Thanks. Using an abstract base class isn't a problem at the
moment, the message types only implement the interface.
However, in case there's a situation in the future where we would
have to use an interface I'll leave the question open.
On Monday, 20 February 2012 at 16:21:37 UTC, Daniel Murphy wrote:
Does message _have_ to be an interface? With an abstract class
the offset
will always be zero, so
T my_cast(T : Message)(Message m)
{
debug
return cast(T)m;
else
return cast(T)cast(void*)m;
}
"Rene Zwanenburg" <[email protected]> wrote in message
news:[email protected]...
A part of the project I'm currently working on relies heavily
on messages being sent between objects. There's some sort of
container object which receives the messages, and notifies any
listeners registered to it. We can't send the messages to the
final recipients directly because the sender doesn't know about
the receivers existence and vice versa.
There are many types of messages, but they all implement a
Message interface. The messageListeners cast the Message to
some ConcreteMessage, but the casting is starting to take a
significant percentage of execution time. Due to the way
MessageListeners are registered, we can be sure a
MessageListener only receives the type of messages it's
interested in, so the cast will never fail.
I'm looking for a static_cast construct, which will be checked
during debug, but in release mode it should minimize overhead
by simply adding the interface offset to the pointer. I've
tried to make use off Type.classinfo's interfaces member, but
for some reason a Type's classinfo isn't available at compile
time.
So my question comes down to: does anyone know of a way to get
static_cast behavior for upcasts?
