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

Some more documentation work #366

Merged
merged 11 commits into from
Apr 2, 2018

Conversation

pitdicker
Copy link
Contributor

  • With #![doc(test(attr(allow(unused_variables), deny(warnings))))] we now make sure all doctest warnings do not get ignored, see Use imported thread_rng() in main doc example. #363.
  • I made small tweaks to the documentation: linkified some things, shortened the initial description of some structs, less or more documentation for re-exports.
  • Better documentation and examples for JitterRng.
  • Reordered the methods of Rng, so they show up in a somewhat more logical order in the documentation.
  • Removed hard-coded links to rand_core on docs.rs. Because Rand re-exports the things that are linked to, we can just link to the documentation of the re-exports.

I think this is best to review per commit.

@@ -88,6 +88,7 @@ impl<X: SampleRange> Distribution<X> for Range<X> {
/// Helper trait for creating objects using the correct implementation of
/// `RangeImpl` for the sampling type; this enables `Range::new(a, b)` to work.
pub trait SampleRange: PartialOrd+Sized {
/// Actual `RangeImpl` implementation for `X`.
Copy link
Member

Choose a reason for hiding this comment

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

I'd prefer you didn't use 'actual'; try "The RangeImpl implementation supporting type X."

Also, it would make sense now to rename XT and T to something else (perhaps Impl; it's not something users will see a lot). This is what Huon suggested I think.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

That sounds better, as always.

#![deny(missing_debug_implementations)]
#![doc(test(attr(allow(unused_variables), deny(warnings))))]
Copy link
Member

Choose a reason for hiding this comment

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

I generally prefer warnings to errors (#![warn(...)]) simply because new compiler versions may add warnings without considering backwards compatibility, thus breaking builds. In the latter case unfortunately we have no choice since warnings from examples are not shown; also "missing docs" is not something the compiler can break so that's fine.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Good point, I'll keep that in mind!

src/lib.rs Outdated
#[cfg(feature="std")] pub use os::OsRng;
#[doc(inline)] pub use jitter::JitterRng;
#[cfg(feature="std")]
#[doc(inline)] pub use os::OsRng;
Copy link
Member

Choose a reason for hiding this comment

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

Keep the OsRng import on a single line please; this causes chaos with diff tools.

src/lib.rs Outdated
@@ -489,61 +388,101 @@ pub trait Rng: RngCore {
Range::sample_single(low, high, self)
}

/// Return a bool with a 1 in n chance of true
/// Sample a new value, using the given distribution.
Copy link
Member

Choose a reason for hiding this comment

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

This commit is just reordering, and this is the diff tool getting confused?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, I made sure not to change anything but the order.

@dhardy
Copy link
Member

dhardy commented Apr 1, 2018

Needs rebase

@pitdicker
Copy link
Contributor Author

That rebase was harder than I thought 😄. I removed the commit that reordered the Rng methods, and recreated it at the end, to make sure I did not lose changes during the rebase.

@pitdicker pitdicker force-pushed the deny_doctest_warnings branch 3 times, most recently from ac4290a to dc381cf Compare April 2, 2018 08:01
@pitdicker
Copy link
Contributor Author

pitdicker commented Apr 2, 2018

Added three more commits, that may need some explanation.

cargo-deadlinks proved really useful to make sure all documentation links are valid. At least for rand, there are two links in rand_core I don't know what to do with. I have added it to the CI, but don't know if installing cargo-deadlinks will take too much time.

I set the dependency to rand_core to a path in the repro so it uses the fixed links. When rand_core 0.1 is released, everything should work out and rand master can again depend on the crates.io version. But it seems like a good idea to me to depend on a relative path on master sometimes.

The line "[rand_core::impls]: ../rand_core/impls/index.html" has the link go out of the rand_core crate, and then back in. This way it also works for all the re-exported functions in rand. At least locally and on GitHub pages. I added a section to cargo.toml to give hints to docs.rs (documentation), so it hopefully builds rand_core together with the rand_crate.

Finally I tried to clean up the list of exported items in libs.rs. At least in the documentation by hiding things, not by breaking the API. A few more things are now directly available like ReadRng and ReseedingRng, instead of under a module with only one item.

Two exceptions: I broke JitterRng by only exporting the module (couldn't get the links to work from both places). We need to have the module public, because it also contains an error type. And I think not having it at the top level import can even be the right thing... Found one very recent dependency on GitHub that uses it directly.

Also the reseeding module is now not exported any more. We can basically consider it a rewrite, and all existing code using it will have to update. No point in making sure the export does not break 😄.

One oddity: even when we build the documentation with "--all-features" XorShiftRng and IsaacRng don't show they implement Serialize and Deserialize. This is due to rust-lang/rust#36922.

Copy link
Member

@dhardy dhardy left a comment

Choose a reason for hiding this comment

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

I see a couple more significant import changes too:

  • no longer expose Hc128Rng directly; 👍
  • make reseeding module private and expose ReseedingRng directly; breaking (except any usage will be broken anyway); not quite sure about this

I'd have preferred it if you'd brought up the breaking changes in a separate issue/PR. The only one I'm really concerned about is the read module.

.travis.yml Outdated
@@ -34,6 +34,8 @@ matrix:
- cargo test --features serde-1,log,nightly
- cargo test --benches
- cargo doc --no-deps --all --all-features
- cargo --list | egrep "^\s*deadlinks$" -q || cargo install cargo-deadlinks
- cargo deadlinks
Copy link
Member

Choose a reason for hiding this comment

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

I've not used it but apparently it only checks local links by default. Is this what we want?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, only the documentation cross-links. But I suppose checking all links can be useful...

src/lib.rs Outdated
pub use rand_core::{RngCore, BlockRngCore, CryptoRng, SeedableRng};
pub use rand_core::{ErrorKind, Error};

// external rngs
#[doc(inline)] pub use jitter::JitterRng;
Copy link
Member

Choose a reason for hiding this comment

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

No longer exporting JitterRng is a breaking change, but I approve.

src/lib.rs Outdated
#[cfg(feature="std")] pub use entropy_rng::EntropyRng;
#[cfg(feature="std")] pub use os::OsRng;
#[cfg(feature="std")] pub use read::ReadRng;
Copy link
Member

Choose a reason for hiding this comment

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

This is more convenient but still a significant change. Do we want it? Potentially a Read wrapper (#254) could also live in rand::read.

@pitdicker
Copy link
Contributor Author

I'd have preferred it if you'd brought up the breaking changes in a separate issue/PR.

Yes, that would have been better. It was kind of necessary to make the links checking work out, so it would have been hard to separate.

@pitdicker
Copy link
Contributor Author

Updated the commits.

@pitdicker pitdicker mentioned this pull request Apr 2, 2018
@pitdicker pitdicker merged commit ba88dad into rust-random:master Apr 2, 2018
@pitdicker pitdicker deleted the deny_doctest_warnings branch April 2, 2018 14:00
pitdicker added a commit that referenced this pull request Apr 4, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants