I am designing a behaviour that extends gen_server and thus I want all of gen_server’s callback specification in addition to some additional ones on top as part of the extension. In the past I have favored using the `-callback` attribute as recommended in the documentation but then I saw someone recommend the following when extending gen_server
> behaviour_info(Type) -> gen_server:behaviour_info(Type).
I tried combining the two approaches but such is not allowed (which makes sense). I believe my options are to append my callbacks to gen_server’s list using `behaviour_info` or to copy gen_server’s callback specifications from the OTP source. Copy-pasting code just feels wrong... how bad is it to use `behaviour_info` over `-callback`? And am I taking the right approach?
Thanks,
Brett
I was trying out the approach of the first answer here (https://stackoverflow.com/questions/6745183/erlang-extended-gen-server). In this approach, the extension behaviour get first crack at all of the gen_server messages and then forwards the user-module-specific stuff to the user module’s gen_server implementation.
Basically I have written a bunch of (tested/working) TCP client/server interaction code as a gen_server and have a bunch of other modules that add specific functionality on top but require the same TCP handshaking. My idea was to wrap my working stuff as a behaviour but I can’t yet decide if I want to expose all of gen_server to my user modules or customize the interface. Right now I have opted for generality and am experimenting with exposing everything but we shall see... I am in that awkward place where I am working out my architecture while also trying to learn Erlang best practices; fun times.
Brett
Hah; Fernando after I replied I (think I) realized what you were saying. I now only specify my specific extensions in my behaviour using the `-callback` attribute and then include both my behaviour and gen_server in the user modules via `-behaivour`. I think this is a good solution to my dilemma.
Thanks!
Brett