/// <summary>
/// Recursively searches the composite node for fields and constructs a hierarchical organization of those fields.
/// </summary>
/// <param name="compositeNode">A node that can be composed of other nodes.</param>
/// <param name="wordFields">A hierarchical list of the Word fields found in the composite node or it's descendants.</param>
private static void ParseFields(CompositeNode parentNode, List <WordField> wordFields)
{
// This will recursively search the document object model for any fields at any depth and constructs a hierarchical organization of the fields
// found.
Node childNode = parentNode.FirstChild;
while (childNode != null)
{
// Composite nodes will be searched recursively for child fields.
if (childNode.IsComposite)
{
MergeDocument.ParseFields(childNode as CompositeNode, wordFields);
}
// When a field is identified based on the starting node, a new WordField is generated from the stream of nodes in the document. These fields
// classes are easier to manage than an distinguised stream of nodes.
if (childNode.NodeType == NodeType.FieldStart)
{
WordField wordField = WordField.CreateField(childNode as FieldStart);
wordFields.Add(wordField);
childNode = wordField.FieldEnd;
}
// Test the next node in the stream for the presence of a field.
childNode = childNode.NextSibling;
}
}
/// <summary>
/// Create a hierarchical group of nodes from the start of the field.
/// </summary>
/// <param name="fieldStart">The node which represents the start of a field.</param>
public WordField(FieldStart fieldStart)
{
// Create the object.
this.endDictionary = new Dictionary <Node, WordField>();
this.startDictionary = new Dictionary <Node, WordField>();
this.fieldStart = fieldStart;
// This will find the end of the field and acts as a kind of delimiter. Since the DOM structure makes heavy use of linked lists, it can also act
// to join the end of a field to the start and back again when navigating through the document.
Boolean parsing = true;
Node node = this.fieldStart.NextSibling;
while (parsing)
{
// This is the field that is going to be parsed out of the stream.
WordField wordField = default(WordField);
// Each node is parsed to determine what kind of an object has been discovered here.
switch (node.NodeType)
{
case NodeType.FieldStart:
// This will recurse into any of the embedded fields creating a hierarchical order of fields that can be evaluated from the inside out.
wordField = CreateField(node as FieldStart);
// Any field found in the field field result area is ignored. All other fields are added to a collection that can be indexed sequentially
// using an enumerator or directly using the dictionary.
this.startDictionary.Add(wordField.FieldStart, wordField);
this.endDictionary.Add(wordField.FieldEnd, wordField);
// This moves the parser to the position after the field no matter how many levels of embedded fields were included.
node = wordField.FieldEnd;
break;
case NodeType.FieldSeparator:
// This indicates where the literal portion of a field starts.
this.fieldSeperator = node as FieldSeparator;
break;
case NodeType.FieldEnd:
// The end of the field is part of this obect and used when navigating forward or backward through the nodes of a paragraph.
this.fieldEnd = node as FieldEnd;
// When the final field is found the parsing of this field is completed.
parsing = false;
break;
}
// The parsing continues until the last field node is discovered.
node = node.NextSibling;
}
}
/// <summary>
/// Clears the literal data from a field.
/// </summary>
public void ClearLiteral()
{
// This will remove the literal portion of a field.
Node literalNode = this.fieldSeperator.NextSibling;
while (literalNode != this.fieldEnd)
{
// When parsed, the fields that appear in the literal area are added to the list of child fields that need to be evaluated. However, this can
// lead to errors, especially if the field doesn't exist any more. Removing the field from the list of children inhibits the evaluation of
// anything that appears after the field seperator.
if (literalNode.NodeType == NodeType.FieldStart)
{
WordField wordField = this.FindByStartNode(literalNode);
this.startDictionary.Remove(wordField.FieldStart);
this.endDictionary.Remove(wordField.FieldEnd);
}
// This will remove all the nodes in the literal part of the field.
Node nextNode = literalNode.NextSibling;
literalNode.Remove();
literalNode = nextNode;
}
}
/// <summary>
/// Creates a strongly-typed field from the tokens in the WordProcessing DOM.
/// </summary>
/// <param name="fieldStart">The starting point for the field.</param>
/// <returns>A strongly typed field based on the tokens found in the stream.</returns>
public static WordField CreateField(FieldStart fieldStart)
{
// This is what is created if the type isn't recognized.
WordField wordField = default(WordField);
// The field type indicates what parser is used to evaluate the field.
switch (fieldStart.FieldType)
{
case FieldType.FieldDate:
// Create a new 'DATE' field.
wordField = new DateField(fieldStart);
break;
case FieldType.FieldIf:
// Create a new 'IF' field.
wordField = new IfField(fieldStart);
break;
case FieldType.FieldMergeField:
// Create a new 'MERGEFIELD' field.
wordField = new MergeField(fieldStart);
break;
default:
// All other fields are unhandled but parsed.
wordField = new UnhandledField(fieldStart);
break;
}
// This is a generic field that can be evaluated.
return(wordField);
}
/// <summary>
/// Create the resources required for this instance.
/// </summary>
/// <param name="startNode"></param>
public IfField(FieldStart startNode)
: base(startNode)
{
// Initialize the object.
this.falseExpression = new List <Node>();
this.trueExpression = new List <Node>();
this.expression = String.Empty;
// The parser below will pull apart the various parts of the conditional statement. The biggest issue is that the 'true' clause and the 'false'
// clause contain a combination of Nodes which must be preserved exactly as they will become part of the document. It is not enought to just extract
// the text from these nodes, the fonts, paragraphs and styles and any other fields must be extracted as well. Parsing nodes is difficult as the
// various tokens can exist all in a single node, or be spread across several nodes. These variables are used for parsing. Note that parsinge starts
// from the end of the statement and works towards the beginning. This is because the 'true' and 'false' clauses are a fixed part of the syntax
// and can be deterministically parsed using the quotes as tokens. The rest of the statement -- the conditional expression -- is much more
// difficult and requires a professional strength parser and analyzer. The reason a RexEx function doesn't work here is because the values
// extracted are in the form of CodeDOM Nodes and can't be tokenized.
Int32 lastIndex = default(Int32);
Node node = this.FieldEnd.PreviousSibling;
Run run = default(Run);
Run runClone = default(Run);
Int32 startIndex = -1;
State state = default(State);
// The parsinge is constructed as a simple state machine. There is no pushback or advanced expression evaluation performed in this section. The
// primary goal is to remove the nodes that will become the 'true' and 'false' expressions. A full-blown expression evaluator will be called with
// the part of the statement that remains after pulling out the clauses.
while (state != State.Final)
{
switch (node.NodeType)
{
case NodeType.Run:
// Each of the run nodes can have a quote character which delineates the 'true' and 'false' clauses. The 'startIndex' acts as a kind of
// cursor as the parser moves through each of the Run nodes.
run = node as Run;
if (startIndex == -1)
{
startIndex = run.Text.Length - 1;
}
// Once a Run node is recognized it is examined for tokens that delinate the clauses. The state drives how each Run node is examined and
// determines what is trying to be extracted from each node.
switch (state)
{
case State.FieldResult:
// Just eat any Run nodes in the Field Result section of the field. This moves the parser past the garbage -- the last evaluated data
// for this field -- and into the useful part of the field.
startIndex = -1;
node = node.PreviousSibling;
break;
case State.FalseEndQuote:
// Look for the quote that finishes the 'False' clause.
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex == -1)
{
// If the delimiter character can't be found then try the previous node.
startIndex = -1;
}
else
{
// If the delimiter is found then move the cursor into the body of the clause.
state = State.FalseStatement;
startIndex = lastIndex - 1;
}
// move to the previous node when there is nothing left to parse in the current node until the delimiter is found.
if (startIndex == -1)
{
node = node.PreviousSibling;
}
break;
case State.FalseStatement:
// This will extract the body of the 'False' condition. The next token that will cause a state change is the quote that opened this
// condition (remember we are parsing backwards).
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex == -1)
{
// If there are no delimiters in the current run then clone it and copy the text up to the end quote (or the entire Run node if the
// end quote was in a previously parsed node). There are times when the false clause is empty and an optimization removes any
// empty Run nodes.
String text = run.Text.Substring(0, startIndex + 1);
if (text != String.Empty)
{
runClone = run.Clone(true) as Run;
runClone.Text = text;
this.falseExpression.Add(runClone);
}
// Since this node doesn't contain a delimiter, the state remains the same. Continue parsing the next node for the start quote.
node = node.PreviousSibling;
startIndex = -1;
}
else
{
// When an open quote is found we've come to the start of the false statement. Copy any remaining text in the current Run node
// into the resulting clause. The quotes will be removed but the rest of the text will be a literal copy of the text found in the
// document complete with all the formatting.
String text = run.Text.Substring(lastIndex + 1, startIndex - lastIndex);
if (runClone == null || runClone.Text != String.Empty)
{
runClone = run.Clone(true) as Run;
runClone.Text = text;
this.falseExpression.Add(runClone);
}
// After the opening quote is found the parser will move to a state where it looks for the opening quote. While this may seem
// unnecessary because we've already found the quote, it made the logic simpler than having to create a 'push back' concept for
// the tokens.
state = State.FalseStartQuote;
}
break;
case State.FalseStartQuote:
// This will search for the start quote of the 'False' clause. Note that there is no check here for an empty Run node. The only way
// for the parser to be in this state is if the previous state had found an open quote in the current Run node.
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex != -1)
{
// If the start quote was the starting character of this Run node then the parsing will continue with the previous sibling when
// the next state is entered. Otherwise, the cursor is moved in front of the opening quote as the parsing continues.
if (lastIndex == 0)
{
node = node.PreviousSibling;
startIndex = -1;
}
else
{
startIndex = lastIndex - 1;
}
// Transition to a state where the parser searches for the opening quote of the 'True' condition.
state = State.TrueEndQuote;
}
break;
case State.TrueEndQuote:
// Look for the quote that finishes the 'True' clause.
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex == -1)
{
// If the delimiter character can't be found then try the previous node.
startIndex = -1;
}
else
{
// If the delimiter is found then move the cursor into the body of the clause.
state = State.TrueStatement;
startIndex = lastIndex - 1;
}
// move to the previous node when there is nothing left to parse in the current node until the delimiter is found.
if (startIndex == -1)
{
node = node.PreviousSibling;
}
break;
case State.TrueStatement:
// This will extract the body of the 'True' condition. The next token that will cause a state change is the quote that opened this
// condition (remember we are parsing backwards).
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex == -1)
{
// If there are no delimiters in the current run then clone it and copy the text up to the end quote (or the entire Run node if the
// end quote was in a previously parsed node). There are times when the true clause is empty and an optimization removes any
// empty Run nodes.
String text = run.Text.Substring(0, startIndex + 1);
if (text != String.Empty)
{
runClone = run.Clone(true) as Run;
runClone.Text = text;
this.trueExpression.Add(runClone);
}
// Since this node doesn't contain a delimiter, the state remains the same. Continue parsing the next node for the start quote.
node = node.PreviousSibling;
startIndex = -1;
}
else
{
// When an open quote is found we've come to the start of the true statement. Copy any remaining text in the current Run node into
// the resulting clause. The quotes will be removed but the rest of the text will be a literal copy of the text found in the
// document complete with all the formatting.
String text = run.Text.Substring(lastIndex + 1, startIndex - lastIndex);
if (runClone.Text != String.Empty)
{
runClone = run.Clone(true) as Run;
runClone.Text = text;
this.trueExpression.Add(runClone);
}
// After the opening quote is found the parser will move to a state where it looks for the opening quote. While this may seem
// unnecessary because we've already found the quote, it made the logic simpler than having to create a 'push back' concept for
// the tokens.
state = State.TrueStartQuote;
}
break;
case State.TrueStartQuote:
// This will search for the start quote of the 'True' clause. Note that there is no check here for an empty Run node. The only way
// for the parser to be in this state is if the previous state had found an open quote in the current Run node.
lastIndex = run.Text.LastIndexOf('\"', startIndex);
if (lastIndex != -1)
{
// If the start quote was the starting character of this Run node then the parsing will continue with the previous sibling when
// the next state is entered. Otherwise, the cursor is moved in front of the opening quote as the parsing continues.
if (lastIndex == 0)
{
node = node.PreviousSibling;
startIndex = -1;
}
else
{
startIndex = lastIndex - 1;
}
// Transition to a state where the parser searches for condition expression.
state = State.Expression;
}
break;
case State.Expression:
// Once the 'True' and 'False' clauses have been parsed, everthing between the start quote and the start of the field can be considered
// part of the expression. There is no formatting that is required to evaluate the expression so there is no need to pull apart the
// Run nodes.
this.expression = run.Text.Substring(0, startIndex + 1) + expression;
// This moves the cursor up to the previous run node. Remember the parsing runs backwards because it is more deterministic to pull out
// the 'True' and 'False' clauses from the end than to parse forward to determine where the condition ended.
startIndex = -1;
node = node.PreviousSibling;
break;
}
break;
case NodeType.FieldSeparator:
// The field seperator divides the field into the part you see in the document when working with fields, and the actual text that is
// displayed when you are not working with fields. Anything after the field seperator is previously evaluated data which is of no interest
// to this parser.
node = node.PreviousSibling;
state = State.FalseEndQuote;
break;
case NodeType.FieldEnd:
// Remember that the parser works from the end of the field to the start. Therefore, the end is just the beginning (of a field).
WordField wordField = this.FindByEndNode(node);
// Each field is handled according to the state of the parser.
switch (state)
{
case State.FalseStatement:
// A field found during the parsing of a false statement is added in its entirety to the literal collection.
while (node != null && node.NextSibling != wordField.FieldStart)
{
this.falseExpression.Insert(0, node.Clone(true));
node = node.PreviousSibling;
}
// Evaluate the previous node in the DOM.
node = wordField.FieldStart.PreviousSibling;
break;
case State.TrueStatement:
// A field found during the parsing of a true statement is added in its entirety to the literal collection.
while (node != null && node.NextSibling != wordField.FieldStart)
{
this.trueExpression.Insert(0, node.Clone(true));
node = node.PreviousSibling;
}
// Evaluate the previous node in the DOM.
node = wordField.FieldStart.PreviousSibling;
break;
case State.Expression:
// When a field is found during the evaluation of an expression, the name of the merge field is used as a general purpose parameter to
// the expression evaluator.
if (wordField is MergeField)
{
MergeField mergeField = wordField as MergeField;
this.expression = mergeField.Reference + expression;
}
// Evaluate the previous node in the DOM.
node = wordField.FieldStart.PreviousSibling;
break;
default:
// Ignore anything whiel processing the field results.
node = wordField.FieldStart.PreviousSibling;
break;
}
break;
case NodeType.FieldStart:
// The parser works from the end of the field to the front because the expression can have many items in it but the 'True' and 'False'
// clauses are relatively easy to delinate. When the cursor has come to the start of the field, then the parsing of this 'IF' statement is
// complete. The last task is to remove the 'IF' keyword from the expression.
this.expression = expression.Substring(expression.IndexOf(IfField.ifText) + IfField.ifText.Length);
state = State.Final;
break;
}
}
}
/// <summary>
/// Evaluate the IF field.
/// </summary>
public override void Evaluate()
{
// This will clear out the literal area of the field. The main idea of evaluation is to replace the contents of this part of the field with the
// evaluated data. The original field -- the stuff before the field seperator -- remains pretty much in tact.
this.ClearLiteral();
// Recursively evaluate each of the child fields before the parent is evaluated.
foreach (WordField wordField in this)
{
wordField.Evaluate();
}
// This is a general purpose expression evaluator. It is used to determine the veracity of the expression in the 'IF' statement and depending on
// the resulting value, will either add the 'True' clause or the 'False' clause to the document.
ExpressionContext context = new ExpressionContext();
// The first step to evaluating a conditional statement to provide distinct variables for all the merge fields.
foreach (WordField wordField in this)
{
if (wordField is MergeField)
{
MergeField mergeField = wordField as MergeField;
if (!context.Variables.ContainsKey(mergeField.Reference))
{
context.Variables.Add(mergeField.Reference, mergeField.Value);
}
}
}
try
{
// This is the part where the expression that was parsed out of the field is evaluated. At this point all the variables having been set to
// values from the merged database and the expression can be compiled and evaluated. Also note that this will catch any compile time errors
// and produce a field in the output document that contains the error.
IDynamicExpression iDynamicExpression = context.CompileDynamic(this.expression);
List <Node> statement = Convert.ToBoolean(iDynamicExpression.Evaluate()) ? this.trueExpression : this.falseExpression;
// The literal part of the field has been cleared out and now that the expression has been evaluated it can be filled in with either the data
// collected from the 'True' or' 'False' statements (depending on the outcome of the evaluation done above).
foreach (Node replacementNode in statement)
{
this.FieldStart.ParentNode.InsertBefore(replacementNode, this.FieldEnd);
}
// Adding elements from the 'True' or 'False' statements has the potential to add child fields to this field. This pass will look generate
// any new fields and immediately evaluate them. Like a real programming language, conditional clauses are not evaluated until the condition
// has been resolved.
foreach (Node replacementNode in statement)
{
if (replacementNode.NodeType == NodeType.FieldStart)
{
WordField wordField = WordField.CreateField(replacementNode as FieldStart);
wordField.Evaluate();
}
}
}
catch (Exception exception)
{
// This inserts the error message into the document.
Run run = new Run(this.FieldStart.Document, String.Format("Error! {0}", exception.Message));
this.FieldStart.ParentNode.InsertAfter(run, this.FieldEnd);
}
}
