I'm playing with the statechart library, in conjunction with Boost 1.33 and VC 7.1. I'm having trouble using the state history functionality; I can't get it to compile, even taking into account the useful static assertions that ought to be helping me. Can someone suggest what I might be missing? Here's a boiled down sample: there are 2 top-level states, StateA and StateB. Event2 transitions between them. When in StateA, Event1 transitions between 2 sub-states, StateA1 and StateA2. When transitioning from StateB to StateA, I want to use StateA's history to determine which of StateA1 and Stae2 become active. From reading the tutorial, references, and sample code, it looks like this should work: //////////////////////////////////////////////////////////////////// #include <boost/statechart/event.hpp> #include <boost/statechart/state_machine.hpp> #include <boost/statechart/simple_state.hpp> #include <boost/statechart/transition.hpp> #include <boost/statechart/shallow_history.hpp> namespace sc = boost::statechart; //Events struct Event1 : sc::event<Event1> {}; struct Event2 : sc::event<Event2> {}; //States struct StateA; struct StateA1; struct StateA2; struct StateB; struct SCTest : sc::state_machine<SCTest, StateA> {}; struct StateA : sc::simple_state<StateA, SCTest, StateA1, sc::has_shallow_history> { typedef sc::transition<Event2, StateB > reactions; }; struct StateB : sc::simple_state<StateB, SCTest> { typedef sc::transition<Event2, sc::shallow_history<StateA> > reactions; }; struct StateA1 : sc::simple_state<StateA1, StateA> { typedef sc::transition<Event1, StateA2> reactions; }; struct StateA2 : sc::simple_state<StateA2, StateA> { typedef sc::transition<Event1, StateA1> reactions; }; int main(int argc, char* argv[]) { SCTest test; return 0; } This gives me the following error: c:\boost\boost\statechart\shallow_history.hpp(34) : error C2027: use of undefined type 'boost::STATIC_ASSERTION_FAILURE<x>' with [ x=false ] <snip> [ DefaultState=StateA ] etc., and comments at that point in the statechart source direct me to: "...pass either statechart::has_deep_history or statechart::has_full_history as the last parameter of DefaultState's context." This doesn't seem to help, though. If I use has_deep_history or has_full_history in StateA's definition, the errors do not go away. Any help would be greatly appreciated! -------------------------------- John Wismar john.wismar@autozone.com

I'm playing with the statechart library, in conjunction with Boost 1.33 and VC 7.1.  I'm having trouble using the state history functionality; I can't get it to compile, even taking into account the useful static assertions that ought to be helping me.  Can someone suggest what I might be missing?Here's a boiled down sample:

Hi John <john.wismar <at> autozone.com> writes: there

are 2 top-level states, StateA and StateB.  Event2 transitions between them.  When in StateA, Event1 transitions between 2 sub-states, StateA1 and StateA2.  When transitioning from StateB to StateA, I want to use StateA's history to determine which of StateA1 and Stae2 become active.

I'm very sorry for the late answer, but I missed your post back in September because it was sent in reply to an unrelated post (Boost Graph Library). Mick's analysis of the problem is correct. When you make a transition to history then you must specify the *default* *state* that is entered in case the history has never been saved before. The *outer* *state* of the default state must then pass the appropriate has_xxx_history. Note that this is very much in line with the UML notation. A transition to history always goes to a H disc first and from that H disc you have another arrow to the default state. The state containing the H disc is said to be the one "having history".

From reading the tutorial, references, and sample code, it looks like this should work: [example snipped]

I checked the docs and I can't find any examples that display the wrong syntax you used. A pointer would be appreciated. HTH & Regards, -- Andreas Huber When replying by private email, please remove the words spam and trap from the address shown in the header.

Mick Hollins wrote:

...

Note that this is very much in line with the UML notation. A transition to history always goes to a H disc first and from that H disc you have another arrow to the default state.

Initially I thought it was not quite in line with UML since the boost::statechart implementation allows a different default arrow to be specified for each transition to a H disc, whereas UML only allows a single arrow out of a H disc.

But I just realised that afaict UML does not disallow having more than one H disc in a parent state. As such you can have 2 H discs each with a different default state, and boost::statechart supports that.

I faintly remember seeing a statechart diagram with more than one history disc in one state. This was enough for me to choose a design that allows for that. I can't say offhand whether or not UML explicitly allows it.

...

I checked the docs and I can't find any examples that display the wrong syntax you used. A pointer would be appreciated.

I don't believe your docs are wrong. It is just that it is easy to misread them as it seems natural to expect that you would say:

sc::shallow_history<StateA>

since you are requesting a transition to one of the sub-states of StateA. Saying:

sc::shallow_history<StateA1>

just reads a little funny since you may or may not be transitioning to StateA1. The following, while redundant, might actually be easier for people to read:

sc::shallow_history<StateA, StateA1>

where the first parameter specifies the parent state to which you are performing a history transition, and the second parameter specifies the default state.

I'm not sure. In the UML diagram the arrow doesn't point to StateA either so you have a very similar problem there. Plus, I don't like the unnecessary redundancy. Regards, -- Andreas Huber When replying by private email, please remove the words spam and trap from the address shown in the header.