The subject of the day seems to be sub-domains and it's great because
I wanted to check that a use I just made of them was correct.
So the problem is:
In eUML, states are terminals of my state machine domain (sm_domain).
The domain's grammar disables as little as possible (address_of):

struct terminal_grammar : proto::not_<proto::address_of<proto::_> >

// Forward-declare an expression wrapper
template<typename Expr>
struct euml_terminal;

struct sm_domain
    : proto::domain< proto::generator<euml_terminal>, terminal_grammar >

Now, I just implemented serialization for state machines using
boost::serialization which happens to define its own DSEL, archive <<
fsm and archive >> fsm.
As state machines are states in eUML, I created a conflict between
both DSELs: << and >> for serialization and my own eUML, like init_ <<

I solved the conflict by creating a sub-domain for states with a
stricter grammar:

struct state_grammar :
        proto::not_<proto::address_of<proto::_> >,
        proto::not_<proto::shift_right<proto::_,proto::_> >,
        proto::not_<proto::shift_left<proto::_,proto::_> >,
        proto::not_<proto::bitwise_and<proto::_,proto::_> >
struct state_domain
    : proto::domain< proto::generator<euml_terminal>, state_grammar,sm_domain >

As init_ is in the super domain sm_domain, init_ << state is allowed,
but as states are in the sub-domain, archive << fsm calls the correct
boost::serialization overload.

However, I have some doubt that what I'm doing really is ok or happens
to be working by pure luck.
If it's an acceptable usage, sub-domains helped me out of a tight spot
and I'm really going to be tempted to push this further ;-)

proto mailing list

Reply via email to