Tomas Abrahamsson

In retrospect, the approach to using `$` in the generated variable names was maybe not that good. I realized the generated code will for example not survive a round-trip via...

Interesting idea. How much performance would you gain? Would it be possible to modify the generated pb module to try out something like that? From the example it looks like...

> it details exactly the same pattern -- both how to do it and the example of OpenTelemetry's protos -- https://blog.najaryan.net/posts/partial-protobuf-encoding/ Indeed it does. The workaround in that post—to pre-encode...

Hi, interesting stuff! About switching `MsgDef` to a map: I'm getting the impression you are using the `gpb` module directly as opposed to generating an encoder/decoder module and using it?...

> Is that not the case, i.e. you're saying that gpb is basically a parallel implementation of the encoder/decoder? Yes, that's correct, it is a parallel implementation, there is no...

Stable in what sense? I'd say both implementations are fairly well tested. Most work has gone into the generated code. Both since it is a bit more complex problem, but...

Yes, definitely, I think they'd be nice improvements.

Thanks for both PRs. I will take a look.

Maybe you could try the [`use_packages`](https://hexdocs.pm/gpb/gpb_compile.html#option-use_packages) option? I noticed your `Foo` messages are in different packages. With the `use_packages`, messages will get prefixed by package, so you will probably need...

I see now, sorry, didn't get originally that it was also the proto files that had the same name (but in different directories), and got also now that you were...