Don't display "hidden" features in `cargo add`

[epage]

docs.rs treats _ prefixed features as hidden. cargo add should reflect that in reporting features to the user.

The question is if we should unconditionally hide the features, hide them if disabled, or do the tracking to hide it unless its directly enabled.

[epage]

See https://internals.rust-lang.org/t/listing-of-cargo-features/16867/6

[epage]

I think the ideal is to hide them if they are not directly enabled.

It should be possible to implement this at end of the preparation when we print the features. See

cargo/src/cargo/ops/cargo_add/mod.rs

Line 699 in 8827baa

if !activated.is_empty() || !deactivated.is_empty() {

[epage]

We've talked about this at the cargo team meeting. We are reluctant to move forward with this for now; we'd like more time before committing to this.

We'd like the decision to be more intentional than "copy what docs.rs did without the background for their decision", including

• We should consider how this impacts any possibility for officially tracking feature visibility or other feature state (unstable, deprecated, etc)
• In a way, the more we codify this, the more we implicitly endorse snake_case as the feature naming style in contrast to kebab-case since _kebab-case would be out of place

[ehuss]

I opened #10882 for a more general discussion of hidden features since I couldn't find an existing issue. This gist contains a list of packages that use features with a leading underscore, which might be interesting to review in more depth to see why and how people are using them.