mediawiki-extensions-VisualEditor/modules/ve/ve.js

/*!
 * VisualEditor namespace.
 *
 * @copyright 2011-2013 VisualEditor Team and others; see AUTHORS.txt
 * @license The MIT License (MIT); see LICENSE.txt
 */

( function ( oo ) {
	var ve, hasOwn;

	/**
	 * Namespace for all VisualEditor classes, static methods and static properties.
	 * @class
	 * @singleton
	 */
	ve = {
		// List of instances of ve.ui.Surface
		'instances': []
	};

	/* Utility Functions */

	hasOwn = Object.prototype.hasOwnProperty;

	/* Static Methods */

	/**
	 * Create an object that inherits from another object.
	 *
	 * @method
	 * @until ES5: Object#create
	 * @inheritdoc Object#create
	 */
	ve.createObject = Object.create;

	/**
	 * @method
	 * @inheritdoc OO#inheritClass
	 */
	ve.inheritClass = oo.inheritClass;

	/**
	 * Checks if an object is an instance of one or more classes.
	 *
	 * @method
	 * @param {Object} subject Object to check
	 * @param {Function[]} classes Classes to compare with
	 * @return {boolean} Object inherits from one or more of the classes
	 */
	ve.isInstanceOfAny = function ( subject, classes ) {
		var i = classes.length;

		while ( classes[--i] ) {
			if ( subject instanceof classes[i] ) {
				return true;
			}
		}
		return false;
	};

	/**
	 * @method
	 * @inheritdoc OO#mixinClass
	 */
	ve.mixinClass = function ( targetFn, originFn ) {
		oo.mixinClass( targetFn, originFn );

		// Track mixins
		targetFn.mixins = targetFn.mixins || [];
		targetFn.mixins.push( originFn );
	};

	/**
	 * Check if a constructor or object contains a certain mixin.
	 *
	 * @param {Function|Object} a Class or object to check
	 * @param {Function} mixin Mixin to check for
	 * @returns {boolean} Class or object uses mixin
	 */
	ve.isMixedIn = function ( subject, mixin ) {
		// Traverse from instances to the constructor
		if ( $.type( subject ) !== 'function' ) {
			subject = subject.constructor;
		}
		return !!subject.mixins && subject.mixins.indexOf( mixin ) !== -1;
	};

	/**
	 * @method
	 * @inheritdoc OO#cloneObject
	 */
	ve.cloneObject = oo.cloneObject;

	/**
	 * @method
	 * @inheritdoc OO#cloneObject
	 */
	ve.getObjectValues = oo.getObjectValues;

	/**
	 * @method
	 * @until ES5: Object#keys
	 * @inheritdoc Object#keys
	 */
	ve.getObjectKeys = Object.keys;

	/**
	 * @method
	 * @inheritdoc OO#compare
	 */
	ve.compare = oo.compare;

	/**
	 * @method
	 * @inheritdoc OO#copy
	 */
	ve.copyArray = oo.copy;

	/**
	 * @method
	 * @inheritdoc OO#copy
	 */
	ve.copyObject = oo.copy;

	/**
	 * Copy an array of DOM elements, optionally into a different document.
	 *
	 * @param {HTMLElement[]} domElements DOM elements to copy
	 * @param {HTMLDocument} [doc] Document to create the copies in; if unset, simply clone each element
	 * @return {HTMLElement[]} Copy of domElements with copies of each element
	 */
	ve.copyDomElements = function ( domElements, doc ) {
		return domElements.map( function ( domElement ) {
			return doc ? doc.importNode( domElement, true ) : domElement.cloneNode( true );
		} );
	};

	/**
	 * Check to see if an object is a plain object (created using "{}" or "new Object").
	 *
	 * @method
	 * @source <http://api.jquery.com/jQuery.isPlainObject/>
	 * @param {Object} obj The object that will be checked to see if it's a plain object
	 * @returns {boolean}
	 */
	ve.isPlainObject = $.isPlainObject;

	/**
	 * Check to see if an object is empty (contains no properties).
	 *
	 * @method
	 * @source <http://api.jquery.com/jQuery.isEmptyObject/>
	 * @param {Object} obj The object that will be checked to see if it's empty
	 * @returns {boolean}
	 */
	ve.isEmptyObject = $.isEmptyObject;

	/**
	 * @method
	 * @until ES5: Array#isArray
	 * @inheritdoc Array#isArray
	 */
	ve.isArray = Array.isArray;

	/**
	 * Wrapper for Function#bind.
	 *
	 * Create a function that calls the given function in a certain context.
	 * If a function does not have an explicit context, it is determined at
	 * execution time based on how it is invoked (e.g. object member, call/apply,
	 * global scope, etc.).
	 *
	 * Performance optimization: <http://jsperf.com/function-bind-shim-perf>
	 *
	 * @method
	 * @source <http://api.jquery.com/jQuery.proxy/>
	 * @until ES5: Function#bind
	 * @param {Function} func Function to bind
	 * @param {Object} context Context for the function
	 * @param {Mixed...} [args] Variadic list of arguments to prepend to arguments
	 *   to the bound function
	 * @returns {Function} The bound
	 */
	ve.bind = $.proxy;

	/**
	 * Wrapper for Array#indexOf.
	 *
	 * Values are compared without type coercion.
	 *
	 * @method
	 * @source <http://api.jquery.com/jQuery.inArray/>
	 * @until ES5: Array#indexOf
	 * @param {Mixed} value Element to search for
	 * @param {Array} array Array to search in
	 * @param {number} [fromIndex=0] Index to being searching from
	 * @returns {number} Index of value in array, or -1 if not found
	 */
	ve.indexOf = $.inArray;

	/**
	 * Compute the union (duplicate-free merge) of a set of arrays.
	 *
	 * Arrays values must be convertable to object keys (strings)
	 *
	 * By building an object (with the values for keys) in parallel with
	 * the array, a new item's existence in the union can be computed faster
	 *
	 * @param {Array...} arrays Arrays to union
	 * @returns {Array} Union of the arrays
	 */
	ve.simpleArrayUnion = function () {
		var i, ilen, j, jlen, arr, obj = {}, result = [];
		for ( i = 0, ilen = arguments.length; i < ilen; i++ ) {
			arr = arguments[i];
			for ( j = 0, jlen = arr.length; j < jlen; j++ ) {
				if ( !obj[arr[j]] ) {
					obj[arr[j]] = true;
					result.push( arr[j] );
				}
			}
		}
		return result;
	};

	/**
	 * Compute the intersection of two arrays (items in both arrays).
	 *
	 * Arrays values must be convertable to object keys (strings)
	 *
	 * @param {Array} a First array
	 * @param {Array} b Second array
	 * @returns {Array} Intersection of arrays
	 */
	ve.simpleArrayIntersection = function ( a, b ) {
		return ve.simpleArrayCombine( a, b, true );
	};

	/**
	 * Compute the difference of two arrays (items in 'a' but not 'b').
	 *
	 * Arrays values must be convertable to object keys (strings)
	 *
	 * @param {Array} a First array
	 * @param {Array} b Second array
	 * @returns {Array} Intersection of arrays
	 */
	ve.simpleArrayDifference = function ( a, b ) {
		return ve.simpleArrayCombine( a, b, false );
	};

	/**
	 * Combine arrays (intersection or difference).
	 *
	 * An intersection checks the item exists in 'b' while difference checks it doesn't.
	 *
	 * Arrays values must be convertable to object keys (strings)
	 *
	 * By building an object (with the values for keys) of 'b' we can
	 * compute the result faster
	 *
	 * @param {Array} a First array
	 * @param {Array} b Second array
	 * @param {boolean} includeB Include items in 'b'
	 * @returns {Array} Combination (intersection or difference) of arrays
	 */
	ve.simpleArrayCombine = function ( a, b, includeB ) {
		var i, ilen, isInB, bObj = {}, result = [];
		for ( i = 0, ilen = b.length; i < ilen; i++ ) {
			bObj[b[i]] = true;
		}
		for ( i = 0, ilen = a.length; i < ilen; i++ ) {
			isInB = !!bObj[a[i]];
			if ( isInB === includeB ) {
				result.push( a[i] );
			}
		}
		return result;
	};

	/**
	 * Merge properties of one or more objects into another.
	 * Preserves original object's inheritance (e.g. Array, Object, whatever).
	 * In case of array or array-like objects only the indexed properties
	 * are copied over.
	 * Beware: If called with only one argument, it will consider
	 * 'target' as 'source' and 'this' as 'target'. Which means
	 * ve.extendObject( { a: 1 } ); sets ve.a = 1;
	 *
	 * @method
	 * @source <http://api.jquery.com/jQuery.extend/>
	 * @param {boolean} [recursive=false]
	 * @param {Mixed} [target] Object that will receive the new properties
	 * @param {Mixed...} [sources] Variadic list of objects containing properties
	 * to be merged into the targe.
	 * @returns {Mixed} Modified version of first or second argument
	 */
	ve.extendObject = $.extend;

	/**
	 * Generates a hash of an object based on its name and data.
	 * Performance optimization: http://jsperf.com/ve-gethash-201208#/toJson_fnReplacerIfAoForElse
	 *
	 * To avoid two objects with the same values generating different hashes, we utilize the replacer
	 * argument of JSON.stringify and sort the object by key as it's being serialized. This may or may
	 * not be the fastest way to do this; we should investigate this further.
	 *
	 * Objects and arrays are hashed recursively. When hashing an object that has a .getHash()
	 * function, we call that function and use its return value rather than hashing the object
	 * ourselves. This allows classes to define custom hashing.
	 *
	 * @param {Object} val Object to generate hash for
	 * @returns {string} Hash of object
	 */
	ve.getHash = function ( val ) {
		return JSON.stringify( val, ve.getHash.keySortReplacer );
	};

	/**
	 * Helper function for ve.getHash which sorts objects by key.
	 *
	 * This is a callback passed into JSON.stringify.
	 *
	 * @param {string} key Property name of value being replaced
	 * @param {Mixed} val Property value to replace
	 * @returns {Mixed} Replacement value
	 */
	ve.getHash.keySortReplacer = function ( key, val ) {
		var normalized, keys, i, len;
		if ( val && typeof val.getHashObject === 'function' ) {
			// This object has its own custom hash function, use it
			val = val.getHashObject();
		}
		if ( !ve.isArray( val ) && Object( val ) === val ) {
			// Only normalize objects when the key-order is ambiguous
			// (e.g. any object not an array).
			normalized = {};
			keys = ve.getObjectKeys( val ).sort();
			i = 0;
			len = keys.length;
			for ( ; i < len; i += 1 ) {
				normalized[keys[i]] = val[keys[i]];
			}
			return normalized;

		// Primitive values and arrays get stable hashes
		// by default. Lets those be stringified as-is.
		} else {
			return val;
		}
	};

	/**
	 * Splice one array into another.
	 *
	 * This is the equivalent of arr.splice( offset, remove, d1, d2, d3, ... ) except that arguments are
	 * specified as an array rather than separate parameters.
	 *
	 * This method has been proven to be faster than using slice and concat to create a new array, but
	 * performance tests should be conducted on each use of this method to verify this is true for the
	 * particular use. Also, browsers change fast, never assume anything, always test everything.
	 *
	 * Includes a replacement for broken implementation of Array.prototype.splice() found in Opera 12.
	 *
	 * @param {Array|ve.dm.BranchNode} arr Object supporting .splice() to remove from and insert into. Will be modified
	 * @param {number} offset Offset in arr to splice at. This may NOT be negative, unlike the
	 *  'index' parameter in Array#splice
	 * @param {number} remove Number of elements to remove at the offset. May be zero
	 * @param {Array} data Array of items to insert at the offset. May not be empty if remove=0
	 * @returns {Array} Array of items removed
	 */
	ve.batchSplice = ( function () {
		var arraySplice;

		// This yields 'true' on Opera 12.15.
		function isSpliceBroken() {
			var n = 256, a = [];
			a[n] = 'a';

			a.splice( n + 1, 0, 'b' );

			return a[n] !== 'a';
		}

		if ( !isSpliceBroken() ) {
			arraySplice = Array.prototype.splice;
		} else {
			// Standard Array.prototype.splice() function implemented using .slice() and .push().
			arraySplice = function ( offset, remove/*, data... */ ) {
				var data, begin, removed, end;

				data = Array.prototype.slice.call( arguments, 2 );

				begin = this.slice( 0, offset );
				removed = this.slice( offset, remove );
				end = this.slice( offset + remove );

				this.length = 0;
				// This polyfill only been discovered to be necessary on Opera
				// and it seems to handle up to 1048575 function parameters.
				this.push.apply( this, begin );
				this.push.apply( this, data );
				this.push.apply( this, end );

				return removed;
			};
		}

		return function ( arr, offset, remove, data ) {
			// We need to splice insertion in in batches, because of parameter list length limits which vary
			// cross-browser - 1024 seems to be a safe batch size on all browsers
			var splice, index = 0, batchSize = 1024, toRemove = remove, spliced, removed = [];

			splice = ve.isArray( arr ) ? arraySplice : arr.splice;

			if ( data.length === 0 ) {
				// Special case: data is empty, so we're just doing a removal
				// The code below won't handle that properly, so we do it here
				return splice.call( arr, offset, remove );
			}

			while ( index < data.length ) {
				// Call arr.splice( offset, remove, i0, i1, i2, ..., i1023 );
				// Only set remove on the first call, and set it to zero on subsequent calls
				spliced = splice.apply(
					arr, [index + offset, toRemove].concat( data.slice( index, index + batchSize ) )
				);
				if ( toRemove > 0 ) {
					removed = spliced;
				}
				index += batchSize;
				toRemove = 0;
			}
			return removed;
		};
	}() );

	/**
	 * Insert one array into another.
	 *
	 * This just a shortcut for `ve.batchSplice( dst, offset, 0, src )`.
	 *
	 * @see #batchSplice
	 */
	ve.insertIntoArray = function ( dst, offset, src ) {
		ve.batchSplice( dst, offset, 0, src );
	};

	/**
	 * Get a deeply nested property of an object using variadic arguments, protecting against
	 * undefined property errors.
	 *
	 * `quux = getProp( obj, 'foo', 'bar', 'baz' );` is equivalent to `quux = obj.foo.bar.baz;`
	 * except that the former protects against JS errors if one of the intermediate properties
	 * is undefined. Instead of throwing an error, this function will return undefined in
	 * that case.
	 *
	 * @param {Object} obj
	 * @param {Mixed...} [keys]
	 * @returns obj[arguments[1]][arguments[2]].... or undefined
	 */
	ve.getProp = function ( obj ) {
		var i, retval = obj;
		for ( i = 1; i < arguments.length; i++ ) {
			if ( retval === undefined || retval === null ) {
				// Trying to access a property of undefined or null causes an error
				return undefined;
			}
			retval = retval[arguments[i]];
		}
		return retval;
	};

	/**
	 * Set a deeply nested property of an object using variadic arguments, protecting against
	 * undefined property errors.
	 *
	 * `ve.setProp( obj, 'foo', 'bar', 'baz' );` is equivalent to `obj.foo.bar = baz;` except that
	 * the former protects against JS errors if one of the intermediate properties is
	 * undefined. Instead of throwing an error, undefined intermediate properties will be
	 * initialized to an empty object. If an intermediate property is null, or if obj itself
	 * is undefined or null, this function will silently abort.
	 *
	 * @param {Object} obj
	 * @param {Mixed...} [keys]
	 * @param {Mixed} [value]
	 */
	ve.setProp = function ( obj /*, keys ... , value */ ) {
		var i, prop = obj;
		if ( Object( obj ) !== obj ) {
			return;
		}
		for ( i = 1; i < arguments.length - 2; i++ ) {
			if ( prop[arguments[i]] === undefined ) {
				prop[arguments[i]] = {};
			}
			if ( prop[arguments[i]] === null || typeof prop[arguments[i]] !== 'object' ) {
				return;
			}
			prop = prop[arguments[i]];
		}
		prop[arguments[arguments.length - 2]] = arguments[arguments.length - 1];
	};

	/**
	 * Log data to the console.
	 *
	 * This implementation does nothing, to add a real implmementation ve.debug needs to be loaded.
	 *
	 * @param {Mixed...} [args] Data to log
	 */
	ve.log = function () {
		// don't do anything, this is just a stub
	};

	/**
	 * Log an object to the console.
	 *
	 * This implementation does nothing, to add a real implmementation ve.debug needs to be loaded.
	 *
	 * @param {Object} obj
	 */
	ve.dir = function () {
		// don't do anything, this is just a stub
	};

	/**
	 * Return a function, that, as long as it continues to be invoked, will not
	 * be triggered. The function will be called after it stops being called for
	 * N milliseconds. If `immediate` is passed, trigger the function on the
	 * leading edge, instead of the trailing.
	 *
	 * Ported from: http://underscorejs.org/underscore.js
	 *
	 * @param {Function} func
	 * @param {number} wait
	 * @param {boolean} immediate
	 * @returns {Function}
	 */
	ve.debounce = function ( func, wait, immediate ) {
		var timeout;
		return function () {
			var context = this,
				args = arguments,
				later = function () {
					timeout = null;
					if ( !immediate ) {
						func.apply( context, args );
					}
				};
			if ( immediate && !timeout ) {
				func.apply( context, args );
			}
			clearTimeout( timeout );
			timeout = setTimeout( later, wait );
		};
	};

	/**
	 * Move the selection to the end of an input.
	 *
	 * @param {HTMLElement} element Input element
	 */
	ve.selectEnd = function ( element ) {
		element.focus();
		if ( element.selectionStart !== undefined ) {
			element.selectionStart = element.selectionEnd = element.value.length;
		} else if ( element.createTextRange ) {
			var textRange = element.createTextRange();
			textRange.collapse( false );
			textRange.select();
		}
	};

	/**
	 * Get a localized message.
	 *
	 * @param {string} key Message key
	 * @param {Mixed...} [params] Message parameters
	 */
	ve.msg = function () {
		// Avoid using ve.bind because ve.init.platform doesn't exist yet.
		// TODO: Fix dependency issues between ve.js and ve.init.platform
		return ve.init.platform.getMessage.apply( ve.init.platform, arguments );
	};

    /**
     * @method
     * @inheritdoc unicodeJS.graphemebreak#splitClusters
     * @see unicodeJS.graphemebreak#splitClusters
     */
	ve.splitClusters = unicodeJS.graphemebreak.splitClusters;

	/**
	 * Determine if the text consists of only unattached combining marks.
	 *
	 * @param {string} text Text to test
	 * @returns {boolean} The text is unattached combining marks
	 */
	ve.isUnattachedCombiningMark = function ( text ) {
		return ( /^[\u0300-\u036F]+$/ ).test( text );
	};

	/**
	 * Convert a grapheme cluster offset to a byte offset.
	 *
	 * @param {string} text Text in which to calculate offset
	 * @param {number} clusterOffset Grapheme cluster offset
	 * @returns {number} Byte offset
	 */
	ve.getByteOffset = function ( text, clusterOffset ) {
		return ve.splitClusters( text ).slice( 0, clusterOffset ).join( '' ).length;
	};

	/**
	 * Convert a byte offset to a grapheme cluster offset.
	 *
	 * @param {string} text Text in which to calculate offset
	 * @param {number} byteOffset Byte offset
	 * @returns {number} Grapheme cluster offset
	 */
	ve.getClusterOffset = function ( text, byteOffset ) {
		return ve.splitClusters( text.substring( 0, byteOffset ) ).length;
	};

	/**
	 * Get a text substring, taking care not to split grapheme clusters.
	 *
	 * @param {string} text Text to take the substring from
	 * @param {number} start Start offset
	 * @param {number} end End offset
	 * @param {boolean} [outer=false] Include graphemes if the offset splits them
	 * @returns {string} Substring of text
	 */
	ve.graphemeSafeSubstring = function ( text, start, end, outer ) {
		// TODO: improve performance by incrementally inspecting characters around the offsets
		var unicodeStart = ve.getByteOffset( text, ve.getClusterOffset( text, start ) ),
			unicodeEnd = ve.getByteOffset( text, ve.getClusterOffset( text, end ) );

		// If the selection collapses and we want an inner, then just return empty
		// otherwise we'll end up crossing over start and end
		if ( unicodeStart === unicodeEnd && !outer ) {
			return '';
		}

		// The above calculations always move to the right of a multibyte grapheme.
		// Depending on the outer flag, we may want to move to the left:
		if ( unicodeStart > start && outer ) {
			unicodeStart = ve.getByteOffset( text, ve.getClusterOffset( text, start ) - 1 );
		}
		if ( unicodeEnd > end && !outer ) {
			unicodeEnd = ve.getByteOffset( text, ve.getClusterOffset( text, end ) - 1 );
		}
		return text.substring( unicodeStart, unicodeEnd );
	};

	/**
	 * Escape non-word characters so they can be safely used as HTML attribute values.
	 *
	 * This method is basically a copy of `mw.html.escape`.
	 *
	 * @see #escapeHtml_escapeHtmlCharacter
	 * @param {string} value Attribute value to escape
	 * @returns {string} Escaped attribute value
	 */
	ve.escapeHtml = function ( value ) {
		return value.replace( /['"<>&]/g, ve.escapeHtml.escapeHtmlCharacter );
	};

	/**
	 * Helper function for #escapeHtml to escape a character for use in HTML.
	 *
	 * This is a callback intended to be passed to String#replace.
	 *
	 * @method escapeHtml_escapeHtmlCharacter
	 * @private
	 * @param {string} key Property name of value being replaced
	 * @returns {string} Escaped charcater
	 */
	ve.escapeHtml.escapeHtmlCharacter = function ( value ) {
		switch ( value ) {
			case '\'':
				return '&#039;';
			case '"':
				return '&quot;';
			case '<':
				return '&lt;';
			case '>':
				return '&gt;';
			case '&':
				return '&amp;';
			default:
				return value;
		}
	};

	/**
	 * Generate an opening HTML tag.
	 *
	 * This method copies part of `mw.html.element` from MediaWiki.
	 *
	 * NOTE: While the values of attributes are escaped, the tag name and the names of
	 * attributes (i.e. the keys in the attributes objects) are NOT ESCAPED. The caller is
	 * responsible for making sure these are sane tag/attribute names and do not contain
	 * unsanitized content from an external source (e.g. from the user or from the web).
	 *
	 * @param {string} tag HTML tag name
	 * @param {Object} attributes Key-value map of attributes for the tag
	 * @returns {string} Opening HTML tag
	 */
	ve.getOpeningHtmlTag = function ( tagName, attributes ) {
		var html, attrName, attrValue;
		html = '<' + tagName;
		for ( attrName in attributes ) {
			attrValue = attributes[attrName];
			if ( attrValue === true ) {
				// Convert name=true to name=name
				attrValue = attrName;
			} else if ( attrValue === false ) {
				// Skip name=false
				continue;
			}
			html += ' ' + attrName + '="' + ve.escapeHtml( String( attrValue ) ) + '"';
		}
		html += '>';
		return html;
	};

	/**
	 * Get the attributes of a DOM element as an object with key/value pairs.
	 *
	 * @param {HTMLElement} element
	 * @returns {Object}
	 */
	ve.getDomAttributes = function ( element ) {
		var result = {}, i;
		for ( i = 0; i < element.attributes.length; i++ ) {
			result[element.attributes[i].name] = element.attributes[i].value;
		}
		return result;
	};

	/**
	 * Set the attributes of a DOM element as an object with key/value pairs.
	 *
	 * @param {HTMLElement} element DOM element to apply attributes to
	 * @param {Object} attributes Attributes to apply
	 * @param {string[]} [whitelist] List of attributes to exclusively allow (all lower case names)
	 */
	ve.setDomAttributes = function ( element, attributes, whitelist ) {
		var key;
		// Duck-typing for attribute setting
		if ( !element.setAttribute || !element.removeAttribute ) {
			return;
		}
		for ( key in attributes ) {
			if ( attributes[key] === undefined || attributes[key] === null ) {
				element.removeAttribute( key );
			} else {
				if ( whitelist && whitelist.indexOf( key.toLowerCase() ) === -1 ) {
					continue;
				}
				element.setAttribute( key, attributes[key] );
			}
		}
	};

	/**
	 * Build a summary of an HTML element.
	 *
	 * Summaries include node name, text, attributes and recursive summaries of children.
	 * Used for serializing or comparing HTML elements.
	 *
	 * @private
	 * @param {HTMLElement} element Element to summarize
	 * @returns {Object} Summary of element.
	 */
	ve.getDomElementSummary = function ( element ) {
		var i,
			$element = $( element ),
			summary = {
				'type': element.nodeName.toLowerCase(),
				'text': $element.text(),
				'attributes': {},
				'children': []
			};

		// Gather attributes
		if ( element.attributes ) {
			for ( i = 0; i < element.attributes.length; i++ ) {
				summary.attributes[element.attributes[i].name] = element.attributes[i].value;
			}
		}
		// Summarize children
		if ( element.childNodes ) {
			for ( i = 0; i < element.childNodes.length; i++ ) {
				if ( element.childNodes[i].nodeType !== Node.TEXT_NODE ) {
					summary.children.push( ve.getDomElementSummary( element.childNodes[i] ) );
				}
			}
		}
		return summary;
	};

	/**
	 * Callback for #copyArray and #copyObject to convert nodes to a comparable summary.
	 *
	 * @private
	 * @param {Object} value Value in the object/array
	 * @returns {Object} DOM element summary if value is a node, otherwise just the value
	 */
	ve.convertDomElements = function ( value ) {
		// Use duck typing rather than instanceof Node; the latter doesn't always work correctly
		if ( value && value.nodeType ) {
			return ve.getDomElementSummary( value );
		}
		return value;
	};

	/**
	 * Check whether a given DOM element is of a block or inline type.
	 *
	 * @param {HTMLElement} element
	 * @returns {boolean} True if element is block, false if it is inline
	 */
	ve.isBlockElement = function ( element ) {
		return ve.isBlockElementType( element.nodeName.toLowerCase() );
	};

	/**
	 * Check whether a given tag name is a block or inline tag.
	 *
	 * @param {string} nodeName All-lowercase HTML tag name
	 * @returns {boolean} True if block, false if inline
	 */
	ve.isBlockElementType = function ( nodeName ) {
		return ve.indexOf( nodeName, ve.isBlockElementType.blockTypes ) !== -1;
	};

	/**
	 * Private data for #isBlockElementType.
	 *
	 */
	ve.isBlockElementType.blockTypes = [
		'div', 'p',
		// tables
		'table', 'tbody', 'thead', 'tfoot', 'caption', 'th', 'tr', 'td',
		// lists
		'ul', 'ol', 'li', 'dl', 'dt', 'dd',
		// HTML5 heading content
		'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'hgroup',
		// HTML5 sectioning content
		'article', 'aside', 'body', 'nav', 'section', 'footer', 'header', 'figure',
		'figcaption', 'fieldset', 'details', 'blockquote',
		// other
		'hr', 'button', 'canvas', 'center', 'col', 'colgroup', 'embed',
		'map', 'object', 'pre', 'progress', 'video'
	];

	/**
	 * Create an HTMLDocument from an HTML string.
	 *
	 * The html parameter is supposed to be a full HTML document with a doctype and an `<html>` tag.
	 * If you pass a document fragment, it may or may not work, this is at the mercy of the browser.
	 *
	 * To create an empty document, pass the empty string.
	 *
	 * @param {string} html HTML string
	 * @returns {HTMLDocument} Document constructed from the HTML string
	 */
	ve.createDocumentFromHtml = function ( html ) {
		// Here's how this function should look:
		//
		//     var newDocument = document.implementation.createHtmlDocument( '' );
		//     newDocument.open();
		//     newDocument.write( html );
		//     newDocument.close();
		//     return newDocument;
		//
		// (Or possibly something involving DOMParser.prototype.parseFromString, but that's Firefox-only
		// for now.)
		//
		// Sadly, it's impossible:
		// * On IE 9, calling open()/write() on such a document throws an "Unspecified error" (sic).
		// * On Firefox 20, calling open()/write() doesn't actually do anything, including writing.
		//   This is reported as Firefox bug 867102.
		// * On Opera 12, calling open()/write() behaves as if called on window.document, replacing the
		//   entire contents of the page with new HTML. This is reported as Opera bug DSK-384486.
		//
		// Funnily, in all of those browsers it's apparently perfectly legal and possible to access the
		// newly created document's DOM itself, including modifying documentElement's innerHTML, which
		// would achieve our goal. But that requires some nasty magic to strip off the <html></html> tag
		// itself, so we're not doing that. (We can't use .outerHTML, either, as the spec disallows
		// assigning to it for the root element.)
		//
		// There is one more way - create an <iframe>, append it to current document, and access its
		// contentDocument. The only browser having issues with that is Opera (sometimes the accessible
		// value is not actually a Document, but something which behaves just like an empty regular
		// object...), so we're detecting that and using the innerHTML hack described above.

		// Create an invisible iframe
		var newDocument, $iframe = $( '<iframe frameborder="0" width="0" height="0" />'),
			iframe = $iframe.get( 0 );
		// Attach it to the document. We have to do this to get a new document out of it
		document.documentElement.appendChild( iframe );
		// Write the HTML to it
		newDocument = ( iframe.contentWindow && iframe.contentWindow.document ) || iframe.contentDocument;
		newDocument.open();
		newDocument.write( html ); // Party like it's 1995!
		newDocument.close();
		// Detach the iframe
		// FIXME detaching breaks access to newDocument in IE
		iframe.parentNode.removeChild( iframe );

		if ( !newDocument.documentElement || newDocument.documentElement.cloneNode() === undefined ) {
			// Surprise! The document is not a document! Only happens on Opera.
			// (Or its nodes are not actually nodes, while the document
			// *is* a document. This only happens when debugging with Dragonfly.)
			newDocument = document.implementation.createHTMLDocument( '' );
			// Carefully unwrap the HTML out of the root node (and doctype, if any).
			// <html> might have some arguments here, but they're apparently not important.
			html = html.replace(/^\s*(?:<!doctype[^>]*>)?\s*<html[^>]*>/i, '' );
			html = html.replace(/<\/html>\s*$/i, '' );
			newDocument.documentElement.innerHTML = html;
		}

		return newDocument;
	};

	/**
	 * Get the actual inner HTML of a DOM node.
	 *
	 * In most browsers, .innerHTML is broken and eats newlines in `<pre>` elements, see
	 * https://bugzilla.mozilla.org/show_bug.cgi?id=838954 . This function detects this behavior
	 * and works around it, to the extent possible. `<pre>\nFoo</pre>` will become `<pre>Foo</pre>`
	 * if the browser is broken, but newlines are preserved in all other cases.
	 *
	 * @param {HTMLElement} element HTML element to get inner HTML of
	 * @returns {string} Inner HTML
	 */
	ve.properInnerHtml = function ( element ) {
		return ve.fixupPreBug( element ).innerHTML;
	};

	/**
	 * Get the actual outer HTML of a DOM node.
	 *
	 * @see ve#properInnerHtml
	 * @param {HTMLElement} element HTML element to get outer HTML of
	 * @returns {string} Outer HTML
	 */
	ve.properOuterHtml = function ( element ) {
		return ve.fixupPreBug( element ).outerHTML;
	};

	/**
	 * Helper function for ve#properInnerHtml and #properOuterHtml.
	 *
	 * Detect whether the browser has broken `<pre>` serialization, and if so return a clone
	 * of the node with extra newlines added to make it serialize properly. If the browser is not
	 * broken, just return the original node.
	 *
	 * @param {HTMLElement} element HTML element to fix up
	 * @returns {HTMLElement} Either element, or a fixed-up clone of it
	 */
	ve.fixupPreBug = function ( element ) {
		var div, $element;
		if ( ve.isPreInnerHtmlBroken === undefined ) {
			// Test whether newlines in `<pre>` are serialized back correctly
			div = document.createElement( 'div' );
			div.innerHTML = '<pre>\n\n</pre>';
			ve.isPreInnerHtmlBroken = div.innerHTML === '<pre>\n</pre>';
		}

		if ( !ve.isPreInnerHtmlBroken ) {
			return element;
		}

		// Workaround for bug 42469: if a `<pre>` starts with a newline, that means .innerHTML will
		// screw up and stringify it with one fewer newline. Work around this by adding a newline.
		// If we don't see a leading newline, we still don't know if the original HTML was
		// `<pre>Foo</pre>` or `<pre>\nFoo</pre>` , but that's a syntactic difference, not a
		// semantic one, and handling that is Parsoid's job.
		$element = $( element ).clone();
		$element.find( 'pre, textarea, listing' ).each( function () {
			var matches;
			if ( this.firstChild && this.firstChild.nodeType === Node.TEXT_NODE ) {
				matches = this.firstChild.data.match( /^(\r\n|\r|\n)/ );
				if ( matches && matches[1] ) {
					// Prepend a newline exactly like the one we saw
					this.firstChild.insertData( 0, matches[1] );
				}
			}
		} );
		return $element.get( 0 );
	};

	// Add more as you need
	ve.Keys = {
		'UNDEFINED': 0,
		'BACKSPACE': 8,
		'DELETE': 46,
		'LEFT': 37,
		'RIGHT': 39,
		'UP': 38,
		'DOWN': 40,
		'ENTER': 13,
		'END': 35,
		'HOME': 36,
		'TAB': 9,
		'PAGEUP': 33,
		'PAGEDOWN': 34,
		'ESCAPE': 27,
		'SHIFT': 16
	};

	// Expose
	window.ve = ve;
}( OO ) );