This is an alternative to upgrading the way rvalues are handled in the
borrow check. Making rvalues handled more like lvalues in the borrow
check caused numerous problems related to double mutable borrows and
rvalue scopes. Rather than come up with more borrow check rules to try
to solve these problems, I decided to just forbid pattern bindings after
`@`. This affected fewer than 10 lines of code in the compiler and
libraries.
This breaks code like:
match x {
y @ z => { ... }
}
match a {
b @ Some(c) => { ... }
}
Change this code to use nested `match` or `let` expressions. For
example:
match x {
y => {
let z = y;
...
}
}
match a {
Some(c) => {
let b = Some(c);
...
}
}
Closes#14587.
[breaking-change]
This ensures that the name of the crate is set from the command line for tests
so the auto-injection of `extern crate <name>` in doc tests works correctly.
When generating a unique symbol for things like closures or glue_drop,
we call token::gensym() to create a crate-unique Name. Recently, Name
changed its Show impl so it no longer prints as a number. This caused
symbols like glue_drop:1542 to become glue_drop:"glue_drop"(1542), or in
mangled form, glue_drop.$x22glue_drop$x22$LP$1542$RP$.
Add a few new free functions to alloc::rc for manipulating
uniquely-owned Rc values. is_unique() can be used to test if the Rc is
uniquely-owned, try_unwrap() can remove the value from a uniquely-owned
Rc, and get_mut() can return a &mut for a uniquely-owned Rc.
These are all free functions, because smart pointers should avoid having
methods when possible. They can't be static methods because UFCS will
remove that distinction. I think we should probably change downgrade()
and make_unique() into free functions as well, but that's out of scope.
Add a few new free functions to alloc::rc for manipulating
uniquely-owned Rc values. is_unique() can be used to test if the Rc is
uniquely-owned, try_unwrap() can remove the value from a uniquely-owned
Rc, and get_mut() can return a &mut for a uniquely-owned Rc.
These are all free functions, because smart pointers should avoid having
methods when possible. They can't be static methods because UFCS will
remove that distinction. I think we should probably change downgrade()
and make_unique() into free functions as well, but that's out of scope.
This ensures that the name of the crate is set from the command line for tests
so the auto-injection of `extern crate <name>` in doc tests works correctly.
A quick and dirty fix for #15036 until we get serious decoder reform.
Right now it is impossible for a `Decodable` to signal a decode error, for example if it has only finitely many allowed values, is a string which must be encoded a certain way, needs a valid checksum, etc. For example in the `libuuid` implementation of `Decodable` an `Option` is unwrapped, meaning that a decode of a malformed UUID will cause the task to fail.
A quick and dirty fix for #15036 until we get serious decoder reform.
Right now it is impossible for a Decodable to signal a decode error,
for example if it has only finitely many allowed values, is a string
which must be encoded a certain way, needs a valid checksum, etc. For
example in the libuuid implementation of Decodable an Option is
unwrapped, meaning that a decode of a malformed UUID will cause the
task to fail.
Since this adds a method to the `Decoder` trait, all users will need
to update their implementations to add it. The strategy used for the
current implementations for JSON and EBML is to add a new entry to
the error enum `ApplicationError(String)` which stores the string
provided to `.error()`.
[breaking-change]
When generating a unique symbol for things like closures or glue_drop,
we call token::gensym() to create a crate-unique Name. Recently, Name
changed its Show impl so it no longer prints as a number. This caused
symbols like glue_drop:1542 to become glue_drop:"glue_drop"(1542), or in
mangled form, glue_drop.$x22glue_drop$x22$LP$1542$RP$.
This was motivated by a desire to remove allocation in the common
pattern of
let old = key.replace(None)
do_something();
key.replace(old);
This also switched the map representation from a Vec to a TreeMap. A Vec
may be reasonable if there's only a couple TLD keys, but a TreeMap
provides better behavior as the number of keys increases.
Like the Vec, this TreeMap implementation does not shrink the container
when a value is removed. Unlike Vec, this TreeMap implementation cannot
reuse an empty node for a different key. Therefore any key that has been
inserted into the TLD at least once will continue to take up space in
the Map until the task ends. The expectation is that the majority of
keys that are inserted into TLD will be expected to have a value for
most of the rest of the task's lifetime. If this assumption is wrong,
there are two reasonable ways to fix this that could be implemented in
the future:
1. Provide an API call to either remove a specific key from the TLD and
destruct its node (e.g. `remove()`), or instead to explicitly clean
up all currently-empty nodes in the map (e.g. `compact()`). This is
simple, but requires the user to explicitly call it.
2. Keep track of the number of empty nodes in the map and when the map
is mutated (via `replace()`), if the number of empty nodes passes
some threshold, compact it automatically. Alternatively, whenever a
new key is inserted that hasn't been used before, compact the map at
that point.
---
Benchmarks:
I ran 3 benchmarks. tld_replace_none just replaces the tld key with None
repeatedly. tld_replace_some replaces it with Some repeatedly. And
tld_replace_none_some simulates the common behavior of replacing with
None, then replacing with the previous value again (which was a Some).
Old implementation:
test tld_replace_none ... bench: 20 ns/iter (+/- 0)
test tld_replace_none_some ... bench: 77 ns/iter (+/- 4)
test tld_replace_some ... bench: 57 ns/iter (+/- 2)
New implementation:
test tld_replace_none ... bench: 11 ns/iter (+/- 0)
test tld_replace_none_some ... bench: 23 ns/iter (+/- 0)
test tld_replace_some ... bench: 12 ns/iter (+/- 0)
Errors can be printed with {}, printing with {:?} does not work very
well.
Not actually related to this PR, but it came up when running the tests
and now is as good a time to fix it as any.
This was motivated by a desire to remove allocation in the common
pattern of
let old = key.replace(None)
do_something();
key.replace(old);
This also switched the map representation from a Vec to a TreeMap. A Vec
may be reasonable if there's only a couple TLD keys, but a TreeMap
provides better behavior as the number of keys increases.
Like the Vec, this TreeMap implementation does not shrink the container
when a value is removed. Unlike Vec, this TreeMap implementation cannot
reuse an empty node for a different key. Therefore any key that has been
inserted into the TLD at least once will continue to take up space in
the Map until the task ends. The expectation is that the majority of
keys that are inserted into TLD will be expected to have a value for
most of the rest of the task's lifetime. If this assumption is wrong,
there are two reasonable ways to fix this that could be implemented in
the future:
1. Provide an API call to either remove a specific key from the TLD and
destruct its node (e.g. `remove()`), or instead to explicitly clean
up all currently-empty nodes in the map (e.g. `compact()`). This is
simple, but requires the user to explicitly call it.
2. Keep track of the number of empty nodes in the map and when the map
is mutated (via `replace()`), if the number of empty nodes passes
some threshold, compact it automatically. Alternatively, whenever a
new key is inserted that hasn't been used before, compact the map at
that point.
---
Benchmarks:
I ran 3 benchmarks. tld_replace_none just replaces the tld key with None
repeatedly. tld_replace_some replaces it with Some repeatedly. And
tld_replace_none_some simulates the common behavior of replacing with
None, then replacing with the previous value again (which was a Some).
Old implementation:
test tld_replace_none ... bench: 20 ns/iter (+/- 0)
test tld_replace_none_some ... bench: 77 ns/iter (+/- 4)
test tld_replace_some ... bench: 57 ns/iter (+/- 2)
New implementation:
test tld_replace_none ... bench: 11 ns/iter (+/- 0)
test tld_replace_none_some ... bench: 23 ns/iter (+/- 0)
test tld_replace_some ... bench: 12 ns/iter (+/- 0)
Closes#16097 (fix variable name in tutorial)
Closes#16100 (More defailbloating)
Closes#16104 (Fix deprecation commment on `core::cmp::lexical_ordering`)
Closes#16105 (fix formatting in pointer guide table)
Closes#16107 (remove serialize::ebml, add librbml)
Closes#16108 (Fix heading levels in pointer guide)
Closes#16109 (rustrt: Don't conditionally init the at_exit QUEUE)
Closes#16111 (hexfloat: Deprecate to move out of the repo)
Closes#16113 (Add examples for GenericPath methods.)
Closes#16115 (Byte literals!)
Closes#16116 (Add a non-regression test for issue #8372)
Closes#16120 (Deprecate semver)
Closes#16124 (Deprecate uuid)
Closes#16126 (Deprecate fourcc)
Closes#16127 (Remove incorrect example)
Closes#16129 (Add note about production deployments.)
Closes#16131 (librustc: Don't ICE when trying to subst regions in destructor call.)
Closes#16133 (librustc: Don't ICE with struct exprs where the name is not a valid struct.)
Closes#16136 (Implement slice::Vector for Option<T> and CVec<T>)
Closes#16137 (alloc, arena, test, url, uuid: Elide lifetimes.)
Note: This PR is motivated by an attempt to write an custom syntax extension that tried to use `syntax::fold`, and that could only do so by fixing bugs in it and copying out private functions.
---
Refactored `syntax::fold`
Prior to this, the code there had a few issues:
- Default implementations inconsistenly either had the prefix `noop_` or
not.
- Some default methods where implemented in terms of a public noop function
for user code to call, others where implemented directly on the trait
and did not allow users of the trait to reuse the code.
- Some of the default implementations where private, and thus not reusable
for other implementors.
- There where some bugs where default implemntations called other default
implementations directly, rather than to the underlying Folder, with the
result of some ast nodes never being visted even if the user implemented that
method. (For example, the current Folder never folded struct fields)
This commit solves this situation somewhat radically by making __all__
`fold_...` functions in the module into Folder methods, and implementing
them all in terms of public `noop_...` functions for other implementors to
call out to.
Some public functions had to be renamed to fit the new system, so this is a
breaking change.
---
Also added a few trait implementations to `ast` types
Not included are two required patches:
* LLVM: segmented stack support for DragonFly [1]
* jemalloc: simple configure patches
[1]: http://reviews.llvm.org/D4705
Our implementation of ebml has diverged from the standard in order
to better serve the needs of the compiler, so it doesn't make much
sense to call what we have ebml anyore. Furthermore, our implementation
is pretty crufty, and should eventually be rewritten into a format
that better suits the needs of the compiler. This patch factors out
serialize::ebml into librbml, otherwise known as the Really Bad
Markup Language. This is a stopgap library that shouldn't be used
by end users, and will eventually be replaced by something better.
[breaking-change]
