Provided that the iterator has finite length and does not trigger user-provided code, this is safe.
What follows is a full list of the DoubleEndedIterators in liballoc/libcore and whether this optimization is safe, and if not, why not.
src/liballoc/boxed.rs
Box: Pass through to avoid defeating optimization of the underlying DoubleIterator implementation. This has no correctness impact.
src/liballoc/collections/binary_heap.rs
Iter: Pass through to avoid defeating optimizations on slice::Iter
IntoIter: Not safe, changes Drop order
Drain: Not safe, changes Drop order
src/liballoc/collections/btree/map.rs
Iter: Safe to call next_back, invokes no user defined code.
IterMut: ditto
IntoIter: Not safe, changes Drop order
Keys: Safe to call next_back, invokes no user defined code.
Values: ditto
ValuesMut: ditto
Range: ditto
RangeMut: ditto
src/liballoc/collections/btree/set.rs
Iter: Safe to call next_back, invokes no user defined code.
IntoIter: Not safe, changes Drop order
Range: Safe to call next_back, invokes no user defined code.
src/liballoc/collections/linked_list.rs
Iter: Safe to call next_back, invokes no user defined code.
IterMut: ditto
IntoIter: Not safe, changes Drop order
src/liballoc/collections/vec_deque.rs
Iter: Safe to call next_back, invokes no user defined code.
IterMut: ditto
IntoIter: Not safe, changes Drop order
Drain: ditto
src/liballoc/string.rs
Drain: Safe because return type is a primitive (char)
src/liballoc/vec.rs
IntoIter: Not safe, changes Drop order
Drain: ditto
Splice: ditto
src/libcore/ascii.rs
EscapeDefault: Safe because return type is a primitive (u8)
src/libcore/iter/adapters/chain.rs
Chain: Not safe, invokes user defined code (Iterator impl)
src/libcore/iter/adapters/flatten.rs
FlatMap: Not safe, invokes user defined code (Iterator impl)
Flatten: ditto
FlattenCompat: ditto
src/libcore/iter/adapters/mod.rs
Rev: Not safe, invokes user defined code (Iterator impl)
Copied: ditto
Cloned: Not safe, invokes user defined code (Iterator impl and T::clone)
Map: Not safe, invokes user defined code (Iterator impl + closure)
Filter: ditto
FilterMap: ditto
Enumerate: Not safe, invokes user defined code (Iterator impl)
Skip: ditto
Fuse: ditto
Inspect: ditto
src/libcore/iter/adapters/zip.rs
Zip: Not safe, invokes user defined code (Iterator impl)
src/libcore/iter/range.rs
ops::Range: Not safe, changes Drop order, but ALREADY HAS SPECIALIZATION
ops::RangeInclusive: ditto
src/libcore/iter/sources.rs
Repeat: Not safe, calling last should iloop.
Empty: No point, iterator is at most one item long.
Once: ditto
OnceWith: ditto
src/libcore/option.rs
Item: No point, iterator is at most one item long.
Iter: ditto
IterMut: ditto
IntoIter: ditto
src/libcore/result.rs
Iter: No point, iterator is at most one item long
IterMut: ditto
IntoIter: ditto
src/libcore/slice/mod.rs
Split: Not safe, invokes user defined closure
SplitMut: ditto
RSplit: ditto
RSplitMut: ditto
Windows: Safe, already has specialization
Chunks: ditto
ChunksMut: ditto
ChunksExact: ditto
ChunksExactMut: ditto
RChunks: ditto
RChunksMut: ditto
RChunksExact: ditto
RChunksExactMut: ditto
src/libcore/str/mod.rs
Chars: Safe, already has specialization
CharIndices: ditto
Bytes: ditto
Lines: Safe to call next_back, invokes no user defined code.
LinesAny: Deprecated
Everything that is generic over P: Pattern: Not safe because Pattern invokes user defined code.
SplitWhitespace: Safe to call next_back, invokes no user defined code.
SplitAsciiWhitespace: ditto
It's natural for `retain` to work in order from beginning to end, but
this wasn't actually documented to be the case. If we actually promise
this, then the caller can do useful things like track the index of each
element being tested, as [discussed in the forum][1]. This is now
documented for `Vec`, `VecDeque`, and `String`.
[1]: https://users.rust-lang.org/t/vec-retain-by-index/27697
`HashMap` and `HashSet` also have `retain`, and the `hashbrown`
implementation does happen to use a plain `iter()` order too, but it's
not certain that this should always be the case for these types.
This updates the `Extend` implementations to use `for_each` for many
collections: `BinaryHeap`, `BTreeMap`, `BTreeSet`, `LinkedList`, `Path`,
`TokenStream`, `VecDeque`, and `Wtf8Buf`.
Folding with `for_each` enables better performance than a `for`-loop for
some iterators, especially if they can just forward to internal
iterators, like `Chain` and `FlatMap` do.
override `VecDeque::try_rfold`, also update iterator
This keeps the slice based iteration and updates the iterator state after each slice. It also uses a loop to reduce the amount of code.
This uses unsafe code, so some thorough review would be appreciated. Cc @RalfJung
This keeps the slice based iteration and updates the iterator
state after each slice. It also uses a loop to reduce the amount
of code.
This uses unsafe code, so some thorough review would be
appreciated.
There are two big categories of changes in here
- Removing lifetimes from common traits that can essentially never user a lifetime from an input (particularly `Drop` & `Debug`)
- Forwarding impls that are only possible because the lifetime doesn't matter (like `impl<R: Read + ?Sized> Read for &mut R`)
I omitted things that seemed like they could be more controversial, like the handful of iterators that have a `Item: 'static` despite the iterator having a lifetime or the `PartialEq` implementations where the flipped one cannot elide the lifetime.
VecDeque: fix for stacked borrows
`VecDeque` violates a version of stacked borrows where creating a shared reference is not enough to make a location *mutably accessible* from raw pointers (and I think that is the version we want). There are two problems:
* Creating a `NonNull<T>` from `&mut T` goes through `&T` (inferred for a `_`), then `*const T`, then `NonNull<T>`. That means in this stricter version of Stacked Borrows, we cannot actually write to such a `NonNull` because it was created from a shared reference! This PR fixes that by going from `&mut T` to `*mut T` to `*const T`.
* `VecDeque::drain` creates the `Drain` struct by *first* creating a `NonNull` from `self` (which is an `&mut VecDeque`), and *then* calling `self.buffer_as_mut_slice()`. The latter reborrows `self`, asserting that `self` is currently the unique pointer to access this `VecDeque`, and hence invalidating the `NonNull` that was created earlier. This PR fixes that by instead using `self.buffer_as_slice()`, which only performs read accesses and creates only shared references, meaning the raw pointer (`NonNull`) remains valid.
It is possible that other methods on `VecDeque` do something similar, miri's test coverage of `VecDeque` is sparse to say the least.
Cc @nikomatsakis @Gankro
