This represents the rule specification but most engines will compile this specification into an abstract machine or processing graph.

The rule specification comprises a list of antecendents (body) and a list of consequents (head). If there is more than one consequent then a backchainer should regard this as a shorthand for several rules, all with the same body but with a singleton head.

Each element in the head or body can be a TriplePattern, a Functor or a Rule. A TriplePattern is just a triple of Nodes but the Nodes can represent variables, wildcards and embedded functors - as well as constant uri or literal graph nodes. A functor comprises a functor name and a list of arguments. The arguments are Nodes of any type except functor nodes (there is no functor nesting). The functor name can be mapped into a registered java class that implements its semantics. Functors play three roles - in heads they represent actions (procedural attachement); in bodies they represent builtin predicates; in TriplePatterns they represent embedded structured literals that are used to cache matched subgraphs such as restriction specifications.

We include a trivial, recursive descent parser but this is just there to allow rules to be embedded in code. External rule syntax based on N3 and RDF could be developed. The embedded syntax supports rules such as: [ (?C rdf:type *), guard(?C, ?P) -> (?c rb:restriction some(?P, ?D)) ].
[ (?s owl:foo ?p) -> [ (?s owl:bar ?a) -> (?s ?p ?a) ] ].
[name: (?s owl:foo ?p) -> (?s ?p ?a)].
only built in namespaces are recognized as such, * is a wildcard node, ?c is a variable, name(node ... node) is a functor, (node node node) is a triple pattern, [..] is an embedded rule, commas are ignore and can be freely used as separators. Functor names may not end in ':'.

那位大侠具体讲解一下，举几个例子。小弟感激不尽！！！