Rewrite in a real programming language about coreos-assembler HOT 8 CLOSED

I agree that we don't want to wait very long, but we also need to have a high level discussion about this project in general and develop a strategy for it. Right now there hasn't really been any design and it's all been tactical. We need to udevadm settle the tactical changes (i.e. get in what we need for now so it can be useful in the short term) and then kinda freeze on that while we rewrite with toplevel design in mind.

from coreos-assembler.

@lucab Are we assuming that we're working with Python 2? Because with the latest versions of Python 3, there's the ability to have the interpreter and bytecode compilation steps enforce the type checking.

from coreos-assembler.

This seems relevant here: http://mypy-lang.org/.
Anyone has experience with this?

from coreos-assembler.

There is a discussion happening at 17:30 UTC today that may help with the high level design a bit.

from coreos-assembler.

FWIW, I favor Python because it's more expressive, and it's much easier to integrate into other systems (as many of them are written in Python today).

from coreos-assembler.

I have some (brief and stale) experience with mypy. I liked it a lot for small codebase, but I realized it can't replace a real static type system because pervasive use of deply-nested-collections quickly results in unwieldy annotations or useless Any everywhere.

Looking at CL SDK, it contains ~8k of bash and ~1k of python (LoC) as of now. If we expect to have external contributors and reach similar size, I'd be happier to have a compiler in a CI step to catch silly mistakes instead of runtime errors in the pipeline.

Regarding integration, I think the scope here is similar to mantle and the primary mean of consumption is via containerized CLI execution, not as a library importing.

from coreos-assembler.

@Conan-Kudo No, just a language-agnostic comment with my view on scope and wishlist. I'd assume that we are speaking about py3 (barring any legacy lib). And you have a good point, I think that #174 is basically that.

from coreos-assembler.

Deduping with #163 (mantle is Go)

from coreos-assembler.