Protobuf wrapper class serialized as nil

[travisofthenorth]

Please describe the issue

Apologies in advance if this has nothing to do with gruf. I have a pretty straight-forward server/client setup and I'm serializing proto3 objects in the response. I'm seeing a strange issue where an int wrapper object with a value of 0, i.e. <Google::Protobuf::Int32Value: value: 0>, comes over the wire to the client as nil.

How to Reproduce

Serialize a proto object with a property as described in the issue.

What should happen?

The response on the client side should contain <Google::Protobuf::Int32Value: value: 0> for the particular attribute instead of nil

Anything else we should know?

Server:

• google-protobuf 3.11
• grpc 1.24
• gruf 2.8.0
• Ruby 2.6.3
• OS: I think it's a debian docker container

[splittingred]

Hi @travisofthenorth - fields in Protobuf are always optional - meaning that any unspecified value will always be nil if not passed. This allows you to do proper presence checks, which allow partial updates to happen.

For instance, take a description - field:

message Product {
   uint64 id = 1;
   string name = 2;
   string description = 3;
}

Lets say I wanted to update Product, but didn't want to touch the description. Right now, description is a native string, which means it will always come over as a string. This means if not set by the client, it will come as "". That's problematic for us, as how would the client differentiate between an empty description (a valid value!) and one not passed?

Google provides wrapper types, such as Google::Protobuf::StringValue, to help with this. Lets change our message:

message Product {
   uint64 id = 1;
   google.protobuf.StringValue name = 2;   
   google.protobuf.StringValue description = 3;
}

Now, we can pass either the name or description optionally, and in our code check for nil to see if the client meant to update that field.

So, in other words - this behavior is expected and intentional. Thanks!

[travisofthenorth]

@splittingred I'm explicitly passing in Google::Protobuf::Int32Value.new(value: 0)

[travisofthenorth]

In any case, another one of my clients doesn't seem to have this problem, so it must be something silly like a gem version in the particular client I'm looking at.

Thanks for taking a look, and apologies for the false alarm!

[travisofthenorth]

If anyone encounters this error again, this was a bug in protobuf fixed here: protocolbuffers/protobuf#7195

[splittingred]

@travisofthenorth Oh! Interesting, I misunderstood your question then (didn't realize you were asking about the wrapper value vs the main message).

Thanks for linking the protobuf fix. Appreciate it!