This is an initial pass at stabilizing the `iter` module. The module is
fairly large, but is also pretty polished, so most of the stabilization
leaves things as they are.
Some changes:
* Due to the new object safety rules, various traits needs to be split
into object-safe traits and extension traits. This includes `Iterator`
itself. While splitting up the traits adds some complexity, it will
also increase flexbility: once we have automatic impls of `Trait` for
trait objects over `Trait`, then things like the iterator adapters
will all work with trait objects.
* Iterator adapters that use up the entire iterator now take it by
value, which makes the semantics more clear and helps catch bugs. Due
to the splitting of Iterator, this does not affect trait objects. If
the underlying iterator is still desired for some reason, `by_ref` can
be used. (Note: this change had no fallout in the Rust distro except
for the useless mut lint.)
* In general, extension traits new and old are following an [in-progress
convention](https://github.com/rust-lang/rfcs/pull/445). As such, they
are marked `unstable`.
* As usual, anything involving closures is `unstable` pending unboxed
closures.
* A few of the more esoteric/underdeveloped iterator forms (like
`RandomAccessIterator` and `MutableDoubleEndedIterator`, along with
various unfolds) are left experimental for now.
* The `order` submodule is left `experimental` because it will hopefully
be replaced by generalized comparison traits.
* "Leaf" iterators (like `Repeat` and `Counter`) are uniformly
constructed by free fns at the module level. That's because the types
are not otherwise of any significance (if we had `impl Trait`, you
wouldn't want to define a type at all).
Closes#17701
Due to renamings and splitting of traits, this is a:
[breaking-change]
It looks like currently kinds required by traits are not propagated when they are wrapped in a TyTrait. Additionally, in SelectionContext::builtin_bound, no attempt is made to check whether the target trait or its supertraits require the kind specified.
This PR alters SelectionContext::builtin_bound to examine all supertraits in the target trait's bounds recursively for required kinds.
Alternatively, the kinds could be added to the TyTrait upon creation (by just setting its builtin_bounds to the union of the bounds requested in this instance and the bounds required by the trait), this option may have less overhead during compilation but information is lost about which kinds were explicitly requested for this instance (vs those specified by traits/supertraits) would be lost.
This patch merges the `libsync` crate into `libstd`, undoing part of the
facade. This is in preparation for ultimately merging `librustrt`, as
well as the upcoming rewrite of `sync`.
Because this removes the `libsync` crate, it is a:
[breaking-change]
However, all uses of `libsync` should be able to reroute through
`std::sync` and `std::comm` instead.
r? @alexcrichton
Code to fragment paths into pieces based on subparts being moved around, e.g. moving `x.1` out of a tuple `(A,B,C)` leaves behind the fragments `x.0: A` and `x.2: C`. Further discussion in borrowck/doc.rs.
Includes differentiation between assigned_fragments and moved_fragments, support for all-but-one array fragments, and instrumentation to print out the moved/assigned/unmmoved/parents for each function, factored out into a separate submodule.
These fragments can then be used by `trans` to inject stack-local dynamic drop flags. (They also can be hooked up with dataflow to reduce the expected number of injected flags.)
The tests use new "//~| ERROR" follow syntax.
Includes a test for moves involving array elements. It was easier
than i realized to get something naive off the ground here.
Includes differentiation between assigned_fragments and
moved_fragments, support for all-but-one array fragments, and
instrumentation to print out the moved/assigned/unmmoved/parents for
each function, factored out into separate submodule.
This is accomplished by:
1. Add `MatchMode` enum to `expr_use_visitor`.
2. Computing the match mode for each pattern via a pre-pass, and then
passing the mode along when visiting the pattern in
expr_use_visitor.
3. Adding a `fn matched_pat` callback to expr_use_visitor, which is
called on interior struct and enum nodes of the pattern (as opposed
to `fn consume_pat`, which is only invoked for identifiers at the
leaves of the pattern), and invoking it accordingly.
Of particular interest are the `cat_downcast` instances established
when matching enum variants.
This is to fix a problem where I could not reliably map attach the
type for each loan-path to the loan-path itself because the same
loan-path was ending up associated with two different types, because
the cmt's had diverged in their interpretation of the path.
To make this clean, refactored old `LoanPath` enum into a
`LoanPath` struct with a `ty::t` and a newly-added `LoanPathVariant` enum.
This enabled me to get rid of the ugly and fragile `LoanPath::to_type`
method, and I can probably also get rid of other stuff that was
supporting it, maybe.
`LpDowncast` carries the `DefId` of the variant itself. To support
this, added the enum variant `DefId` to the `cat_downcast` variant in
`mem_categorization::categorization`.
(updated to fix mem_categorization to handle downcast of enum
struct-variants properly.)
This change applies the conventions to unwrap listed in [RFC 430][rfc] to rename
non-failing `unwrap` methods to `into_inner`. This is a breaking change, but all
`unwrap` methods are retained as `#[deprecated]` for the near future. To update
code rename `unwrap` method calls to `into_inner`.
[rfc]: https://github.com/rust-lang/rfcs/pull/430
[breaking-change]
cc #19091
A slice iterator is isomorphic to a slice, just with a slightly
different form: storing start and end pointers rather than start pointer
and length. This patch reflects this by making converting between them
as easy as `iter.as_slice()` (or even `iter[]` if the shorter lifetime
is ok). That is, `slice.iter().as_slice() == slice`.
r? @aturon
A slice iterator is isomorphic to a slice, just with a slightly
different form: storing start and end pointers rather than start pointer
and length. This patch reflects this by making converting between them
as easy as `iter.as_slice()` (or even `iter[]` if the shorter lifetime
is ok). That is, `slice.iter().as_slice() == slice`.
It turns out that rustrt::at_exit() doesn't actually occur after all pthread
threads have exited (nor does atexit()), so there's not actually a known point
at which we can deallocate these keys. It's not super critical that we do so,
however, because we're about to exit anyway!
Closes#19280
At the same time remove the `pub use` of the variants in favor of accessing
through the enum type itself. This is a breaking change as the `Found` and
`NotFound` variants must now be imported through `BinarySearchResult` instead of
just `std::slice`.
[breaking-change]
Closes#19272
It turns out that rustrt::at_exit() doesn't actually occur after all pthread
threads have exited (nor does atexit()), so there's not actually a known point
at which we can deallocate these keys. It's not super critical that we do so,
however, because we're about to exit anyway!
Closes#19280
"_" should keep the default syntax class (symbol, not word). This
allows, e.g., `forward-word' to behave in the familiar way, jumping to
underscores within a function or variable name.
All of the enum components had a redundant 'Type' specifier: TypeSymlink, TypeDirectory, TypeFile. This change removes them, replacing them with a namespace: FileType::Symlink, FileType::Directory, and FileType::RegularFile.
RegularFile is used instead of just File, as File by itself could be mistakenly thought of as referring to the struct.
[breaking-change]
