code smell?

[nrdxp]

Isn't the need for such a thing a blatant code smell? Is there any upstream issues to address this?

[blaggacao]

Might be.

I have raised the issue in the flake matrix chat and the Unofficial Flakes Roadmap has an idea about "input groups", which is somewhat related.

I'm not sure, however, if there already is (strong) consensus that the flake.lock file should be concise / terse and not hold extra entries.

I think the only way to achieve this ahead of full eval time is to define input groups who correspond to an output such as:

myOutputInputs -> self.myOutput

This way, we could declare which outputs a consumer intents to use and only fetch inputs from this particular output.

On the other hand, it might be that the correct answer to "input groups", might actually be subflakes.