bad naming for eventless transitions

[jp-embedded]

'sc::state::unconditional' is bad naming. The transition is not unconditional, but eventless. Consider to change this to for example 'sc::state::eventless' or 'sc::state::none'

[ringlej]

Agreed. It seems silly to have:

template<> bool sc::transition_actions<&sc::state::unconditional, sc::state_A>::condition(sc::data_model& m);

Seems oxymoronic :)
What kind of usecase is this meant for?

[jp-embedded]

Well, your example condition's transition is both target less and event less. I don't have a use case in mind for this. However, in the timer switch example, there is a condition for a transition which does have a target but no event:

template<> bool sc::transition_actions<&sc::state::unconditional, sc::state_on, sc::state_off>::condition(sc::data_model &m)
{
	return m.user->timer >= 5;
}

This will transition from state_on to state_off when m.user->timer >= 5

[ringlej]

Oh I see. So, the use case is when you want a guard condition to be evaluated upon any event that may occur in the source state. I wouldn’t say that this is really eventless. An event has to occur, but you just don’t care which event it is. I’d offer the suggestion to call this sc::state::any_event and to use this also as the default argument to dispatch

void dispatch(event e = &state::any_event)
{
	bool cont = dispatch_event(e);
	while (cont) {
		if ((cont = dispatch_event(&state::initial)));
		else if ((cont = dispatch_event(&state::any_event)));
		else if (model.event_queue.size()) cont = dispatch_event(model.event_queue.front()), model.event_queue.pop(), cont |= !model.event_queue.empty();
		else break;
	}
}

If you like this, I'd be happy to prepare a PR

[sstiller]

I think, sc::state::eventless is a good name because "eventless" is also used in the scxml standard.

But it's a little bit worse:
If you have multiple eventless transitions with same source and target state, the generated code can not be compiled.
An example:

<?xml version="1.0" encoding="UTF-8"?>
<scxml xmlns="http://www.w3.org/2005/07/scxml" version="1.0" binding="early" name="cond" datamodel="cplusplus">
    <state id="State_1">
        <transition type="external" target="State_2" cond="isOn"/>
        <transition type="external" target="State_2" cond="isOn2"/>
    </state>
    <state id="State_2"/>
    <datamodel>
        <data id="bool isOn" expr="false"/>
        <data id="bool isOn2" expr="false"/>
    </datamodel>
</scxml>

cond.h:176:17: error: redefinition of ‘bool sc_cond::transition_actions<E, S, D0>::condition(sc_cond::data_model&) [with sc_cond::state* (sc_cond::state::* E)(sc_cond&) = &sc_cond::state::unconditional; S = sc_cond::state_State_1; D0 = sc_cond::state_State_2]’
 template<> bool sc_cond::transition_actions<&sc_cond::state::unconditional, sc_cond::state_State_1, sc_cond::state_State_2>::condition(sc_cond::data_model &m)
                 ^~~~~~~
cond.h:169:17: note: ‘bool sc_cond::transition_actions<E, S, D0>::condition(sc_cond::data_model&) [with sc_cond::state* (sc_cond::state::* E)(sc_cond&) = &sc_cond::state::unconditional; S = sc_cond::state_State_1; D0 = sc_cond::state_State_2]’ previously declared here
 template<> bool sc_cond::transition_actions<&sc_cond::state::unconditional, sc_cond::state_State_1, sc_cond::state_State_2>::condition(sc_cond::data_model &m)

I think, if the condition is defined in the scxml, an eventless transition should get some kind of identifier (simple counter).

[jp-embedded]

Thanks for your observation :-)
I will look into this. Your suggested solution is also what I think could be a solution.
Transitions are made unique using S, D and E (source state, destination state and event), but as you show, this is not enough. The same problem will exists for transitions which have an event if S, D & E are equal.