Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

RFC for doc_cfg, doc_cfg_auto, doc_cfg_hide and doc_cfg_show features #3631

Open
wants to merge 6 commits into
base: master
Choose a base branch
from
281 changes: 281 additions & 0 deletions text/000-rustdoc-cfgs-handling.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,281 @@
Rustdoc: stabilization of the `doc(cfg*)` attributes

- Features Name: `doc_cfg`
- Start Date: 2022-12-07
- RFC PR: [rust-lang/rfcs#0000](https://github.com/rust-lang/rfcs/pull/0000)
- Rust Issue: [rust-lang/rust#43781](https://github.com/rust-lang/rust/issues/43781)


# Summary
[summary]: #summary

This RFC aims at providing rustdoc users the possibility to add visual markers to the rendered documentation to know under which conditions an item is available (currently possible through the following unstable features: `doc_cfg`, `doc_auto_cfg` and `doc_cfg_hide`).

It does not aim to allow having a same item with different `cfg`s to appear more than once in the generated documentation.

It does not aim to document items which are *inactive* under the current configuration (i.e., “`cfg`'ed out”).
Comment on lines +12 to +16
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

...I'm slightly confused by this preamble, honestly, because I would figure the most important time to know when something is available is when it is, in fact, not available... when it has been "cfg'd out" but can be enabled by other means.

Copy link
Member

@fmease fmease Oct 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are right, this sentence doesn't tell the full story and should be improved (however, it's technically correct). I'm pretty sure the RFC does elaborate on it further down in the document (Ctrl+F doc (Match Case, Match Full Words)).

It implicitly refers to rust-lang/rust#1998.

As you can probably guess, rustdoc doesn't and won't magically recompile the to-be-documented crate under all possible cfgs and combinations thereof to gather the information necessary to display the "portability info box".

So for the following crate, function wouldn't show up in the generated docs unless you actually passed --cfg special to rustdoc:

#![feature(doc_auto_cfg)] // (A)
// #![feature(doc_cfg)] // (B)

// #[doc(cfg(special))] (B)
#[cfg(special)]
pub fn function() {}

Therefore, the common and offical workaround is the use of the semi-special cfg doc:

#![feature(doc_auto_cfg)] // (A)
// #![feature(doc_cfg)] // (B)

// #[doc(cfg(special))] (B)
#[cfg(any(doc, special))]
pub fn function() {}

This gets more hairy if there are "multiple" functions called function (where the cfg specs differ for obvious reasons), one needs to choose which of the many functions to annotate with doc.

Of course, this doesn't scale to more complicated cases. E.g., if the parameter/return types differ by cfg spec. So, yeah. The general problem isn't solved. That's what I meant with the sentence you highlighted.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I mean, I've certainly written documentation that needs that invocation (or similar) to work properly. I guess I'm just feeling a bit perplexed that there is seemingly quite a lot of new concepts in this RFC and that's left unsolved. I would expect an RFC that has the sense of adding a Lot Of Stuff to also have an answer to that critical missing piece, because it's a right pain in the ass! ...or I would expect it to be smaller and more tightly scoped.

I realize that's a Pure Vibes thing, but I do think it can be a bit hard to evaluate a solution... and thus people would come to consensus slower... if it feels incorrectly scoped.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

rustdoc doesn't and won't magically recompile the to-be-documented crate under all possible cfgs and combinations thereof

i mean, rustc now tracks enough information that it doesn’t have to
rust-lang/rust#109005

Copy link
Member

@fmease fmease Oct 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i mean, rustc now tracks enough information that it doesn’t have to

Well, rustc only stores the name of cfg'ed out items but rustdoc would need to have access to so much more (struct fields, function signatures, ...). See StrippedCfgItem.

Edit: And doing that would drastically increase size of the crate metadata / rustc's memory use.


# Motivation
[motivation]: #motivation

The goal of this RFC is to stabilize the possibility to add visual markers to the rendered documentation to know under which conditions an item is available.

Providing this information to users will solve a common issue: “Why can I see this item in the documentation and yet can't use it in my code?”.
The end goal being to provide this information automatically so that the documentation maintenance cost won't increase.


# Guide-level explanation
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I mentioned this as a regular comment, but I'll move it to a review so it gets noticed: I think that the guide-level explanation should clarify the cfg(any(doc, ...)) trick for making auto_cfg work properly, since it's not immediately clear otherwise. Basically, conditional compilation is still applied before rustdoc runs, meaning you have to ensure that rustdoc can see code to document it. Then, because auto_cfg by default ignores cfg(doc), only the other cfgs are documented.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It was not ignored. You can see the RFC mentioning this here. If it's not clear enough, how would you word it?

[guide-level-explanation]: #guide-level-explanation

This RFC proposes to add the following attributes:

* `#![doc(auto_cfg(enable))]`/`#[doc(auto_cfg(disable))]`
GuillaumeGomez marked this conversation as resolved.
Show resolved Hide resolved

When this is turned on, `#[cfg]` attributes are shown in documentation just like `#[doc(cfg)]` attributes are. By default, `auto_cfg` will be enabled.

* `#[doc(cfg(...))]`

This attribute is used to document the operating systems, feature flags, and build profiles where an item is available. For example, `#[doc(cfg(unix))` will add a tag that says "this is supported on **unix** only" to the item.

The syntax of this attribute is the same as the syntax of the [`#[cfg(unix)]` attribute][cfg attribute] used for conditional compilation.

* `#![doc(cfg_hide(...))]` / `#[doc(cfg_show(...))]`

These attributes suppress or un-suppress the `auto_cfg` behavior for a particular configuration predicate.

For example, `#[doc(cfg_hide(windows))]` could be used in newer versions of the [`windows` crate] to prevent the "this is supported on **windows** only" tag from being shown on every single item.

[cfg attribute]: https://doc.rust-lang.org/reference/conditional-compilation.html
[`windows` crate]: https://docs.rs/windows/latest/windows/

All of these attributes can be added to a module or to the crate root, and they will be inherited by the child items unless another attribute overrides it (except that `doc(cfg)` cannot be added to the crate root). This is why "opposite" attributes like `cfg_hide` and `cfg_show` are provided: they allow a child item to override its parent.
GuillaumeGomez marked this conversation as resolved.
Show resolved Hide resolved


# Reference-level explanation
[reference-level-explanation]: #reference-level-explanation

## The attributes

### `#[doc(auto_cfg(enable))]`/`#[doc(auto_cfg(disable))]`

This is a crate-level attribute. By default, `#[doc(auto_cfg)]` is enabled at the crate-level. When it's enabled, Rustdoc will automatically display `cfg(...)` compatibility information as-if the same `#[doc(cfg(...))]` had been specified.
GuillaumeGomez marked this conversation as resolved.
Show resolved Hide resolved

So if we take back the previous example:

```rust
#[cfg(feature = "futures-io")]
pub mod futures {}
```

There's no need to "duplicate" the `cfg` into a `doc(cfg())` to make Rustdoc display it.

In some situations, the detailed conditional compilation rules used to implement the feature might not serve as good documentation (for example, the list of supported platforms might be very long, and it might be better to document them in one place). To turn it off, add the `#[doc(auto_cfg(disable))]` attribute at the crate-level.

### `#[doc(cfg(...))]`

This attribute provides a standardized format to override `#[cfg()]` attributes to document conditionally available items. Example:

```rust
// the "real" cfg condition
#[cfg(feature = "futures-io")]
// the `doc(cfg())` so it's displayed to the readers
#[doc(cfg(feature = "futures-io"))]
pub mod futures {}
```

It will display in the documentation for this module:

![This is supported on feature="futures-io" only.](https://user-images.githubusercontent.com/81079/89731116-d7b7ce00-da44-11ea-87c6-022d192d6eca.png)

This attribute has the same syntax as conditional compilation, but it only causes documentation to be added. This means `#[doc(cfg(not(windows)))]` will not cause your docs to be hidden on non-windows targets, even though `#[cfg(not(windows))]` does do that.

This attribute works on modules and on items but cannot be used at the crate root level.

### `#[doc(cfg_hide(...))]`

This attribute is used to prevent some `cfg` to be generated in the visual markers. It only applies to `#[doc(auto_cfg(enable))]`, not to `#[doc(cfg(...))]`. So in the previous example:

```rust
#[cfg(any(unix, feature = "futures-io"))]
pub mod futures {}
```

It currently displays both `unix` and `feature = "futures-io"` into the documentation, which is not great. To prevent the `unix` cfg to ever be displayed, you can use this attribute at the crate root level:

```rust
#![doc(cfg_hide(unix))]
```

Or directly on a given item/module as it covers any of the item's descendants:
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Again the current implementation only supports being set at the crate-level, are there usecases that want to be able to limit it to certain sub-trees?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The one case I'm thinking about is to reduce complexity of rendered doc cfg as described in this comment.


```rust
#[doc(cfg_hide(unix))]
#[cfg(any(unix, feature = "futures-io"))]
pub mod futures {
// `futures` and all its descendants won't display "unix" in their cfgs.
}
```

Then, the `unix` cfg will never be displayed into the documentation.

Rustdoc currently hides `doc` and `doctest` attributes by default and reserves the right to change the list of "hidden by default" attributes.
GuillaumeGomez marked this conversation as resolved.
Show resolved Hide resolved

The attribute accepts only a list of identifiers or key/value items. So you can write:

```rust
#[doc(cfg_hide(unix, doctest, feature = "something"))]
#[doc(cfg_hide())]
```

But you cannot write:

```rust
#[doc(cfg_hide(not(unix)))]
```

If `cfg_show` and `cfg_hide` are used to show/hide a same `cfg` on a same item, it'll emit an error. Example:

```rust
#[doc(cfg_hide(unix))]
#[doc(cfg_show(unix))] // Error!
pub fn foo() {}
```

### `#[doc(cfg_show(...))]`

This attribute does the opposite of `#[doc(cfg_hide(...))]`: if you used `#[doc(cfg_hide(...))]` and want to revert its effect on an item and its descendants, you can use `#[doc(cfg_show(...))]`.
It only applies to `#[doc(auto_cfg(enable))]`, not to `#[doc(cfg(...))]`.

For example:

```rust
#[doc(cfg_hide(unix))]
#[cfg(any(unix, feature = "futures-io"))]
pub mod futures {
// `futures` and all its descendants won't display "unix" in their cfgs.
#[doc(cfg_show(unix))]
pub mod child {
// `child` and all its descendants will display "unix" in their cfgs.
}
}
```

The attribute accepts only a list of identifiers or key/value items. So you can write:

```rust
#[doc(cfg_show(unix, doctest, feature = "something"))]
#[doc(cfg_show())]
```

But you cannot write:

```rust
#[doc(cfg_show(not(unix)))]
```

If `cfg_show` and `cfg_hide` are used to show/hide a same `cfg` on a same item, it'll emit an error. Example:

```rust
#[doc(cfg_hide(unix))]
#[doc(cfg_show(unix))] // Error!
pub fn foo() {}
```

## Inheritance

Rustdoc merges `cfg` attributes from parent modules to its children. For example, in this case, the module `non_unix` will describe the entire compatibility matrix for the module, and not just its directly attached information:

```rust
#[doc(cfg(any(windows, unix)))]
pub mod desktop {
#[doc(cfg(not(unix)))]
pub mod non_unix {
//
}
}
```

> ![Available on (Windows or Unix) and non-Unix only.](https://hackmd.io/_uploads/SJrmwYeF2.png)

[Future versions of rustdoc][boolean simplification] may simplify this display down to "available on **Windows** only."

### Re-exports and inlining

`cfg` attributes of a re-export are never merged with the re-exported item(s) attributes except if the re-export has the `#[doc(inline)]` attribute. In this case, the `cfg` of the re-exported item will be merged with the re-export's.

When talking about "attributes merge", we mean that if the re-export has `#[cfg(unix)]` and the re-exported item has `#[cfg(feature = "foo")]`, you will only see `cfg(unix)` on the re-export and only `cfg(feature = "foo")` on the re-exported item, unless the re-export has `#[doc(inline)]`, then you will only see the re-exported item with both `cfg(unix)` and `cfg(feature = "foo")`.

Example:

```rust
#[doc(cfg(any(windows, unix)))]
pub mod desktop {
#[doc(cfg(not(unix)))]
pub mod non_unix {
//
}
}

#[doc(cfg(target_os = "freebsd"))]
pub use desktop::non_unix as non_unix_desktop;
#[doc(cfg(target_os = "macos"))]
#[doc(inline)]
pub use desktop::non_unix as inlined_non_unix_desktop;
```

In this example, `non_unix_desktop` will only display `cfg(target_os = "freeebsd")` and not display any `cfg` from `desktop::non_unix`.

On the contrary, `inlined_non_unix_desktop` will have cfgs from both the re-export and the re-exported item.

# Drawbacks
[drawbacks]: #drawbacks

A potential drawback is that it adds more attributes, making documentation more complex.


# Rationale and alternatives
[rationale-and-alternatives]: #rationale-and-alternatives

## Why not merging cfg and doc(cfg) attributes by default?

It was debated and implemented in [rust-lang/rust#113091](https://github.com/rust-lang/rust/pull/113091).

When re-exporting items with different cfgs there are two things that can happen:

1. The re-export uses a subset of cfgs, this subset is sufficient so that the item will appear exactly with the subset
2. The re-export uses a non-subset of cfgs like in this code:
```rust
#![feature(doc_auto_cfg)]

#[cfg(target_os = "linux")]
mod impl_ {
pub fn foo() { /* impl for linux */ }
}

#[cfg(target_os = "macos")]
mod impl_ {
pub fn foo() { /* impl for darwin */ }
}

pub use impl_::foo;
```
If the non-subset cfgs are active (e.g. compiling this example on windows), then this will be a compile error as the item doesn't exist to re-export. If the subset cfgs are active it behaves like described in 1.


# Unresolved questions
[unresolved-questions]: #unresolved-questions


# Future possibilities
[future possibilities]: #future-possibilities

## Boolean simplification
[boolean simplification]: #boolean-simplification

> ![Available on (Windows or Unix) and non-Unix only.](https://hackmd.io/_uploads/SJrmwYeF2.png)

Of course, the above example is equivalent to "available on **Windows** only."

We probably don't want to make promises one way or the other about whether rustdoc does this, but for compatibility's sake, Rustdoc does promise that `#[doc(cfg(false))]` will not hide the documentation. This means simplification can be added, and it won't cause docs to mysteriously vanish.

This is tracked in issue [rust-lang/rust#104991](https://github.com/rust-lang/rust/issues/104991).