This commit is an implementation of [RFC 576][rfc] which adds back the `std::io`
module to the standard library. No functionality in `std::old_io` has been
deprecated just yet, and the new `std::io` module is behind the same `io`
feature gate.
[rfc]: https://github.com/rust-lang/rfcs/pull/576
A good bit of functionality was copied over from `std::old_io`, but many tweaks
were required for the new method signatures. Behavior such as precisely when
buffered objects call to the underlying object may have been tweaked slightly in
the transition. All implementations were audited to use composition wherever
possible. For example the custom `pos` and `cap` cursors in `BufReader` were
removed in favor of just using `Cursor<Vec<u8>>`.
A few liberties were taken during this implementation which were not explicitly
spelled out in the RFC:
* The old `LineBufferedWriter` is now named `LineWriter`
* The internal representation of `Error` now favors OS error codes (a
0-allocation path) and contains a `Box` for extra semantic data.
* The io prelude currently reexports `Seek` as `NewSeek` to prevent conflicts
with the real prelude reexport of `old_io::Seek`
* The `chars` method was moved from `BufReadExt` to `ReadExt`.
* The `chars` iterator returns a custom error with a variant that explains that
the data was not valid UTF-8.
This removes the `ByRef` iterator adaptor to stay in line with the changes to
`std::io`. The `by_ref` method instead just returns `&mut Self`.
This also removes the implementation of `Iterator for &mut Iterator` and instead
generalizes it to `Iterator for &mut I` where `I: Iterator + ?Sized`. The
`Box<I>` implementations were also updated.
This is a breaking change due to the removal of the `std::iter::ByRef` type. All
mentions of `ByRef<'a, T>` should be replaced with `&mut T` to migrate forward.
[breaking-change]
upgrade the inference based on expected type so that it is able to
infer the fn kind in isolation even if the full signature is not
available (and we could perhaps do better still in some cases, such as
extracting just the types of the arguments but not the return value).
closure kind, thereby detecting what happens if there are
mismatches. Simply removing the `:` annotations caused most of these
tests to pass or produce other errors, because the inference would
convert the closure into a more appropriate kind. (The ability to
override the inference by using the expected type is an important
backdoor partly for this reason.)
possible. There is some amount of duplication as a result (similar to
select) -- I am not happy about this but not sure how to fix it
without deeper rewrites.
The extra check caused by the expect() call can, in general, not be
optimized away, because the length of the iterator is unknown at compile
time, causing a noticable slow-down. Since the check only triggers if
the element isn't actually found in the iterator, i.e. it isn't
guaranteed to trigger for ill-behaved ExactSizeIterators, it seems
reasonable to switch to an implementation that doesn't need the check
and just always returns None if the value isn't found.
Benchmark:
````rust
let v: Vec<u8> = (0..1024*65).map(|_| 0).collect();
b.iter(|| {
v.as_slice().iter().rposition(|&c| c == 1)
});
````
Before:
````
test rposition ... bench: 49939 ns/iter (+/- 23)
````
After:
````
test rposition ... bench: 33306 ns/iter (+/- 68)
````
That is, when offering suggestions for unresolved method calls, avoid
suggesting traits for which implementing the trait for the receiver type
either makes little sense (e.g. type errors, or sugared unboxed
closures), or violates coherence.
The latter is approximated by ensuring that at least one of `{receiver
type, trait}` is local. This isn't precisely correct due to
multidispatch, but the error messages one encounters in such situation
are useless more often than not; it is better to be conservative and
miss some cases, than have overly many false positives (e.g. writing
`some_slice.map(|x| ...)` uselessly suggested that one should implement
`IteratorExt` for `&[T]`, while the correct fix is to call `.iter()`).
Closes#21420.
That is, when offering suggestions for unresolved method calls, avoid
suggesting traits for which implementing the trait for the receiver type
either makes little sense (e.g. type errors, or sugared unboxed
closures), or violates coherence.
The latter is approximated by ensuring that at least one of `{receiver
type, trait}` is local. This isn't precisely correct due to
multidispatch, but the error messages one encounters in such situation
are useless more often than not; it is better to be conservative and
miss some cases, than have overly many false positives (e.g. writing
`some_slice.map(|x| ...)` uselessly suggested that one should implement
`IteratorExt` for `&[T]`, while the correct fix is to call `.iter()`).
Closes#21420.
I’d kind of like to be able to use HashState in AnyMap, which I can’t do without a stability attribute on it. While I was at it I looked around and found a few more missing.
The existence of these two functions is at odds with our current [error
conventions][conventions] which recommend that panicking and `Result`-like
variants should not be provided together.
[conventions]: https://github.com/rust-lang/rfcs/blob/master/text/0236-error-conventions.md#do-not-provide-both-result-and-fail-variants
This commit adds a new `borrow_state` function returning a `BorrowState` enum to
`RefCell` which serves as a replacemnt for the `try_borrow` and `try_borrow_mut`
functions.
Now that associated types are fully implemented the iterator adaptors only need
type parameters which are associated with actual storage. All other type
parameters can either be derived from these (e.g. they are an associated type)
or can be bare on the `impl` block itself.
This is a breaking change due to the removal of type parameters on these
iterator adaptors, but code can fairly easily migrate by just deleting the
relevant type parameters for each adaptor. Other behavior should not be
affected.
Closes#21839
[breaking-change]
Unicode escapes were changed in [this RFC](28aeb3c391/text/0446-es6-unicode-escapes.md) to use the ES6 \u{00FFFF} syntax with a variable number of digits from 1-6, eliminating the need for two different syntaxes for unicode literals.
I have updated The Reference and grammar.md to reflect these changes.
For "symmetric" binary operators, meaning the types of two sides must be
equal, if the type of LHS doesn't know yet but RHS does, use that as an
hint to infer LHS' type.
Closes#21634
