(Most of these are from a review by joshtriplett. Thanks!)
Fix errors in `as_pin_ref` and `as_pin_mut` in the "Adapters for
working with references" overview.
Reword some headings about transformation methods.
Reclassify `map`, `map_or`, `map_or_else`, `map_err`, etc. to more
accurately reflect which variants they transform.
Document `Debug` requirement for `get_or_insert_default`.
Reword text about `take` and `replace` to be more accurate.
Add examples for the `Product` and `Sum` traits.
Also:
Move link reference definitions closer to their uses. Warn about making
link reference definintions for `err` and `ok`. Avoid making other link
reference definitions that might conflict in the future (foreign methods
that share a name with local ones, etc.)
Write out the generics of `Option` and `Result` when the following
text refers to the type parameters.
Due to the std/alloc split, it is not possible to make
`alloc::collections::TryReserveError::AllocError` non-exhaustive without
having an unstable, doc-hidden method to construct (which negates the
benefits from `#[non_exhaustive]`.
Document associativity of iterator folds.
Document the associativity of `Iterator::fold` and
`DoubleEndedIterator::rfold` and add examples demonstrating this.
Add links to direct users to the fold of the opposite associativity.
Better errors for Debug and Display traits
Currently, if someone tries to pass value that does not implement `Debug` or `Display` to a formatting macro, they get a very verbose and confusing error message. This PR changes the error messages for missing `Debug` and `Display` impls to be less overwhelming in this case, as suggested by #85844. I was a little less aggressive in changing the error message than that issue proposed. Still, this implementation would be enough to reduce the number of messages to be much more manageable.
After this PR, information on the cause of an error involving a `Debug` or `Display` implementation would suppressed if the requirement originated within a standard library macro. My reasoning was that errors originating from within a macro are confusing when they mention details that the programmer can't see, and this is particularly problematic for `Debug` and `Display`, which are most often used via macros. It is possible that either a broader or a narrower criterion would be better. I'm quite open to any feedback.
Fixes#85844.
Add comments around code where ordering is important due for panic-safety
Iterators contain arbitrary code which may panic. Unsafe code has to be
careful to do its state updates at the right point between calls that may panic.
As requested in https://github.com/rust-lang/rust/pull/86452#discussion_r655153948
r? `@RalfJung`
Iterators contain arbitrary code which may panic. Unsafe code has to be
careful to do its state updates at the right point between calls
that may panic.
Rollup of 11 pull requests
Successful merges:
- #85054 (Revert SGX inline asm syntax)
- #85182 (Move `available_concurrency` implementation to `sys`)
- #86037 (Add `io::Cursor::{remaining, remaining_slice, is_empty}`)
- #86114 (Reopen#79692 (Format symbols under shared frames))
- #86297 (Allow to pass arguments to rustdoc-gui tool)
- #86334 (Resolve type aliases to the type they point to in intra-doc links)
- #86367 (Fix comment about rustc_inherit_overflow_checks in abs().)
- #86381 (Add regression test for issue #39161)
- #86387 (Remove `#[allow(unused_lifetimes)]` which is now unnecessary)
- #86398 (Add regression test for issue #54685)
- #86493 (Say "this enum variant takes"/"this struct takes" instead of "this function takes")
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Add MIR pass to lower call to `core::slice::len` into `Len` operand
During some larger experiment with range analysis I've found that code like `let l = slice.len()` produces different MIR then one found in bound checks. This optimization pass replaces terminators that are calls to `core::slice::len` with just a MIR operand and Goto terminator.
It uses some heuristics to remove the outer borrow that is made to call `core::slice::len`, but I assume it can be eliminated, just didn't find how.
Would like to express my gratitude to `@oli-obk` who helped me a lot on Zullip
fix panic-safety in specialized Zip::next_back
This was unsound since a panic in a.next_back() would result in the
length not being updated which would then lead to the same element
being revisited in the side-effect preserving code.
fixes#86443
Alter std::cell::Cell::get_mut documentation
I felt that there was some inconsistency between between Cell and RefCell with regards to their `get_mut` method documentation: `RefCell` flags this method as "unusual" in that it takes `&mut self`, while `Cell` does not. I attempted to flag this in `Cell`s documentation as well, and point to `RefCell`s method in the case where it is required.
Find relevant parts of docs and the new version below.
The current docs for `Cell::get_mut`:
> Returns a mutable reference to the underlying data.
This call borrows Cell mutably (at compile-time) which guarantees that we possess the only reference.
And `RefCell::get_mut`:
> Returns a mutable reference to the underlying data.
This call borrows `RefCell` mutably (at compile-time) so there is no need for dynamic checks.
However be cautious: this method expects self to be mutable, which is generally not the case when using a `RefCell`. Take a look at the `borrow_mut` method instead if self isn’t mutable.
Also, please be aware that this method is only for special circumstances and is usually not what you want. In case of doubt, use `borrow_mut` instead.
My attempt to make `Cell::get_mut` clearer:
> Returns a mutable reference to the underlying data.
This call borrows `Cell` mutably (at compile-time) which guaranteesthat we possess the only reference.
However be cautious: this method expects `self` to be mutable, which is generally not the case when using a `Cell`. If you require interior mutability by reference, consider using `RefCell` which provides run-time checked mutable borrows through its `borrow_mut` method.
This was unsound since a panic in a.next_back() would result in the
length not being updated which would then lead to the same element
being revisited in the side-effect preserving code.
Document the associativity of `Iterator::fold` and
`DoubleEndedIterator::rfold` and add examples demonstrating this.
Add links to direct users to the fold of the opposite associativity.
Make `sum()` and `product()` documentation hyperlinks refer to `Iterator` methods.
The previous linking seemed confusing: within "the sum() method on iterators", "sum()" was linked to `Sum::sum`, not `Iterator::sum`, even though the sentence is talking about the latter. I have rewritten the sentence to be, I believe, clearer, as well as changing the link destinations; applying the same change to the `Product` documentation as well as `Sum`.
I reviewed other traits in the same module and did not see similar issues, and previewed the results using `./x.py doc library/std`.
This method on the Iterator trait is doc(hidden), and about half of
implementations were doc(hidden). This adds the attribute to the
remaining implementations.
The previous linking seemed confusing: within "the sum() method on
iterators", "sum()" was linked to `Sum::sum`, not `Iterator::sum`, even
though the sentence is talking about the latter.
I have rewritten the sentence to be, I believe, clearer, as well as
changing the link destinations; applying the same change to the
`Product` documentation as well as `Sum`.
Mention the `Borrow` guarantee on the `Hash` implementations for Arrays and `Vec`
To remind people like me who forget about it and send PRs to make them different, and to (probably) get a test failure if the code is changed to no longer uphold it.
Updates `Clone` docs for `Copy` comparison.
Quite a few people (myself included) have come under the impression that the difference between `Copy` and `Clone` is that `Copy` is cheap and `Clone` is expensive, where the actual difference is that `Copy` constrains the type to bit-wise copying, and `Clone` allows for more expensive operations. The source of this misconception is in the `Clone` docs, where the following line is in the description:
> Differs from `Copy` in that `Copy` is implicit and extremely inexpensive, while `Clone` is always explicit and may or may not be expensive.
The `Clone` documentation page also comes up before the `Copy` page on google when searching for "the difference between `Clone` and `Copy`".
This PR updates the documentation to clarify that "extremely inexpensive" means an "inexpensive bit-wise copy" to hopefully prevent future rust users from falling into this misunderstanding.
Integrate binary search codes of binary_search_by and partition_point
For now partition_point has own binary search code piece.
It is because binary_search_by had called the comparer more times and the author (=me) wanted to avoid it.
However, now binary_search_by uses the comparer minimum times. (#74024)
So it's time to integrate them.
The appearance of the codes are a bit different but both use completely same logic.
Stabilize {std, core}::prelude::rust_*.
This stabilizes the `{core, std}::prelude::{rust_2015, rust_2018, rust_2021}` modules.
The usage of these modules as the prelude in those editions was already stabilized. This just stabilizes the modules themselves, making it possible for a user to explicitly refer to them.
Tracking issue: https://github.com/rust-lang/rust/issues/85684
FCP on the RFC that included this finished here: https://github.com/rust-lang/rfcs/pull/3114#issuecomment-840577395
Add functions `Duration::try_from_secs_{f32, f64}`
These functions allow constructing a Duration from a floating point value that could be out of range without panicking.
Tracking issue: #83400
Revert #85176 addition of `clone_from` for `ManuallyDrop`
Forwarding `clone_from` to the inner value changes the observable behavior, as previously the inner value would *not* be dropped by the default implementation.
Frankly, this is a super-niche case, so #85176 is welcome to argue the behavior should be otherwise! But if we overrride it, IMO documenting the behavior would be good.
Example: https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=c5d0856686fa850c1d7ee16891014efb
Make minor wording changes in a few places. Move `filter` to the
"transformations" section. Add `zip` methods to the "transformations"
section. Clarify the section about `Option` iterators, and add a section
about collecting into `Option`.
Clarify that for `Result`, `or` and `or_else` can also produce a
`Result` having a different type.
Implement nonzero arithmetics for NonZero types.
Hello'all, this is my first PR to this repo.
Non-zero natural numbers are stable by addition/multiplication/exponentiation, so it makes sense to make this arithmetic possible with `NonZeroU*`.
The major pitfall is that overflowing underlying `u*` types possibly lead to underlying `0` values, which break the major invariant of `NonZeroU*`. To accommodate it, only `checked_` and `saturating_` operations are implemented.
Other variants allowing wrapped results like `wrapping_` or `overflowing_` are ruled out *de facto*.
`impl Add<u*> for NonZeroU* { .. }` was considered, as it panics on overflow which enforces the invariant, but it does not so in release mode. I considered forcing `NonZeroU*::add` to panic in release mode by deferring the check to `u*::checked_add`, but this is less explicit for the user than directly using `NonZeroU*::checked_add`.
Following `@Lokathor's` advice on zulip, I have dropped the idea.
`@poliorcetics` on Discord also suggested implementing `_sub` operations, but I'd postpone this to another PR if there is a need for it. My opinion is that it could be useful in some cases, but that it makes less sense because non-null natural numbers are not stable by subtraction even in theory, while the overflowing problem is just about technical implementation.
One thing I don't like is that the type of the `other` arg differs in every implementation: `_add` methods accept any raw positive integer, `_mul` methods only accept non-zero values otherwise the invariant is also broken, and `_pow` only seems to accept `u32` for a reason I ignore but that seems consistent throughout `std`. Maybe there is a better way to harmonize this?
This is it, Iope I haven't forgotten anything and I'll be happy to read your feedback.
Fix an error in `map_or_else`. Use more descriptive text for
"don't care" in truth tables. Make minor corrections to truth tables.
Rename `makeiter` to `make_iter` in examples.
std: Stabilize wasm simd intrinsics
This commit performs two changes to stabilize Rust support for
WebAssembly simd intrinsics:
* The stdarch submodule is updated to pull in rust-lang/stdarch#1179.
* The `wasm_target_feature` feature gate requirement for the `simd128`
feature has been removed, stabilizing the name `simd128`.
This should conclude the FCP started on #74372 and...
Closes#74372
This commit performs two changes to stabilize Rust support for
WebAssembly simd intrinsics:
* The stdarch submodule is updated to pull in rust-lang/stdarch#1179.
* The `wasm_target_feature` feature gate requirement for the `simd128`
feature has been removed, stabilizing the name `simd128`.
This should conclude the FCP started on #74372 and...
Closes#74372
to_digit simplification (less jumps)
I just realised we might be able to make use of the fact that changing case in ascii is easy to help simplify to_digit some more.
It looks a bit cleaner and it looks like it's less jumps and there's less instructions in the generated assembly:
https://godbolt.org/z/84Erh5dhz
The benchmarks don't really tell me much. Maybe a slight improvement on the var radix.
Before:
```
test char::methods::bench_to_digit_radix_10 ... bench: 53,819 ns/iter (+/- 8,314)
test char::methods::bench_to_digit_radix_16 ... bench: 57,265 ns/iter (+/- 10,730)
test char::methods::bench_to_digit_radix_2 ... bench: 55,077 ns/iter (+/- 5,431)
test char::methods::bench_to_digit_radix_36 ... bench: 56,549 ns/iter (+/- 3,248)
test char::methods::bench_to_digit_radix_var ... bench: 43,848 ns/iter (+/- 3,189)
test char::methods::bench_to_digit_radix_10 ... bench: 51,707 ns/iter (+/- 10,946)
test char::methods::bench_to_digit_radix_16 ... bench: 52,835 ns/iter (+/- 2,689)
test char::methods::bench_to_digit_radix_2 ... bench: 51,012 ns/iter (+/- 2,746)
test char::methods::bench_to_digit_radix_36 ... bench: 53,210 ns/iter (+/- 8,645)
test char::methods::bench_to_digit_radix_var ... bench: 40,386 ns/iter (+/- 4,711)
test char::methods::bench_to_digit_radix_10 ... bench: 54,088 ns/iter (+/- 5,677)
test char::methods::bench_to_digit_radix_16 ... bench: 55,972 ns/iter (+/- 17,229)
test char::methods::bench_to_digit_radix_2 ... bench: 52,083 ns/iter (+/- 2,425)
test char::methods::bench_to_digit_radix_36 ... bench: 54,132 ns/iter (+/- 1,548)
test char::methods::bench_to_digit_radix_var ... bench: 41,250 ns/iter (+/- 5,299)
```
After:
```
test char::methods::bench_to_digit_radix_10 ... bench: 48,907 ns/iter (+/- 19,449)
test char::methods::bench_to_digit_radix_16 ... bench: 52,673 ns/iter (+/- 8,122)
test char::methods::bench_to_digit_radix_2 ... bench: 48,509 ns/iter (+/- 2,885)
test char::methods::bench_to_digit_radix_36 ... bench: 50,526 ns/iter (+/- 4,610)
test char::methods::bench_to_digit_radix_var ... bench: 38,618 ns/iter (+/- 3,180)
test char::methods::bench_to_digit_radix_10 ... bench: 54,202 ns/iter (+/- 6,994)
test char::methods::bench_to_digit_radix_16 ... bench: 56,585 ns/iter (+/- 8,448)
test char::methods::bench_to_digit_radix_2 ... bench: 50,548 ns/iter (+/- 1,674)
test char::methods::bench_to_digit_radix_36 ... bench: 52,749 ns/iter (+/- 2,576)
test char::methods::bench_to_digit_radix_var ... bench: 40,215 ns/iter (+/- 3,327)
test char::methods::bench_to_digit_radix_10 ... bench: 50,233 ns/iter (+/- 22,272)
test char::methods::bench_to_digit_radix_16 ... bench: 50,841 ns/iter (+/- 19,981)
test char::methods::bench_to_digit_radix_2 ... bench: 50,386 ns/iter (+/- 4,555)
test char::methods::bench_to_digit_radix_36 ... bench: 52,369 ns/iter (+/- 2,737)
test char::methods::bench_to_digit_radix_var ... bench: 40,417 ns/iter (+/- 2,766)
```
I removed the likely as it resulted in a few less instructions. (It's not been in there long - I added it in the last to_digit iteration).
fix off by one in `std::iter::Iterator` documentation
the range `(0..10)` is documented as "The even numbers from zero to ten." - should be ".. to nine".
Make copy/copy_nonoverlapping fn's again
Make copy/copy_nonoverlapping fn's again, rather than intrinsics.
This a short-term change to address issue #84297.
It effectively reverts PRs #81167#81238 (and part of #82967), #83091, and parts of #79684.
Revert "implement TrustedRandomAccess for Take iterator adapter"
This reverts commit 37a5b515e9 (#83990).
The original change unintentionally caused side-effects from certain iterator chains combining `take`, `zip` and `next_back()` to be omitted which is observable by user code and thus likely a breaking change
Technically one could declare it not a breaking change since `Zip`'s API contract is silent about about its backwards iteration behavior but on the other hand there is nothing in the stable Iterator API that could justify the currently observable behavior. And either way, this impact wasn't noticed or discussed in the original PR.
Fixes#85969
To remind people like me who forget about it and send PRs to make them different, and to (probably) get a test failure if the code is changed to no longer uphold it.
Clarify documentation of slice sorting methods
After reading about [this](https://polkadot.network/a-polkadot-postmortem-24-05-2021/), I realized that although the documentation of these methods is not ambiguous in its current state, it is very easy to read it and erroneously assume that their exact behaviour can be relied upon to be deterministic. Although the docs make no guarantees about which index is returned when there are multiple matches, being more explicit about when and how their determinism can be relied upon should help prevent people from making this mistake in the future.
r? ``@steveklabnik``
Update standard library for IntoIterator implementation of arrays
This PR partially resolves issue #84513 of updating the standard library part.
I haven't found any remaining doctest examples which are using iterators over e.g. &i32 instead of just i32 in the standard library. Can anyone point me to them if there's remaining any?
Thanks!
r? ```@m-ou-se```
Add a map method to Bound
Add a map method to std::ops::range::Bound, patterned off of the method
of the same name on Option.
Have left off creating a tracking issue initially, but as soon as I get the go-ahead from a reviewer I'll make that right away 😄
Remove `doc(include)`
This nightly feature is redundant now that `extended_key_value_attributes` is stable (https://github.com/rust-lang/rust/pull/83366). `@rust-lang/rustdoc` not sure if you think this needs FCP; there was already an FCP in #82539, but technically it was for deprecating, not removing the feature altogether.
This should not be merged before #83366.
cc `@petrochenkov`
## User-facing changes
- Intra-doc links to primitives that currently go to rust-lang.org/nightly/std/primitive.x.html will start going to channel that rustdoc was built with. Nightly will continue going to /nightly; Beta will link to /beta; stable compilers will link to /1.52.1 (or whatever version they were built as).
- Cross-crate links from std to core currently go to /nightly unconditionally. They will start going to /1.52.0 on stable channels (but remain the same on nightly channels).
- Intra-crate links from std to std (or core to core) currently go to the same URL they are hosted at; they will continue to do so. Notably, this is different from everything else because it can preserve the distinction between /stable and /1.52.0 by using relative links.
Note that "links" includes both intra-doc links and rustdoc's own
automatically generated hyperlinks.
## Implementation changes
- Update the testsuite to allow linking to /beta and /1.52.1 in docs
- Use an html_root_url for the standard library that's dependent on the channel
This avoids linking to nightly docs on stable.
- Update rustdoc to use channel-dependent links for primitives from an
unknown crate
- Set DOC_RUST_LANG_ORG_CHANNEL from bootstrap to ensure it's in sync
- Include doc.rust-lang.org in the channel
Mention "null pointer optimization" in option docs.
I had seen people discuss "null pointer optimization," but when I tried to find official documentation (using Google), the `std::option` page didn't show up, because it doesn't use that term. Hopefully adding the term will help others find it in the future.
Revert "Auto merge of #83770 - the8472:tra-extend, r=Mark-Simulacrum"
Due to a performance regression that didn't show up in the original perf run
this reverts commit 9111b8ae97 (#83770), reversing
changes made to 9a700d2947.
Since since is expected to have the inverse impact it should probably be rollup=never.
r? `@Mark-Simulacrum`
Make `Step` trait safe to implement
This PR makes a few modifications to the `Step` trait that I believe better position it for stabilization in the short term. In particular,
1. `unsafe trait TrustedStep` is introduced, indicating that the implementation of `Step` for a given type upholds all stated invariants (which have remained unchanged). This is gated behind a new `trusted_step` feature, as stabilization is realistically blocked on min_specialization.
2. The `Step` trait is internally specialized on the `TrustedStep` trait, which avoids a serious performance regression.
3. `TrustedLen` is implemented for `T: TrustedStep` as the latter's invariants subsume the former's.
4. The `Step` trait is no longer `unsafe`, as the invariants must not be relied upon by unsafe code (unless the type implements `TrustedStep`).
5. `TrustedStep` is implemented for all types that implement `Step` in the standard library and compiler.
6. The `step_trait_ext` feature is merged into the `step_trait` feature. I was unable to find any reasoning for the features being split; the `_unchecked` methods need not necessarily be stabilized at the same time, but I think it is useful to have them under the same feature flag.
All existing implementations of `Step` will be broken, as it is not possible to `unsafe impl` a safe trait. Given this trait only exists on nightly, I feel this breakage is acceptable. The blanket `impl<T: Step> TrustedLen for T` will likely cause some minor breakage, but this should be covered by the equivalent impl for `TrustedStep`.
Hopefully these changes are sufficient to place `Step` in decent position for stabilization, which would allow user-defined types to be used with `a..b` syntax.
Mention workaround for floats in Iterator::{min, max}
`Iterator::{min, max}` can't be used with iterators of floats due to NaN issues. This suggests a workaround in the documentation of those functions.
Forwarding `clone_from` to the inner value changes the observable
behavior, as previously the inner value would *not* be dropped by the
default implementation.
Remove Iterator #[rustc_on_unimplemented]s that no longer apply.
Now that `IntoIterator` is implemented for arrays, all the `rustc_on_unimplemented` for arrays of ranges (e.g. `for _ in [1..3] {}`) no longer apply, since they are now valid Rust.
Separated these from #85670, because we should discuss a potential new (clippy?) lint for these.
Until Rust 1.52, `for _ in [1..3] {}` produced:
```
error[E0277]: `[std::ops::Range<{integer}>; 1]` is not an iterator
--> src/main.rs:2:14
|
2 | for _ in [1..3] {}
| ^^^^^^ if you meant to iterate between two values, remove the square brackets
|
= help: the trait `std::iter::Iterator` is not implemented for `[std::ops::Range<{integer}>; 1]`
= note: `[start..end]` is an array of one `Range`; you might have meant to have a `Range` without the brackets: `start..end`
= note: required by `std::iter::IntoIterator::into_iter`
```
But in Rust 1.53 and later, it compiles fine. It iterates over the array by value, for one iteration with the element `1..3`.
This is probably a mistake, which is no longer caught. Should we have a lint for it? Should Clippy have a lint for it?
cc ```@estebank``` ```@flip1995```
cc https://github.com/rust-lang/rust/issues/84513
While stdlib implementations of the unchecked methods require unchecked
math, there is no reason to gate it behind this for external users. The
reasoning for a separate `step_trait_ext` feature is unclear, and as
such has been merged as well.
Add `TrustedRandomAccess` specialization for `Vec::extend()`
This should do roughly the same as the `TrustedLen` specialization but result in less IR by using `__iterator_get_unchecked`
instead of `Iterator::for_each`
Conflicting specializations are manually prioritized by grouping them under yet another helper trait.
Fix typo in core::array::IntoIter comment
Saw a small typo reading some internal comments and decided to just throw this up to fix it for future readers.
Remove num_as_ne_bytes feature
From the discussion in #76976, it is determined that eventual results of the safe transmute work as a more general mechanism will let these conversions happen in safe code without needing specialized methods.
Merging this PR closes#76976 and resolves#64464. Several T-libs members have raised their opinion that it doesn't pull its weight as a standalone method, and so we should not track it as a specific thing to add.
fix `matches!` and `assert_matches!` on edition 2021
Previously this code failed to compile on edition 2021. [(Playground)](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=53960f2f051f641777b9e458da747707)
```rust
fn main() {
matches!((), ());
}
```
```
Compiling playground v0.0.1 (/playground)
error: `$pattern:pat` may be followed by `|`, which is not allowed for `pat` fragments
|
= note: allowed there are: `=>`, `,`, `=`, `if` or `in`
error: aborting due to previous error
error: could not compile `playground`
To learn more, run the command again with --verbose.
```
Demote `ControlFlow::{from|into}_try` to `pub(crate)`
They have mediocre names and non-obvious semantics, so personally I don't think they're worth trying to stabilize, and thus might as well just be internal (they're used for convenience in iterator adapters), not something shown in the rustdocs.
I don't think anyone actually wanted to use them outside `core` -- they just got made public-but-unstable along with the whole type in https://github.com/rust-lang/rust/pull/76204 that promoted `LoopState` from an internal type to the exposed `ControlFlow` type.
cc https://github.com/rust-lang/rust/issues/75744, the tracking issue they mention.
cc https://github.com/rust-lang/rust/pull/85608, the PR where I'm proposing stabilizing the type.
Fix pointer provenance in <[T]>::copy_within
Previously the `self.as_mut_ptr()` invalidated the pointer created by the first `self.as_ptr()`. This also triggered miri when run with `-Zmiri-track-raw-pointers`
