Chapter 24

XML as a Distributed Object Technology

Since XML is simple and easy to work with, it has become popular as a mechanism for serialization. XML serialized objects can be sent across a network, and the sender can be confident that the recipient will be able to parse them, regardless of their platform. For example, consider the simple class shown here:

class Simple

{

public:

std::string mName;

int mPriority;

std::string mData;

};

An object of type Simple could be serialized to the following XML:

<Simple name=”some name” priority=”7”>this is the data</Simple>

Of course, since XML doesn’t specify how individual nodes should be used, you could just as easily serialize it as follows:

<Simple name=”some name” priority=”7” data=”this is the data” />

As long as the recipient of the serialized XML is aware of the rules you are using to serialize the object, they should be able to deserialize it.

XML serialization has increased in popularity as a simpler alternative to heavyweight distributed object technologies such as CORBA. XML has a much more gradual learning curve than CORBA and offers many of the same benefits, such as platform and language independence.

Generating and Parsing XML in C++

Because XML is merely a file format, and not an object description language, the task of converting data to and from XML is left to the programmer. In general, writing XML is the easy part. Reading XML is usually aided by a third-party XML library.

Generating XML

To use XML as a serialization technology, your objects will need to be able to convert themselves into XML. In many cases, building a stream of XML on the fly is the easiest way to output XML. In fact, the notion that XML elements are “wrapped” in other elements makes things even easier. You can build new XML documents as amalgams of existing ones. If that sounds a bit complicated, consider the following example. Assume that you have a function called getNextSentenceXML(), which asks the user for a sentence and returns it as an XML representation of the sentence. Because that function returns the sentence as a valid XML element, you could create a dialogue of sentences by wrapping the results of multiple calls to getNextSentenceXML() in a dialogue element tag:

712

Exploring Distributed Objects

string getDialogueXML()

{

sstringstream outStream;

// Begin the dialogue element. outStream << “<dialogue>”;

while (true) {

// Get the next sentence.

string sentenceXML = getNextSentenceXML(); if (sentenceXML == “”) break;

// Add the sentence element. outStream << sentenceXML;

}

// End the dialogue element. outStream << “</dialogue>”;

return outStream.toString();

}

If subsequent calls to getNextSentenceXML() returned the sentences from the preceding example, the result of this function would be:

<dialogue><sentence speaker=”Marni”>Let’s go get some ice cream.</sentence><sentence speaker=”Scott”>After I’m done writing this C++ book.</sentence></dialogue>

The output is a bit strange because it wasn’t formatted with line breaks and tabs. It is, however, valid XML. If you wanted to beautify the output a bit, you have a few options:

You could use a third-party tool after the fact. For example, the open-source command-line program tidy (http://tidy.sourceforge.net) has an XML pretty-print feature among its many useful tools.

You could include carriage returns and spaces manually in your code. This quickly gets complicated because inside of getNextSentenceXML(), the code has no idea how many tabs to use.

You could use (or write) a simple XML generation class library that is aware of nested elements and formats them appropriately.

An XML Output Class

Even though outputting XML is straightforward, there are several good reasons to factor XML output code into a separate class or set of classes. In addition to the formatting issue seen previously, separating out the code for XML generation provides the following benefits:

Cleaner code. Who wants < all over the place?!

A central location to implement escaping of special characters.

713

Chapter 24

A more object-oriented approach. XML elements could be objects, which can then be stored, passed to methods, and organized.

Reduction of the possibility of XML syntax errors by centralizing output.

Writing an XML generation class is also temptingly simple. The class definition of a simple XML Element class is shown here:

// XMLElement.h

#include <string> #include <vector> #include <map> #include <iostream>

class XMLElement

{

public:

XMLElement();

void setElementName(const std::string& inName);

void setAttribute(const std::string& inAttributeName, const std::string& inAttributeValue);

void addSubElement(const XMLElement* inElement);

// Setting a text node will override any nested elements. void setTextNode(const std::string& inValue);

friend std::ostream& operator<<(std::ostream& outStream, const XMLElement& inElem);

protected:

void writeToStream(std::ostream& outStream, int inIndentLevel = 0) const;

void indentStream(std::ostream& outStream, int inIndentLevel) const;

private:
std::string	mElementName;
std::map<std::string, std::string>	mAttributes;
std::vector<const XMLElement*>	mSubElements;
std::string	mTextNode;

};

Using this class, a user could easily create XMLElement objects, set their attributes, and set text nodes or subelements. At any time, the client can call operator<< to get the XML representation of the current state of the element.

A sample implementation is shown next. Because it uses C++ syntax, which you’re a pro at by now, we won’t explain every single line. Take a look at the inline comments if it doesn’t make sense at first glance.

714

Exploring Distributed Objects

#include “XMLElement.h”

using namespace std;

XMLElement::XMLElement() : mElementName(“unnamed”)

{

}

void XMLElement::setElementName(const string& inName)

{

mElementName = inName;

}

void XMLElement::setAttribute(const string& inAttributeName, const string& inAttributeValue)

{

// Set the key/value pair, replacing the existing one if it exists. mAttributes[inAttributeName] = inAttributeValue;

}

void XMLElement::addSubElement(const XMLElement* inElement)

{

// Add the new element to the vector of subelements. mSubElements.push_back(inElement);

}

void XMLElement::setTextNode(const string& inValue)

{

mTextNode = inValue;

}

ostream& operator<<(ostream& outStream, const XMLElement& inElem)

{

inElem.writeToStream(outStream); return (outStream);

}

void XMLElement::writeToStream(ostream& outStream, int inIndentLevel) const

{

indentStream(outStream, inIndentLevel);

outStream << “<” << mElementName; // open the start tag

// Output any attributes.

for (map<string, string>::const_iterator it = mAttributes.begin(); it != mAttributes.end(); ++it) {

outStream << “ “ << it->first << “=\”” << it->second << “\””;

}

// Close the start tag. outStream << “>”;

if (mTextNode != “”) {

// If there’s a text node, output it.

715

Chapter 24

outStream << mTextNode; } else {

outStream << endl;

// Call writeToStream at inIndentLevel+1 for any subelements.

for (vector<const XMLElement*>::const_iterator it = mSubElements.begin(); it != mSubElements.end(); ++it) {

(*it)->writeToStream(outStream, inIndentLevel + 1);

}

indentStream(outStream, inIndentLevel);

}

// Write the close tag.

outStream << “</” << mElementName << “>” << endl;

}

void XMLElement::indentStream(ostream& outStream, int inIndentLevel) const

{

for (int i = 0; i < inIndentLevel; i++) { outStream << “\t”;

}

}

The preceding implementation is a great starting point and is perfect for simple XML applications. One of the features that is missing is the escaping of special characters. For example, the character & needs to be escaped as & inside of an XML document. Here is a sample program that shows the use of the XMLElement class to build the document that was output manually in the previous example:

int main(int argc, char** argv)

{

XMLElement dialogueElement; dialogueElement.setElementName(“dialogue”);

XMLElement sentenceElement1; sentenceElement1.setElementName(“sentence”); sentenceElement1.setAttribute(“speaker”, “Marni”); sentenceElement1.setTextNode(“Let’s go get some ice cream.”);

XMLElement sentenceElement2; sentenceElement2.setElementName(“sentence”);

sentenceElement2.setAttribute(“speaker”, “Scott”); sentenceElement2.setTextNode(“After I’m done writing this C++ book.”);

//Add the sentence elements as subelements of the dialogue element. dialogueElement.addSubElement(&sentenceElement1); dialogueElement.addSubElement(&sentenceElement2);

//Output the dialogue element to stdout.

cout << dialogeElement;

return 0;

}

716

Exploring Distributed Objects

The output of this program is:

<dialogue>

<sentence speaker=”Marni”>Let’s go get some ice cream.</sentence> <sentence speaker=”Scott”>After I’m done writing this C++ book.</sentence>

</dialogue>

Many XML Parsing libraries also include XML output facilities. If you are using an XML parser for input (described next), check into its output capabilities before writing your own.

Parsing XML

To deserialize XML objects, you’ll need to interpret, or parse, the document. Unless the XML you are reading is extremely simple and rigidly defined, you will most likely want to use a third-party XML parsing library. XML parsing libraries typically come in two flavors, SAX and DOM.

A SAX (Simple API for XML) parser uses an event-based parsing model. To use a SAX parser, you register callback functions or an object that implements certain methods. As the document is parsed, the appropriate functions or methods are called, giving you a chance to perform an action. For example, if you wanted to look for duplicate XML element names in a document, you could register a callback that is triggered upon reaching an element start tag. Internally, you would keep a list of elements that had already been encountered. Using that list, you could detect the duplicates.

A DOM (Document Object Model) parser converts an XML document into a treelike structure that you can easily walk through in code. To programmers accustomed to object-oriented hierarchies and tree data structures, the DOM approach may seem more natural. The disadvantage of the DOM approach is performance. Because it parses the entire document and builds a structure, it is generally slower and more memory-intensive than SAX. Though the rest of this section deals only with DOM parsers, you will find that most XML parsers support both SAX and DOM.

The Xerces XML Library

One of the most popular XML parsers is Xerces, which is part of the Apache XML project. Xerces is an open-source parser and is available for several languages, including C++. You can download the Xerces- C++ library from http://xml.apache.org/.

Once you have Xerces installed and added to your C++ project, you can offload the work of parsing XML. Xerces is easy to get started with even though it has a wealth of functionality — a sign of a welldesigned library!

The most important class in the Xerces DOM parser is DOMNode. A DOMNode is a single unit of XML data, possibly including other nodes. The subclasses of DOMNode include DOMDocument, DOMElement, DOMAttr, DOMText, an so on. Working with an Xerces DOM generally involves starting with the root node (a DOMDocument) and walking through the tree of nodes to find the desired data. Figure 24-4 shows a slightly simplified version of the node tree for the <dialogue> XML document. It is simplified in that it only shows nodes that actually contain data.

717

Chapter 24

DocumentRoot (DOMDocument)

<dialogue>

(DOMElement)

<sentence>

(DOMElement)

Let's go get...

(DOMText)

Figure 24-4

<sentence>

(DOMElement)

After I'm done...

(DOMText)

The XML attributes are not shown in Figure 24-4 because they are properties of an element, not children of the element.

Using Xerces

The one tricky aspect that you will face first is the way that Xerces represents strings. Because XML can be encoded in various ways, the library has its own character type: XMLch. It also has a utility class called XMLString that makes it easy to work with XMLch strings and convert them to more familiar chars. For example, if a Xerces method returns data as an XMLch* string, you can output it by using

XMLString::transcode() to get a C-style string:

void outputXercesString(XMLch* inXercesString)

{

char* familiarString = XMLString::transcode(inXercesString); cout << familiarString << endl;

}

Because transcode() allocates memory for the C-style string, you must also release it with XMLString::release(), which (in a somewhat bizarre design choice) takes a pointer to the C-style string. The modified version below avoids a memory leak:

void outputXercesString(XMLch* inXercesString)

{

char* familiarString = XMLString::transcode(inXercesString); cout << familiarString << endl; XMLString::release(&familiarString);

}

718

Exploring Distributed Objects

With that bit of oddness out of the way, it’s time to parse some XML. This example parses a file named test.xml into a DOM tree, and then loops through all of the nodes, printing out the names of all elements that are encountered, any attributes contained within those elements, and the contents of any text nodes.

The program begins by including the necessary standard headers and Xerces headers. It also declares XERCES_CPP_NAMESPACE_USE, which is a #define in Xerces that gives the correct namespace to the file.

#include <xercesc/util/PlatformUtils.hpp>

#include <xercesc/dom/DOM.hpp>

#include <xercesc/parsers/XercesDOMParser.hpp> #include <xercesc/util/XMLString.hpp>

#include <iostream>

XERCES_CPP_NAMESPACE_USE using namespace std;

void printNode(const DOMNode* inNode);

The program’s main() is fairly straightforward, even though this is where the actual parsing is taking place. It begins by initializing the Xerces library. Next, it creates a new DOM parser and tells it to parse the file. The result of this operation is a DOMNode that represents the document as a whole. To obtain the root element, getDocumentElement() is called. This value is passed to printNode(), which walks through the tree, printing out the data. Finally, the program cleans up the XML library before exiting.

int main(int argc, char** argv)

{

XMLPlatformUtils::Initialize();

XercesDOMParser* parser = new XercesDOMParser(); parser->parse(“test.xml”);

DOMNode* node = parser->getDocument();

DOMDocument* document = dynamic_cast<DOMDocument*>(node);

if (document != NULL) { printNode(document->getDocumentElement());

}

delete parser; XMLPlatformUtils::Terminate();

return 0;

}

The printNode() function is where things get interesting. Because the parameter inNode can be any type of XML node, the function tries its two known node types in sequence. It first attempts to dynamically cast the node into a text node, catching the cast error in case the node is a different type:

719

Chapter 24

void printNode(const DOMNode* inNode)

{

try {

const DOMText& textNode = dynamic_cast<const DOMText&>(*inNode); char* text = XMLString::transcode(textNode.getData());

cout << “Found text data: “ << text << endl; XMLString::release(&text);

}catch (bad_cast) {

//Not a text node . . .

Next, it tries to cast to an element node. If this cast is successful, the element’s name and any attributes are printed out.

try {

const DOMElement& elementNode = dynamic_cast<const DOMElement&>(*inNode); char* tagName = XMLString::transcode(elementNode.getTagName());

cout << “Found tag named: “ << tagName << endl; XMLString::release(&tagName);

// Look at the attribute list.

DOMNamedNodeMap* attributes = elementNode.getAttributes(); for (int i = 0; i < attributes->getLength(); i++) {

try {

const DOMAttr& attrNode =

dynamic_cast<const DOMAttr&>(*attributes->item(i));

char* name = XMLString::transcode(attrNode.getName()); char* value = XMLString::transcode(attrNode.getValue());

cout << “Found attribute pair: (“ << name << “=” << value << “)” << endl;

XMLString::release(&name);

XMLString::release(&value); } catch (bad_cast) {

cerr << “Error converting attribute!” << endl;

}

}

}catch (bad_cast) {

//Not an element node . . .

Finally, the function calls printNode() recursively on children nodes. In practice, children nodes will only exist on element nodes.

// Print any subelements.

DOMNodeList* children = inNode->getChildNodes(); for (int i = 0; i < children->getLength(); i++) {

printNode(children->item(i));

}

}

720