Improve `Ord` violation help
Recent experience in #128083 showed that the panic message when an Ord violation is detected by the new sort implementations can be confusing. So this PR aims to improve it, together with minor bug fixes in the doc comments for sort*, sort_unstable* and select_nth_unstable*.
Is it possible to get these changes into the 1.81 release? It doesn't change behavior and would greatly help when users encounter this panic for the first time, which they may after upgrading to 1.81.
Tagging `@orlp`
`<[T]>::first_chunk` became const stable in 1.77, but `<[T]>::last_chunk` was
left out. This was fixed in 3488679768, which reached stable in 1.80,
making `<[T]>::last_chunk` const stable as of that version, but it is
documented as being const stable as 1.77. While this is what should have
happened, the documentation should reflect what actually did happen.
- Use if the implementation of [`Ord`] for `T`
language
- Link to total order wiki page
- Rework total order help and examples
- Improve language to be more precise and less
prone to misunderstandings.
- Fix usage of `sort_unstable_by` in `sort_by`
example
- Fix missing author mention
- Use more consistent example input for sort
- Use more idiomatic assert_eq! in examples
- Use more natural "comparison function" language
instead of "comparator function"
This restores the original binary search implementation from #45333
which has the nice property of having a loop count that only depends on
the size of the slice. This, along with explicit conditional moves
from #128250, means that the entire binary search loop can be perfectly
predicted by the branch predictor.
Additionally, LLVM is able to unroll the loop when the slice length is
known at compile-time. This results in a very compact code sequence of
3-4 instructions per binary search step and zero branches.
Fixes#53823
Clean and enable `rustdoc::unescaped_backticks` for `core/alloc/std/test/proc_macro`
I am not sure if the lint is supposed to be "ready enough" (since it is `allow` by default), but it does catch a couple issues in `core` (`alloc`, `std`, `test` and `proc_macro` are already clean), so I propose making it `warn` in all the crates rendered in the website.
Cc: `@GuillaumeGomez`
There are only 3 cases across the crates rendered in the website (`core`,
`alloc`, `std`, `proc_macro` and `test`), and they are all in `core`.
Clean them up, so that the lint can be enabled in the next commit.
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Fix doc nits
Many tiny changes to stdlib doc comments to make them consistent (for example "Returns foo", rather than "Return foo"), adding missing periods, paragraph breaks, backticks for monospace style, and other minor nits.
from_ref, from_mut: clarify documentation
This was brought up [here](https://github.com/rust-lang/rust/issues/56604#issuecomment-2143193486). The domain of quantification is generally always constrained by the type in the type signature, and I am not sure it's always worth spelling that out explicitly as that makes things exceedingly verbose. But since this was explicitly brought up, let's clarify.
The new sort implementations have the ability to
detect Ord violations in many cases. This commit
improves the message in a way that should help
users realize what went wrong in their program.
This includes a `From<legacy::RangeInclusive> for RangeInclusive` impl for convenience, instead of the `TryFrom` impl from the RFC.
Having `From` is highly convenient and the assertion is unlikely to be a problem in practice.
This includes re-exports of all existing `Range` types under `core::range`, plus the range-related traits (`RangeBounds`, `Step`, `OneSidedRange`) and the `Bound` enum.
Currently the iterators are just wrappers around the old range types,
and most other trait impls delegate to the old rage types as well.
Also includes an `.iter()` shorthand for `.clone().into_iter()`
This is possible now that inline const blocks are stable; the idea was
even mentioned as an alternative when `uninit_array()` was added:
<https://github.com/rust-lang/rust/pull/65580#issuecomment-544200681>
> if it’s stabilized soon enough maybe it’s not worth having a
> standard library method that will be replaceable with
> `let buffer = [MaybeUninit::<T>::uninit(); $N];`
Const array repetition and inline const blocks are now stable (in the
next release), so that circumstance has come to pass, and we no longer
have reason to want `uninit_array()` other than convenience. Therefore,
let’s evaluate the inconvenience by not using `uninit_array()` in
the standard library, before potentially deleting it entirely.
Account for things that optimize out in inlining costs
This updates the MIR inlining `CostChecker` to have both bonuses and penalties, rather than just penalties.
That lets us add bonuses for some things where we want to encourage inlining without risking wrapping into a gigantic cost. For example, `switchInt(const …)` we give an inlining bonus because codegen will actually eliminate the branch (and associated dead blocks) once it's monomorphized, so measuring both sides of the branch gives an unrealistically-high cost to it. Similarly, an `unreachable` terminator gets a small bonus, because whatever branch leads there doesn't actually exist post-codegen.
Replace sort implementations
This PR replaces the sort implementations with tailor-made ones that strike a balance of run-time, compile-time and binary-size, yielding run-time and compile-time improvements. Regressing binary-size for `slice::sort` while improving it for `slice::sort_unstable`. All while upholding the existing soft and hard safety guarantees, and even extending the soft guarantees, detecting strict weak ordering violations with a high chance and reporting it to users via a panic.
* `slice::sort` -> driftsort [design document](https://github.com/Voultapher/sort-research-rs/blob/main/writeup/driftsort_introduction/text.md), includes detailed benchmarks and analysis.
* `slice::sort_unstable` -> ipnsort [design document](https://github.com/Voultapher/sort-research-rs/blob/main/writeup/ipnsort_introduction/text.md), includes detailed benchmarks and analysis.
#### Why should we change the sort implementations?
In the [2023 Rust survey](https://blog.rust-lang.org/2024/02/19/2023-Rust-Annual-Survey-2023-results.html#challenges), one of the questions was: "In your opinion, how should work on the following aspects of Rust be prioritized?". The second place was "Runtime performance" and the third one "Compile Times". This PR aims to improve both.
#### Why is this one big PR and not multiple?
* The current documentation gives performance recommendations for `slice::sort` and `slice::sort_unstable`. If for example only one of them were to be changed, this advice would be misleading for some Rust versions. By replacing them atomically, the advice remains largely unchanged, and users don't have to change their code.
* driftsort and ipnsort share a substantial part of their implementations.
* The implementation of `select_nth_unstable` uses internals of `slice::sort_unstable`, which makes it impractical to split changes.
---
This PR is a collaboration with `@orlp.`
Due to refactoring the const_trait usage, the CopyMarker impl was
accidentally deleted, which had the consequence that the Copy
specialization for the small-sort was never picked.
Freeze + Copy types should be allowed to choose between all three
small-sort variants. With the recent changes to small-sort selection,
a regression was added that only let such types choose between network
and fallback. It can now also choose general where appropriate.
const_trait in conjunction with specialization was
deemed not ready for usage in this scenario. So
instead a two-stage trait specialization approach
is used. This approach is likely worse for
compile-times. Future work that enables
const_trait can revert back to the previous
version as outlined in the comment marked
FIXME(effects).
Rollup of 16 pull requests
Successful merges:
- #123374 (DOC: Add FFI example for slice::from_raw_parts())
- #124514 (Recommend to never display zero disambiguators when demangling v0 symbols)
- #125978 (Cleanup: HIR ty lowering: Consolidate the places that do assoc item probing & access checking)
- #125980 (Nvptx remove direct passmode)
- #126187 (For E0277 suggest adding `Result` return type for function when using QuestionMark `?` in the body.)
- #126210 (docs(core): make more const_ptr doctests assert instead of printing)
- #126249 (Simplify `[T; N]::try_map` signature)
- #126256 (Add {{target}} substitution to compiletest)
- #126263 (Make issue-122805.rs big endian compatible)
- #126281 (set_env: State the conclusion upfront)
- #126286 (Make `storage-live.rs` robust against rustc internal changes.)
- #126287 (Update a cranelift patch file for formatting changes.)
- #126301 (Use `tidy` to sort crate attributes for all compiler crates.)
- #126305 (Make PathBuf less Ok with adding UTF-16 then `into_string`)
- #126310 (Migrate run make prefer rlib)
- #126314 (fix RELEASES: we do not support upcasting to auto traits)
r? `@ghost`
`@rustbot` modify labels: rollup
Add `size_of` and `size_of_val` and `align_of` and `align_of_val` to the prelude
(Note: need to update the PR to add `align_of` and `align_of_val`, and remove the second commit with the myriad changes to appease the lint.)
Many, many projects use `size_of` to get the size of a type. However,
it's also often equally easy to hardcode a size (e.g. `8` instead of
`size_of::<u64>()`). Minimizing friction in the use of `size_of` helps
ensure that people use it and make code more self-documenting.
The name `size_of` is unambiguous: the name alone, without any prefix or
path, is self-explanatory and unmistakeable for any other functionality.
Adding it to the prelude cannot produce any name conflicts, as any local
definition will silently shadow the one from the prelude. Thus, we don't
need to wait for a new edition prelude to add it.
Add assert_unsafe_precondition to unchecked_{add,sub,neg,mul,shl,shr} methods
(Old PR is haunted, opening a new one. See #117494 for previous discussion.)
This ensures that these preconditions are actually checked in debug mode, and hopefully should let people know if they messed up. I've also replaced the calls (I could find) in the code that use these intrinsics directly with those that use these methods, so that the asserts actually apply.
More discussions on people misusing these methods in the tracking issue: https://github.com/rust-lang/rust/issues/85122.
Actually use TAIT instead of emulating it
`core`'s `impl_fn_for_zst` macro is just a hacky way of emulating TAIT. TAIT has become stable enough to be used [in other places](e8fbd99128/library/std/src/backtrace.rs (L431)) inside the standard library, so let's use it in `core` as well.
