Can protocol buffer be partial updated? - c++

Question

I am new to protobuf and here is my question: Can protocol buffer support partial update?
For example, I have such messages:
package model.test;
message Person{
required int32 id = 1;
required string name = 2;
repeated PhoneNumber phone = 3;
}
enum PhoneType{
MOBILE = 0;
HOME = 1;
WORK = 2;
}
message PhoneNumber{
required string number = 1;
optional PhoneType type = 2 [default = HOME];
}
Now the data I have like that:
model::test::Person person;
person.set_id(1);
person.set_name("Jack");
model::test::PhoneNumber* _phone3 = person.add_phone();
_phone3->set_number("123567");
_phone3->set_type(model::test::MOBILE);
model::test::PhoneNumber* _phone4 = person.add_phone();
_phone4->set_number("347890");
_phone4->set_type(model::test::WORK);
The case is that when only work phone number is changed, I have to update the whole person object with the following codes.
fstream out("User.txt", ios::out | ios::binary | ios::trunc);
person.SerializePartialToOstream(&out);
But it is not efficient to do that. I want to only update the PhoneNumber, Is there any partial update in protolbuf or something like that?

Answer 1

Protocol buffers are actually designed such that concatenation is the same as merge, and that the last field wins when merging (except for repeated, which are added). In your case, you should actually be able to serialize a blob containing just the phone-number set, and append this data, and it will over-ride the earlier value. This, however, only works well for the root object. Which yours: isn't. And it doesn't work for repeated, which yours: is.

Answer 2

I don't think there is any support for what you want to do. If you think about it, it doesn't really make sense that some sort of partial update serialization would exist in the first place. For protobuf to be able to manipulate an object that is serialized in a file on disk, it needs to read and deserialize the whole object so it knows what fields have been previously populated. Then when serializing and writing the updated object back to disk, you're going to have to overwrite the old file no matter what you do (i.e. you can't shove extra bytes into a file on the file system without overwriting the original file completely).