kennethnym
commented
4 months ago thanks for your question. i have indeed looked into cap'n'proto before deciding on creating nanopack. here are my thoughts:
- its schema syntax is more complicated (for example, you don't need to pollute your schema with field numbers)
- it has features i don't need, such as nested structs, reflection, or dynamic schema loading, thus introducing bloat in the runtime
- nanopack has no runtime - the nanopack library simply exports helper methods for reading from a nanopack-formatted buffer
- nanopack exports a much more minimal api.
since nanopack is such a lightweight serialization format, it is easy on my end to support. you can try using nanopack for your project, but it is not stable, and i recommend you compile nanoc from the latest commit on main.
please let me know if you run into any issue.
Good day,
Appreciate your work, I was looking into your Poly project to take some approaches, since I'm working on same architecture kind of project. Basically I need my separate view to interact with native backend via messages and I need to generate code in C++ and TS. So I'm wondering if you ever considered to use capnproto (https://github.com/capnproto/capnproto), since it basically supports all you need (https://polygui.org/nanopack/introduction/), like inheritance in opposite to protobuf. It have custom generators, so what you need to do is to write your own generator (https://capnproto.org/otherlang.html#how-to-write-compiler-plugins).
I'm not saying that you're wrong, but does it really worth to support your own message format is there any significant benefits?