Herbie Ong

Cycle detection can be expensive, but simple depth detection (levels of subfields) is not expensive and is easier. I'm not familiar with the other language implementations, and I can't seem...

For versions >= 1.20, i.e. https://github.com/protocolbuffers/protobuf-go/, the generated Go code for WKT uses the proto files from a downloaded version of the protobuf source repository that @puellanivis pointed out. The...

You'll need to bring this up on the main Protocol Buffers site at https://github.com/protocolbuffers/protobuf/issues as the Go implementation is simply following the guidelines mentioned in the proto definition. For a...

I do have a use for such helper. Our particular usage will have a `descriptorpb.FileDescriptorSet` as input and we'll need to have a resolver for `prototext.UnmarshalOptions`. I'm good with the...

`EmitUnpopulated` is indeed meant for migration from using jsonpb's `EmitDefaults`. The term "default" is misleading because in proto2, even if there is a default field option, what is emitted for...

> The "consumers" aren't always machines, sometimes they're data scientists ... True. In this case, the feature @neild described above would help in terms of what to project out from...

Fyi. Interesting enough, I noticed there's `FieldMaskUtil::ToJsonString` and somehow that function call returns false. So, I'm guessing that MessageToJsonString does not use the same function to encode FieldMask.

I'm assuming you're asking for the field name. In which case, it may be argued that providing the full path to the field will be better. While that seems nice,...

Yes, the enum name is there and the message can include the enum name easily. I had thought you were asking about the field name/path, hence I had to clarify.

I don't have the bandwidth right now. Earliest I can do this is at the end of the year. If someone else can pick this up sooner, that will be...