Would a port of dataclasses to micropython-lib be welcome?

[winni2k]

I think dataclasses are the best. I would be interested in trying my hand at porting the module to micropython-lib. Is that a terrible idea or a great idea?

[davefes]

Rewriting some of my applications using dataclass was my next learning exercise. I had found this:
https://github.com/orgs/micropython/discussions/9327
which seemed to suggest that it was possible.

Have I misunderstood?

[winni2k]

Right! I think I missed a few of the earlier comments when I first read that post and it made no sense to me. Now on a second read, it appears as if there is an implementation of dataclasses, but it has not been shared with micropython-lib?

[davefes]

I will await further developments.

[winni2k]

It looks like there is no implementation available from that discussion (https://github.com/orgs/micropython/discussions/9327#discussioncomment-8635740).

[mattytrentini]

Dataclasses are desirable and would certainly be a welcome feature if they could be implemented with reasonable resource usage - and high compatibility with CPython's implementation.

[winni2k]

I'll see if I can have a go then. However, I am new to micropython (not python). Can you give me a pointer or two on how to measure relevant resource usage?

[winni2k]

I have started by porting a pre-requisite, functools.total_ordering (micropython/micropython-lib#818). I could use a PR review.

Out of the gate, it appears as if dataclasses will not work if they are used to extend, that is, inherit from builtin types because builtin types do not appear to have comparison operators as attributes (for example int.__lt__()).