[Feature Request] Use @doc_private on initializers from MLIR types to hide them from the API docs

[soraros]

Review Mojo's priorities

• I have read the roadmap and priorities and I believe this request falls within the priorities.

What is your request?

As title. Suggested by @lattner on Discord.

More concretely, we should change

fn __init__(inout self, value: Self._mlir_type): ...

to

fn __init__(inout self, *, _mlir: Self._mlir_type): ...

What is your motivation for this change?

So we can hide them in docgen or maybe intellisense to avoid, well, this:

Any other details?

We lose implicit conversion from MLIR values this way, I don't know how important that is.

[lattner]

+1 thank you for filing this! This would be a great improvement to make MLIR stuff an implementation detail. We should also look for the var value : Self._mlir_type pattern and upgrade it to _value for similar reasons.

[soraros]

@lattner Actually, we already have a decorator @doc_private to do this, let's just put them on all the MLIR constructors.

I don't think making the arguments keyword only is that good though. Having implicit conversion in this case can be quite useful.

[lattner]

Oh nice, I didn't realize we had @doc_private. I agree, that seems like a good solution!

[JoeLoser]

Yeah, we should more liberally apply @doc_private for things like this.

[JoeLoser]

A simple grep like `git grep 'mlir' stdlib | grep -i 'fn init' should find all the ones to go chase down if anyone is interested.

[miguelcsx]

Hi @JoeLoser , I started working on this issue, could you assign it to me? thanks 🚀

[JoeLoser]

Hi @JoeLoser , I started working on this issue, could you assign it to me? thanks 🚀

Done, thanks for starting on this!