The Parser and the Application
This section shows you how to integrate the parser in your applications. It discusses the various interfaces available to the programmer.
The Architecture of an XML Program
Figure 7.1 illustrates the architecture of XML programs. As you can see, it is divided into two parts:
The parser deals with the XML file.
The application consumes the content of the file through the parser.
Note that the application can be very simple (such as printing information on the screen) or quite complex (such as a browser, an editor, or an XSL processor).
This chapter and the next one concentrate on the dotted line between the two elements. This is the API (application programming interface) or the communication path between the parser and the application.
The parser and the application must share a common model for XML data. In practice, the common model is always some variation on a tree in memory that matches the tree in the XML document.
The parser reads the XML document and populates the tree in memory. This tree built by the parser is an exact match of the tree in the XML document. The application manipulates it as if it were the XML document. In fact, for the application, it is the XML document.
There are two basic ways to interface a parser with an application: using object-based APIs and using event-based APIs. In practice, the two approaches are more complementary than competitive.
Using an object-based interface, the parser explicitly builds a tree of objects that contains all the elements in the XML document.
This is probably the most natural interface for the application because it is handed a tree in memory that exactly matches the file on disk.
Obviously, it's more convenient for the application to work with the tree in memory, if only because it doesn't have to worry about the XML syntax. Furthermore, if using a validating parser, the tree may have been validated against a DTD or a schema.
Listing 7.1 is a list of products, with their prices in U.S. dollars, presented in an XML document. The structure for this document is shown in Figure 7.2.
Listing 7.1: products.xml
<?xml version="1.0"?> <xbe:products xmlns:xbe="http://www.psol.com/xbe2/listing7.1"> <xbe:product price="499.00">XML Editor</xbe:product> <xbe:product price="199.00">DTD Editor</xbe:product> <xbe:product price="29.99">XML Book</xbe:product> <xbe:product price="699.00">XML Training</xbe:product> </xbe:products>
The parser reads this document and gradually builds a tree of objects that matches the document. Figure 7.3 illustrates how the tree is being built.
When the XML parser reads the document in Listing 7.1, it recognizes that the top-level element is named products. Therefore, it constructs an object to represent the products element.Figure 7.2: The structure of the price list.
Figure 7.3: Building the tree of objects.
The next element is a product. The parser creates another object to represent the product element. Because this is a tree, it attaches the product object to the products object.
The next step is to recognize the price attribute. Again, the parser creates an object for the price and adds it to the tree being built.
In the product, there is some text that the parser translates in another object, a text node, in the tree.
The parser then moves to another product element, which also contains a price attribute and a text node. This results in more objects in the tree.
The process continues until the document has been completely read. By the time the parser reaches the end of the document, it has built a tree of objects in memory that matches the tree of the document
The second approach to interfacing the parser and the application is through events. An event-based interface is natural for the parser but it is more complex for the application. Yet, with some practice, event-based interfaces prove very powerful. More programmers (and more parsers) are turning to event-based APIs for this reason.
With an event-based interface, the parser does not explicitly build a tree of objects. Instead, it reads the file and generates events as it finds elements, attributes, or text in the file. There are events for element starts, element ends, attributes, text content, entities, and so on. Figure 7.4 illustrates how it works.
At first sight, this solution is less natural for the application because it is not given an explicit tree that matches the file. Instead, the application has to listen to events and determine which tree is being described.Figure 7.4: An event-based API.
In practice, both forms of interfaces are helpful, but they serve different goals. Object-based interfaces are ideal for applications that manipulate XML documents such as browsers, editors, XSL processors, and so on.
Event-based interfaces are geared toward applications that maintain their own data structure. For example, event-based interfaces are well adapted to applications that import XML documents in databases. These applications have their own data structure and they map directly from the XML structure to their structure.
An event-based interface is therefore more efficient because it does not explicitly build the XML tree in memory. Fewer objects are required and less memory is being used.
» Chapter 8 discusses event-based interfaces in greater detail ("Alternative API: SAX," page 253).
The Need for Standards
Ideally, the interface between the parser and the application should be standardized. A standard interface allows you to write software using one parser and to deploy the software with another parser.
Again, there is a similarity with databases. Relational databases use SQL as their standard interface. Because they all share the same interface, developers can write software with one database and later move to another database (for price reasons, availability, and so on) without changing the application.
That's the theory, at least. In practice, small differences, vendor extensions, and other issues mean that moving from one product to another requires more work than just recompiling the application. At the minimum, even if they follow the same standards, vendors tend to differ on bugs.
But even if different vendors are not 100% compatible with one another, standards are a good thing.
For one thing, it is still easier to adapt an application from a vendor-tainted version of the standard to another vendor-tainted version of the same standard than to port the application between vendors that use completely different APIs.
Furthermore, standards make it easier to learn new tools. It is easier to learn a new interface when 90% of it is similar to the interface of another product.
Obviously, the two different approaches for interfaces translate into two different standards. The standard for object-based interfaces is DOM, Document Object Model, published by the W3C (http://www.w3.org/TR/DOM-Level-2-Core).
The standard for event-based interface is SAX, Simple API, developed collaboratively by members of the XML-DEV mailing list and edited by David Megginson (http://www.megginson.com)
The two standards are not really in opposition because they serve different needs. Sun has integrated DOM and SAX in the Java API.
This chapter concentrates on DOM. The next chapter discusses SAX. Chapter 9, "Writing XML," looks at how to create XML documents.