The primary focus of Rust's stability story at 1.0 is the standard library.
All other libraries distributed with the Rust compiler are planned to
be #[unstable] and therfore only accessible on the nightly channel of Rust. One
of the more widely used libraries today is libserialize, Rust's current solution
for encoding and decoding types.
The current libserialize library, however, has a number of drawbacks:
* The API is not ready to be stabilize as-is and we will likely not have enough
resources to stabilize the API for 1.0.
* The library is not necessarily the speediest implementations with alternatives
being developed out-of-tree (e.g. serde from erickt).
* It is not clear how the API of Encodable/Decodable can evolve over time while
maintaining backwards compatibility.
One of the major pros to the current libserialize, however, is
`deriving(Encodable, Decodable)` as short-hands for enabling serializing and
deserializing a type. This is unambiguously useful functionality, so we cannot
simply deprecate the in-tree libserialize in favor of an external crates.io
implementation.
For these reasons, this commit starts off a stability story for libserialize by
following these steps:
1. The deriving(Encodable, Decodable) modes will be deprecated in favor of a
renamed deriving(RustcEncodable, RustcDecodable).
2. The in-tree libserialize will be deprecated in favor of an external
rustc-serialize crate shipped on crates.io. The contents of the crate will be
the same for now (but they can evolve separately).
3. At 1.0 serialization will be performed through
deriving(RustcEncodable, RustcDecodable) and the rustc-serialize crate. The
expansions for each deriving mode will change from `::serialize::foo` to
`::rustc_serialize::foo`.
This story will require that the compiler freezes its implementation of
`RustcEncodable` deriving for all of time, but this should be a fairly minimal
maintenance burden. Otherwise the crate in crates.io must always maintain the
exact definition of its traits, but the implementation of json, for example, can
continue to evolve in the semver-sense.
The major goal for this stabilization effort is to pave the road for a new
official serialization crate which can replace the current one, solving many of
its downsides in the process. We are not assuming that this will exist for 1.0,
hence the above measures. Some possibilities for replacing libserialize include:
* If plugins have a stable API, then any crate can provide a custom `deriving`
mode (will require some compiler work). This means that any new serialization
crate can provide its own `deriving` with its own backing
implementation, entirely obsoleting the current libserialize and fully
replacing it.
* Erick is exploring the possibility of code generation via preprocessing Rust
source files in the near term until plugins are stable. This strategy would
provide the same ergonomic benefit that `deriving` does today in theory.
So, in summary, the current libserialize crate is being deprecated in favor of
the crates.io-based rustc-serialize crate where the `deriving` modes are
appropriately renamed. This opens up space for a later implementation of
serialization in a more official capacity while allowing alternative
implementations to be explored in the meantime.
Concretely speaking, this change adds support for the `RustcEncodable` and
`RustcDecodable` deriving modes. After a snapshot is made warnings will be
turned on for usage of `Encodable` and `Decodable` as well as deprecating the
in-tree libserialize crate to encurage users to use rustc-serialize instead.
Using a type alias for iterator implementations is fragile since this
exposes the implementation to users of the iterator, and any changes
could break existing code.
This commit changes the iterators of `VecMap` to use
proper new types, rather than type aliases. However, since it is
fair-game to treat a type-alias as the aliased type, this is a:
[breaking-change].
This PR moves almost all our current uses of closures, both in public API and internal uses, to the new "unboxed" closures system.
In most cases, downstream code that *only uses* closures will continue to work as it is. The reason is that the `|| {}` syntax can be inferred either as a boxed or an "unboxed" closure according to the context. For example the following code will continue to work:
``` rust
some_option.map(|x| x.transform_with(upvar))
```
And will get silently upgraded to an "unboxed" closure.
In some other cases, it may be necessary to "annotate" which `Fn*` trait the closure implements:
```
// Change this
|x| { /* body */}
// to either of these
|: x| { /* body */} // closure implements the FnOnce trait
|&mut : x| { /* body */} // FnMut
|&: x| { /* body */} // Fn
```
This mainly occurs when the closure is assigned to a variable first, and then passed to a function/method.
``` rust
let closure = |: x| x.transform_with(upvar);
some.option.map(closure)
```
(It's very likely that in the future, an improved inference engine will make this annotation unnecessary)
Other cases that require annotation are closures that implement some trait via a blanket `impl`, for example:
- `std::finally::Finally`
- `regex::Replacer`
- `std::str::CharEq`
``` rust
string.trim_left_chars(|c: char| c.is_whitespace())
//~^ ERROR: the trait `Fn<(char,), bool>` is not implemented for the type `|char| -> bool`
string.trim_left_chars(|&: c: char| c.is_whitespace()) // OK
```
Finally, all implementations of traits that contain boxed closures in the arguments of their methods are now broken. And will need to be updated to use unboxed closures. These are the main affected traits:
- `serialize::Decoder`
- `serialize::DecoderHelpers`
- `serialize::Encoder`
- `serialize::EncoderHelpers`
- `rustrt::ToCStr`
For example, change this:
``` rust
// libserialize/json.rs
impl<'a> Encoder<io::IoError> for Encoder<'a> {
fn emit_enum(&mut self,
_name: &str,
f: |&mut Encoder<'a>| -> EncodeResult) -> EncodeResult {
f(self)
}
}
```
to:
``` rust
// libserialize/json.rs
impl<'a> Encoder<io::IoError> for Encoder<'a> {
fn emit_enum<F>(&mut self, _name: &str, f: F) -> EncodeResult where
F: FnOnce(&mut Encoder<'a>) -> EncodeResult
{
f(self)
}
}
```
[breaking-change]
---
### How the `Fn*` bound has been selected
I've chosen the bounds to make the functions/structs as "generic as possible", i.e. to let them allow the maximum amount of input.
- An `F: FnOnce` bound accepts the three kinds of closures: `|:|`, `|&mut:|` and `|&:|`.
- An `F: FnMut` bound only accepts "non-consuming" closures: `|&mut:|` and `|&:|`.
- An `F: Fn` bound only accept the "immutable environment" closures: `|&:|`.
This means that whenever possible the `FnOnce` bound has been used, if the `FnOnce` bound couldn't be used, then the `FnMut` was used. The `Fn` bound was never used in the whole repository.
The `FnMut` bound was the most used, because it resembles the semantics of the current boxed closures: the closure can modify its environment, and the closure may be called several times.
The `FnOnce` bound allows new semantics: you can move out the upvars when the closure is called. This can be effectively paired with the `move || {}` syntax to transfer ownership from the environment to the closure caller.
In the case of trait methods, is hard to select the "right" bound since we can't control how the trait may be implemented by downstream users. In these cases, I have selected the bound based on how we use these traits in the repository. For this reason the selected bounds may not be ideal, and may require tweaking before stabilization.
r? @aturon
