Another example

Consider the following action rule:

   p(X,Y), {event(X,O),event(Y,O)} => write(O).

An agent, which is attached to both X and Y, echoes the event object when the agent is activated by an event. The following gives several sample queries and their expected outputs:

#	Query	Output
1	p(X,Y),post_event(X,a)	a
2	p(X,Y),post_event(Y,b)	b
3	p(X,Y),post_event(X,a),post_event(Y,b)	ab
4	p(X,Y),post_event(X\/Y,c)	c
5	p(X,Y),post_event(X/\Y,c)	c
6	p(X,Y),p(U,V),post_event(X\/U,c)	cc
7	p(X,Y),p(U,V),post_event(X/\U,c)
8	p(X,Y),p(U,V),X=U,post_event(X/\U,c)	cc

Query number 7 does not generate output, since no agent is attached to either of the channels X and U. When two channels are unified, the younger variable is set to reference the older one,^12.2 and all of the agents that are attached to the younger variable are copied to the older one. So, in query number 8, after X=U, X and U become one variable, and the two agents p(X,Y) and p(U,V) become attached to the variable. Therefore, after post_event(X/\U,c), both of the agents are activated. In the examples, the queries will give the same outputs if post_event_df is used instead of post_event. This is generally not the case, if an action rule also posts events.