Extensions Support

[sbchisholm]

trafficstars

Any thoughts on how to implement extensions?

[tanmaykm]

@sbchisholm, I have encountered extensions rarely and haven't really thought of implementing it yet. Any suggestions / ideas?

[sbchisholm]

For the use case when serializing objects to protobuf, extensions provide a way to mirror the objects in the protobuf message and support inheritance. Here's an example of how it can be accomplished in python:

import message_pb2


class A(object):
  def __init__(self, x, y, z):
    self.x = x
    self.y = y
    self.z = z

  def to_protobuf(self):
    proto = message_pb2.A()
    proto.x = self.x
    proto.y = self.y
    proto.z = self.z
    return proto

class B(A):
  def __init__(self, x, y, z, b):
    super(B, self).__init__(x, y, z)
    self.b = b

  def to_protobuf(self):
    proto = super(B, self).to_protobuf()
    ext = proto.Extensions[message_pb2.b_ext]
    ext.b = b
    return proto

message A{
  required string x = 1;
  required string y = 2;
  required string z = 3;
  extensions 100 to max;
}

message B {
  required string b = 1;
}

extend A {
  optional B b_ext = 101;
}

Extensions in python seem to be supported by the ProtoBuf object using a dictionary indexed by the extension number. I think something similar might work for Julia. We would probably want to have accessors for get_extenstion, set_extension, and has_extension.

[tanmaykm]

proto3 deprecates Extensions and introduces an Any type which seems simpler. May be it's better to jump to proto3 instead? Unless there is an immediate requirement.

[sbchisholm]

Moving ahead with proto3 sounds like a good plan, Any does look like a much cleaner approach than the Extensions.