I'm working on improving an in-house messaging library, designed to send messages internally within our applications, and to external consumers. A message consists of a MessageType
(enum class
) and some data (struct
). Each MessageType::
corresponds to a particular data type (e.g., MessageType::TypeA
will always contain Foo
). However, multiple message types could use the same struct (e.g., MessageType::TypeM
could also use Foo
).
We have a class that can send messages. Our previous implementation of the message sender class defines a method for each type:
SendMessageTypeA(Foo data)
SendMessageTypeB(Bar data)
SendMessageTypeM(Foo data)
When there are lots of messages, this can result in a lot of code duplication (the method body is essentially the same, with the exception of the different parameter types).
I've implemented a new method:
template<typename structType>
void Send(MessageType msgType, const structType & messageData)
This single method can send any message, depending on the appropriate template parameter being provided.
The problem is that this new method does not enforce the relationship between MessageType
and struct
. For example, Send<Foo>(MessageType::TypeB, data)
will compile, even though MessageType::TypeB
should contain Bar
. The mismatch will be detected at runtime, but I'd like to make it a compile time error.
I'm not sure how to achieve this. I've considered:
- Declaring all the
SendMessageX()
methods, and use them to callSend<MessageX>()
. This does reduce the duplication, but I still have to create a new method every time a message is defined. - Attempting to use
static_assert
to catch the mismatch. I'm not sure how to mapMessageType
s to their desiredstruct
. - I'm barking up the wrong tree
Aucun commentaire:
Enregistrer un commentaire