I had a funny situation today (simplified):
Ive a communication object of type Adapter, with a #send and a #receive method.
My application communicates with it by DRb.
So in my application, I have a DRbObject, lets say foo.
Now, calling
foo.send(msg, dest)
calls the #send on DRbObject, instead of Adapter.
Most easy fix would be of course to rename the send method. But, I'd like to stay as close to my documentation/underlying API as possible.
What do you think?
Should I rename the send method, or is there a neat (local) hack for this?
DRbObject does it's remote message routing magic with #method_missing. So, clearly, we should undefine the #send method from foo so it will delegate to #method_missing instead!
foo.singleton_class.class_eval { undef_method :send }
Or to do it in a more object-oriented way:
AdapterDRbObject < DRbObject
undef_method :send
end
As for whether you should do this, that's up for debate.
It's actually "okay" to override/remove #send, because everyone is supposed to know that you should always call #__send__ instead. (In fact, if you look at DRbObject#method_missing, it calls #__send__.)
On the other hand, #send is a pretty core concept of Ruby and might confuse future maintainers of the code.