Is this a good implementation of a loop in Prolog?

Question

First of all, let me tell you that this happens to be the first time I ask something here, so if it's not the right place to do so, please forgive me.

I'm developing a rather complex software that has a Prolog core implementing a FSM. Since I don't want it to stop (ever), I'm trying to write a good loop-like predicate that would work using Prolog's recursion. After a few unsuccessful tries (mainly because of stack problems) I ended up having something similar to this:

/* Finite State Transition Network */
transition(st0,evnt0,st1).
transition(st1,evnt1,st2).
transition(st2,evnt2,st0).

fsm_state(state(st0),system(Energy,ActivePayloads),[]) :-
    /* ... */
    transition(st0,evnt0,NextState),
    !,
    fsm_state(state(NextState),system(Energy,ActivePayloads),[]).

fsm_state(state(st1),system(Energy,ActivePayloads),[]) :-
    /* ... */
    transition(st1,evnt1,NextState),
    !,
    fsm_state(state(NextState),system(Energy,ActivePayloads),[0,1,2]).

fsm_state(state(st2),system(Energy,ActivePayloads),[P|Params]) :-
    /* ... */
    transition(st2,evnt2,NextState),
    !,
    fsm_state(state(NextState),system(Energy,ActivePayloads),[]).

start :- 
    Sys = system(10,[]),
    fsm_state(state(s0),Sys,[]).

Is this a good approach?

Answer 1

First a general remark: The way you go about this is a perfectly natural Prolog approach, declaratively (well, more or less, if it were not for the many !/0 that prevent you to use the predicate in all directions) describing a relation between successive states. This relation is naturally implemented with a recursive predicate.

What strikes me as odd is that you need so many clauses that all seem in essence the same. Why can you not use a single clause like:

fsm(State0, System0, Params0) :-
    /* ... */
    transition(State0, Event, State),
    /* ... */
    fsm(State, System, Params).

where maybe some additional goals relate Params0 with Params, System0 with System and maybe some other entities. The general way is usually to have a relation between states before and after a single "step" of the computation, and another relation (consisting of a single clause like above) that describes a sequence of successive states.