Base class for parsers that use a BNF tree to parse an input stream.

The BNF tree form used is a variation on Extended BNF described in the XML1.0 specification. The general form of the tree is as follows:

Each node in the tree defines a rule that the input stream must meet to satisfy the grammar.

1. a node type is set to the rule type, as defined in TParserNodeTypes

2. node data stores any string required by the rule: e.g. for a comparison rule, the string to match against

3. the parser allows callback functions to be called either before or after the rule is processed. If these are present, they are stored as attributes of the node.

4. some rules allow sub-rules: for example, the AND rule expects a number of sub-rules, all of which must be successful if the AND rule itself is to succeed. Each sub-rule is represented as a child node of the parent rule. Sub-rules in turn can have sub-rules.

5. reference rule nodes are also allowed: these do not define themselves rules, but direct the parser to another rule. They can link rules to each other and so build rule sequences more complex than a simple tree.

All the top-level rules are stored as attributes of the root node. The attribute type is a string that names the rule; the attribute value is a pointer to the node that implements the rule.

The class supplies functions that encapsulate adding rules appropriately to the tree. The parser provider creates a derived class that implements the virtual method TreeL() that uses these functions to create a BNF rule tree.

The user of the parser initialises the parser with ResetL(), and then passes input data to the parser using ProcessData(). The parser supports partial parsing: the input stream does not have to completed before parsing can begin. As soon as data is added, the parser attempts to parse it.