In many ways, XmlReader is analogous to the Simple API for XML (SAX). They both work by reporting events to the client. There is one major difference between XmlReader and a SAX parser, however. While SAX implements a push parser model, XmlReader is a pull parser.

In a push parser, events are pushed to you. Typically, a push parser requires you to register a callback method to handle each event. As the parser reads data, the callback method is dispatched as each appropriate event occurs. Control remains with the parser until the end of the document is reached. Since you don’t have control of the parser, you have to maintain knowledge of the parser’s state so your callback knows the context from which it has been called. For example, in order to decide on a particular action, you may need to know how deep you are in an XML tree, or be able to locate the parent of the current element. Figure 2-1 shows the flow of events in a push parser model application.

Figure 2-1. Push parser model

In a pull parser, your code explicitly pulls events from the parser. Running in an event loop, your code requests the next event from the parser. Because you control the parser, you can write a program with well-defined methods for handling specific events, and even completely skip over events you are not interested in. Figure 2-2 shows the flow of events in a pull parser model application.

Figure 2-2. Pull parser model

A pull parser also enables you to write your client code as a recursive descent parser. This is a top-down approach in which the parser (XmlReader, in this case) is called by one or more methods, depending on the context. The recursive descent model is also known as mutual recursion. A neat feature of recursive descent parsers is that the structure of the parser code usually mirrors that of the data stream being parsed. As you’ll see later in this chapter, the structure of a program using XmlReader can be very similar to the structure of the XML document it reads.