Message
.prototype.parser
;
1256 mw
.Message
.prototype.parser = function () {
1257 var messageFunction
;
1259 // TODO: should we cache the message function so we don't create a new one every time? Benchmark this maybe?
1260 // Caching is somewhat problematic, because we do need different message functions for different maps, so
1261 // we'd have to cache the parser as a member of this.map, which sounds a bit ugly.
1262 // Do not use mw.jqueryMsg unless required
1263 if ( this.format
=== 'plain' || !/\{\{|[\[<>]/.test( this.map
.get( this.key
) ) ) {
1264 // Fall back to mw.msg's simple parser
1265 return oldParser
.apply( this );
1268 messageFunction
= mw
.jqueryMsg
.getMessageFunction( {
1269 'messages': this.map
,
1270 // For format 'escaped', escaping part is handled by mediawiki.js
1271 'format': this.format
1273 return messageFunction( this.key
, this.parameters
);
1276 }( mediaWiki
, jQuery
) );