mediawiki-extensions-Visual.../modules/ve/dm/ve.dm.TransactionProcessor.js
Catrope 7465b670e1 Add and update an offset map in DocumentNode
This has some TODOs still but I want to land it now anyway, and fix the
TODOs later.

* Add this.offsetMap which maps each linear model offset to a model tree node
* Refactor createNodesFromData()
** Rename it to buildSubtreeFromData()
** Have it build an offset map as well as a node subtree
** Have it set the root on the fake root node so that when the subtree
   is attached to the main tree later, we don't get a rippling root
   update all the way down
** Normalize the way the loop processes content, that way adding offsets
   for content is easier
* Add rebuildNodes() which uses buildSubtreeFromData() to rebuild stuff
* Use rebuildNodes() in DocumentSynchronizer
* Use pushRebuild() in TransactionProcessor
* Optimize setRoot() for the case where the root is already set correctly

Change-Id: I8b827d0823c969e671615ddd06e5f1bd70e9d54c
2012-04-13 16:46:02 -07:00

599 lines
19 KiB
JavaScript

/**
* Creates an ve.dm.TransactionProcessor object.
*
* @class
* @constructor
*/
ve.dm.TransactionProcessor = function( model, transaction ) {
this.model = model;
this.transaction = transaction;
this.cursor = 0;
this.set = [];
this.clear = [];
};
/* Static Members */
ve.dm.TransactionProcessor.operationMap = {
// Retain
'retain': {
'commit': function( op ) {
this.retain( op );
},
'rollback': function( op ) {
this.retain( op );
}
},
// Insert
'insert': {
'commit': function( op ) {
this.insert( op );
},
'rollback': function( op ) {
this.remove( op );
}
},
// Remove
'remove': {
'commit': function( op ) {
this.remove( op );
},
'rollback': function( op ) {
this.insert( op );
}
},
'replace': {
'commit': function( op ) {
this.replace( op );
},
'rollback': function( op ) {
this.replace( op );
}
},
// Change element attributes
'attribute': {
'commit': function( op ) {
this.attribute( op );
},
'rollback': function( op ) {
this.attribute( op );
}
},
// Change content annotations
'annotate': {
'commit': function( op ) {
this.mark( op );
},
'rollback': function( op ) {
this.mark( op );
}
}
};
/* Static Methods */
ve.dm.TransactionProcessor.commit = function( doc, transaction ) {
var tp = new ve.dm.TransactionProcessor( doc, transaction );
tp.process( 'commit' );
};
ve.dm.TransactionProcessor.rollback = function( doc, transaction ) {
var tp = new ve.dm.TransactionProcessor( doc, transaction );
tp.process( 'rollback' );
};
/* Methods */
ve.dm.TransactionProcessor.prototype.nextOperation = function() {
return this.operations[this.operationIndex++] || false;
};
ve.dm.TransactionProcessor.prototype.executeOperation = function( op ) {
if ( op.type in ve.dm.TransactionProcessor.operationMap ) {
ve.dm.TransactionProcessor.operationMap[op.type][this.method].call( this, op );
} else {
throw 'Invalid operation error. Operation type is not supported: ' + operation.type;
}
};
ve.dm.TransactionProcessor.prototype.process = function( method ) {
var op;
this.synchronizer = new ve.dm.DocumentSynchronizer( this.model );
// Store the method (commit or rollback) and the operations array so executeOperation()
// can access them easily
this.method = method;
this.operations = this.transaction.getOperations();
// This loop is factored this way to allow operations to be skipped over or executed
// from within other operations
this.operationIndex = 0;
while ( ( op = this.nextOperation() ) ) {
this.executeOperation( op );
}
this.synchronizer.synchronize();
};
/**
* Get the parent node that would be affected by inserting given data into it's child.
*
* This is used when inserting data that closes and reopens one or more parent nodes into a child
* node, which requires rebuilding at a higher level.
*
* @method
* @param {ve.Node} node Child node to start from
* @param {Array} data Data to inspect for closings
* @returns {ve.Node} Lowest level parent node being affected
*/
ve.dm.TransactionProcessor.prototype.getScope = function( node, data ) {
var i,
length,
level = 0,
max = 0;
for ( i = 0, length = data.length; i < length; i++ ) {
if ( typeof data[i].type === 'string' ) {
level += data[i].type.charAt( 0 ) === '/' ? 1 : -1;
max = Math.max( max, level );
}
}
if ( max > 0 ) {
for ( i = 0; i < max; i++ ) {
node = node.getParent() || node;
}
}
return node;
};
ve.dm.TransactionProcessor.prototype.applyAnnotations = function( to, update ) {
var i,
j,
k,
length,
annotation,
changes = 0,
index;
// Handle annotations
if ( this.set.length ) {
for ( i = 0, length = this.set.length; i < length; i++ ) {
annotation = this.set[i];
// Auto-build annotation hash
if ( annotation.hash === undefined ) {
annotation.hash = ve.dm.DocumentNode.getHash( annotation );
}
for ( j = this.cursor; j < to; j++ ) {
// Auto-convert to array
if ( ve.isArray( this.model.data[j] ) ) {
this.model.data[j].push( annotation );
} else {
this.model.data[j] = [this.model.data[j], annotation];
}
}
}
changes++;
}
if ( this.clear.length ) {
for ( i = 0, length = this.clear.length; i < length; i++ ) {
annotation = this.clear[i];
if ( annotation instanceof RegExp ) {
for ( j = this.cursor; j < to; j++ ) {
var matches = ve.dm.DocumentNode.getMatchingAnnotations(
this.model.data[j], annotation
);
for ( k = 0; k < matches.length; k++ ) {
index = this.model.data[j].indexOf( matches[k] );
if ( index !== -1 ) {
this.model.data[j].splice( index, 1 );
}
}
// Auto-convert to string
if ( this.model.data[j].length === 1 ) {
this.model.data[j] = this.model.data[j][0];
}
}
} else {
// Auto-build annotation hash
if ( annotation.hash === undefined ) {
annotation.hash = ve.dm.DocumentNode.getHash( annotation );
}
for ( j = this.cursor; j < to; j++ ) {
index = ve.dm.DocumentNode.getIndexOfAnnotation(
this.model.data[j], annotation
);
if ( index !== -1 ) {
this.model.data[j].splice( index, 1 );
}
// Auto-convert to string
if ( this.model.data[j].length === 1 ) {
this.model.data[j] = this.model.data[j][0];
}
}
}
}
changes++;
}
if ( update && changes ) {
var fromNode = this.model.getNodeFromOffset( this.cursor ),
toNode = this.model.getNodeFromOffset( to );
this.model.traverseLeafNodes( function( node ) {
node.emit( 'update' );
if ( node === toNode ) {
return false;
}
}, fromNode );
}
};
ve.dm.TransactionProcessor.prototype.retain = function( op ) {
this.applyAnnotations( this.cursor + op.length, true );
this.cursor += op.length;
};
ve.dm.TransactionProcessor.prototype.insert = function( op ) {
var node,
index,
offset,
scope;
node = this.model.getNodeFromOffset( this.cursor );
// Shortcut 1: we're inserting content. We don't need to bother with any structural stuff
if ( !ve.dm.DocumentNode.containsElementData( op.data ) ) {
// TODO should we check whether we're at a structural offset, and throw an exception
// if that's the case? Or can we assume that the transaction is valid at this point?
// Insert data into linear model
ve.insertIntoArray( this.model.data, this.cursor, op.data );
this.applyAnnotations( this.cursor + op.data.length );
// Update the length of the containing node
this.synchronizer.pushResize( node, op.data.length );
// Move the cursor
this.cursor += op.data.length;
// All done
return;
}
// Determine the scope of the inserted data. If the data is an enclosed piece of structure,
// this will return node. Otherwise, the data closes one or more nodes, and this will return
// the first ancestor of node that isn't closed, which is the node that will contain the
// inserted data entirely.
scope = this.getScope( node, op.data );
// Shortcut 2: we're inserting an enclosed piece of structural data at a structural offset
// that isn't the end of the document.
// TODO why can't it be at the end of the document?
if (
ve.dm.DocumentNode.isStructuralOffset( this.model.data, this.cursor ) &&
this.cursor != this.model.data.length &&
scope == node
) {
// We're inserting an enclosed element into something else, so we don't have to rebuild
// the parent node. Just build a node from the inserted data and stick it in
ve.insertIntoArray( this.model.data, this.cursor, op.data );
this.applyAnnotations( this.cursor + op.data.length );
offset = this.model.getOffsetFromNode( node );
index = node.getIndexFromOffset( this.cursor - offset );
this.synchronizer.pushRebuild( new ve.Range( this.cursor, this.cursor ),
new ve.Range( this.cursor, this.cursor + op.data.length ) );
} else {
// This is the non-shortcut case
// Rebuild all children of scope, which is the node that encloses everything we might have to rebuild
node = scope.getChildren()[0];
offset = this.model.getOffsetFromNode( node );
if ( offset === -1 ) {
throw 'Invalid offset error. Node is not in model tree';
}
// Perform insert on linear data model
ve.insertIntoArray( this.model.data, this.cursor, op.data );
this.applyAnnotations( this.cursor + op.data.length );
// Synchronize model tree
this.synchronizer.pushRebuild( new ve.Range( offset, offset + scope.getContentLength() ),
new ve.Range( offset, offset + scope.getContentLength() + op.data.length ) );
}
this.cursor += op.data.length;
};
ve.dm.TransactionProcessor.prototype.remove = function( op ) {
if ( ve.dm.DocumentNode.containsElementData( op.data ) ) {
// TODO rewrite all this
// Figure out which nodes are covered by the removal
var ranges = this.model.selectNodes(
new ve.Range( this.cursor, this.cursor + op.data.length )
);
// Build the list of nodes to rebuild and the data to keep
var oldNodes = [],
newData = [],
parent = null,
index = null,
firstKeptNode,
lastKeptNode,
i;
for ( i = 0; i < ranges.length; i++ ) {
oldNodes.push( ranges[i].node );
if ( ranges[i].range !== undefined ) {
// We have to keep part of this node
if ( firstKeptNode === undefined ) {
// This is the first node we're keeping
firstKeptNode = ranges[i].node;
}
// Compute the start and end offset of this node
// We could do that with getOffsetFromNode() but
// we already have all the numbers we need so why would we
var startOffset = ranges[i].globalRange.start - ranges[i].range.start,
endOffset = startOffset + ranges[i].node.getContentLength(),
// Get this node's data
nodeData = this.model.data.slice( startOffset, endOffset );
// Remove data covered by the range from nodeData
nodeData.splice(
ranges[i].range.start, ranges[i].range.end - ranges[i].range.start
);
// What remains in nodeData is the data we need to keep
// Append it to newData
newData = newData.concat( nodeData );
lastKeptNode = ranges[i].node;
}
}
// Surround newData with the right openings and closings if needed
if ( firstKeptNode !== undefined ) {
// There are a number of conceptually different cases here,
// but the algorithm for dealing with them is the same.
// 1. Removal within one node: firstKeptNode === lastKeptNode
// 2. Merge of siblings: firstKeptNode.getParent() === lastKeptNode.getParent()
// 3. Merge of arbitrary depth: firstKeptNode and lastKeptNode have a common ancestor
// Because #1 and #2 are special cases of #3 (merges with depth=0 and depth=1,
// respectively), the code below that deals with the general case (#3) and automatically
// covers #1 and #2 that way as well.
// Simultaneously traverse upwards from firstKeptNode and lastKeptNode
// to find the common ancestor. On our way up, keep the element of each
// node we visit and verify that the transaction is a valid merge (i.e. it satisfies
// the merge criteria in prepareRemoval()'s canMerge()).
// FIXME: The code is essentially the same as canMerge(), merge these algorithms
var openings = [],
closings = [],
paths = ve.Node.getCommonAncestorPaths( firstKeptNode, lastKeptNode ),
prevN1,
prevN2;
if ( !paths ) {
throw 'Removal is not a valid merge: ' +
'nodes do not have a common ancestor or are not at the same depth';
}
for ( i = 0; i < paths.node1Path.length; i++ ) {
// Verify the element types are equal
if ( paths.node1Path[i].getElementType() !== paths.node2Path[i].getElementType() ) {
throw 'Removal is not a valid merge: ' +
'corresponding parents have different types ( ' +
paths.node1Path[i].getElementType() + ' vs ' +
paths.node2Path[i].getElementType() + ' )';
}
// Record the opening of n1 and the closing of n2
openings.push( paths.node1Path[i].getElement() );
closings.push( { 'type': '/' + paths.node2Path[i].getElementType() } );
}
// Surround newData with the openings and closings
newData = openings.reverse().concat( newData, closings );
// Rebuild oldNodes' ancestors if needed
// This only happens for merges with depth > 1
prevN1 = paths.node1Path.length ? paths.node1Path[paths.node1Path.length - 1] : null;
prevN2 = paths.node2Path.length ? paths.node2Path[paths.node2Path.length - 1] : null;
if ( prevN1 && prevN1 !== oldNodes[0] ) {
oldNodes = [ prevN1 ];
parent = paths.commonAncestor;
index = parent.indexOf( prevN1 ); // Pass to rebuildNodes() so it's not recomputed
if ( index === -1 ) {
throw "Tree corruption detected: node isn't in its parent's children array";
}
var foundPrevN2 = false;
for ( var j = index + 1; !foundPrevN2 && j < parent.getChildren().length; j++ ) {
oldNodes.push( parent.getChildren()[j] );
foundPrevN2 = parent.getChildren()[j] === prevN2;
}
if ( !foundPrevN2 ) {
throw "Tree corruption detected: node isn't in its parent's children array";
}
}
}
// Update the linear model
this.model.data.splice( this.cursor, op.data.length );
// Perform the rebuild. This updates the model tree
// TODO index of oldNodes[0] in its parent should be computed
// on the go by selectNodes() above
if ( parent == null ) {
parent = oldNodes[0].getParent();
}
if ( index == null ) {
index = parent.indexOf( oldNodes[0] );
}
// TODO better offset computation
// TODO allow direct parameter passing in pushRebuild()
var startOffset = this.model.getOffsetFromNode( oldNodes[0] );
var endOffset = this.model.getOffsetFromNode( oldNodes[oldNodes.length-1] ) +
oldNodes[oldNodes.length-1].getElementLength();
this.synchronizer.pushRebuild( new ve.Range( startOffset, endOffset ),
new ve.Range( startOffset, startOffset + newData.length ) );
} else {
// We're removing content only. Take a shortcut
// Get the node we are removing content from
var node = this.model.getNodeFromOffset( this.cursor );
// Update the linear model
this.model.data.splice( this.cursor, op.data.length );
// Queue a resize
this.synchronizer.pushResize( node, -op.data.length );
}
};
ve.dm.TransactionProcessor.prototype.replace = function( op ) {
var invert = this.method == 'rollback',
remove = invert ? op.replacement : op.remove,
replacement = invert ? op.remove : op.replacement,
removeHasStructure = ve.dm.DocumentNode.containsElementData( remove ),
replacementHasStructure = ve.dm.DocumentNode.containsElementData( replacement ),
node;
// remove is provided only for OT / conflict resolution and for
// reversibility, we don't actually verify it here
// Tree synchronization
// Figure out if this is a structural replacement or a content replacement
if ( !ve.dm.DocumentNode.containsElementData( remove ) && !ve.dm.DocumentNode.containsElementData( replacement ) ) {
// Content replacement
// Update the linear model
ve.batchedSplice( this.model.data, this.cursor, remove.length, replacement );
this.applyAnnotations( this.cursor + replacement.length );
// Get the node containing the replaced content
node = this.model.getNodeFromOffset( this.cursor );
// Queue a resize for this node
this.synchronizer.pushResize( node, replacement.length - remove.length );
// Advance the cursor
this.cursor += replacement.length;
} else {
// Structural replacement
// TODO generalize for insert/remove
// It's possible that multiple replace operations are needed before the
// model is back in a consistent state. This loop applies the current
// replace operation to the linear model, then keeps applying subsequent
// operations until the model is consistent. We keep track of the changes
// and queue a single rebuild after the loop finishes.
var operation = op,
removeLevel = 0,
replaceLevel = 0,
startOffset = this.cursor,
adjustment = 0,
i,
type;
while ( true ) {
if ( operation.type == 'replace' ) {
var opRemove = invert ? operation.replacement : operation.remove,
opReplacement = invert ? operation.remove : operation.replacement;
// Update the linear model for this replacement
ve.batchedSplice( this.model.data, this.cursor, opRemove.length, opReplacement );
this.cursor += opReplacement.length;
adjustment += opReplacement.length - opRemove.length;
// Walk through the remove and replacement data
// and keep track of the element depth change (level)
// for each of these two separately. The model is
// only consistent if both levels are zero.
for ( i = 0; i < opRemove.length; i++ ) {
type = opRemove[i].type;
if ( type === undefined ) {
// This is content, ignore
} else if ( type.charAt( 0 ) === '/' ) {
// Closing element
removeLevel--;
} else {
// Opening element
removeLevel++;
}
}
for ( i = 0; i < opReplacement.length; i++ ) {
type = opReplacement[i].type;
if ( type === undefined ) {
// This is content, ignore
} else if ( type.charAt( 0 ) === '/' ) {
// Closing element
replaceLevel--;
} else {
// Opening element
replaceLevel++;
}
}
} else {
// We're assuming that other operations will not cause
// adjustments.
// TODO actually make this the case by folding insert
// and delete into replace
this.executeOperation( operation );
}
if ( removeLevel === 0 && replaceLevel === 0 ) {
// The model is back in a consistent state, so we're done
break;
}
// Get the next operation
operation = this.nextOperation();
if ( !operation ) {
throw 'Unbalanced set of replace operations found';
}
}
// Queue a rebuild for the replaced node
this.synchronizer.pushRebuild( new ve.Range( startOffset, this.cursor - adjustment ),
new ve.Range( startOffset, this.cursor ) );
}
};
ve.dm.TransactionProcessor.prototype.attribute = function( op ) {
var invert = this.method == 'rollback',
element = this.model.data[this.cursor];
if ( element.type === undefined ) {
throw 'Invalid element error. Can not set attributes on non-element data.';
}
var to = invert ? op.from : op.to;
if ( to === undefined ) {
// Clear
if ( element.attributes ) {
delete element.attributes[op.key];
}
// Automatically clean up attributes object
var empty = true;
for ( var key in element.attributes ) {
empty = false;
break;
}
if ( empty ) {
delete element.attributes;
}
} else {
// Automatically initialize attributes object
if ( !element.attributes ) {
element.attributes = {};
}
// Set
element.attributes[op.key] = to;
}
var node = this.model.getNodeFromOffset( this.cursor + 1 );
if ( node.hasChildren() ) {
node.traverseLeafNodes( function( leafNode ) {
leafNode.emit( 'update' );
} );
} else {
node.emit( 'update' );
}
};
ve.dm.TransactionProcessor.prototype.mark = function( op ) {
var invert = this.method == 'rollback',
target;
if ( ( op.method === 'set' && !invert ) || ( op.method === 'clear' && invert ) ) {
target = this.set;
} else if ( ( op.method === 'clear' && !invert ) || ( op.method === 'set' && invert ) ) {
target = this.clear;
} else {
throw 'Invalid method error. Can not operate attributes this way: ' + method;
}
if ( op.bias === 'start' ) {
target.push( op.annotation );
} else if ( op.bias === 'stop' ) {
var index;
if ( op.annotation instanceof RegExp ) {
index = target.indexOf( op.annotation );
} else {
index = ve.dm.DocumentNode.getIndexOfAnnotation( target, op.annotation );
}
if ( index === -1 ) {
throw 'Annotation stack error. Annotation is missing.';
}
target.splice( index, 1 );
}
};