Struct futures_locks::RwLock

pub struct RwLock<T: ?Sized> { /* fields omitted */ }

A Futures-aware RwLock.

std::sync::RwLock cannot be used in an asynchronous environment like Tokio, because an acquisition can block an entire reactor. This class can be used instead. It functions much like std::sync::RwLock. Unlike that class, it also has a builtin Arc, making it accessible from multiple threads. It's also safe to clone. Also unlike std::sync::RwLock, this class does not detect lock poisoning.

Methods

impl<T> RwLock<T>

pub fn new(t: T) -> RwLock<T>

Create a new RwLock in the unlocked state.

pub fn try_unwrap(self) -> Result<T, RwLock<T>>

Consumes the RwLock and returns the wrapped data. If the RwLock still has multiple references (not necessarily locked), returns a copy of self instead.

impl<T: ?Sized> RwLock<T>

pub fn get_mut(&mut self) -> Option<&mut T>

Returns a reference to the underlying data, if there are no other clones of the RwLock.

Since this call borrows the RwLock mutably, no actual locking takes place -- the mutable borrow statically guarantees no locks exist. However, if the RwLock has already been cloned, then None will be returned instead.

Examples

let mut lock = RwLock::<u32>::new(0);
*lock.get_mut().unwrap() += 5;
assert_eq!(lock.try_unwrap().unwrap(), 5);

pub fn read(&self) -> RwLockReadFut<T>

Acquire the RwLock nonexclusively, read-only, blocking the task in the meantime.

When the returned Future is ready, then this task will have read-only access to the protected data.

Examples

let rwlock = RwLock::<u32>::new(42);
let fut = rwlock.read().map(|mut guard| { *guard });
assert_eq!(spawn(fut).wait_future(), Ok(42));

pub fn write(&self) -> RwLockWriteFut<T>

Acquire the RwLock exclusively, read-write, blocking the task in the meantime.

When the returned Future is ready, then this task will have read-write access to the protected data.

Examples

let rwlock = RwLock::<u32>::new(42);
let fut = rwlock.write().map(|mut guard| { *guard = 5;});
spawn(fut).wait_future().expect("spawn");
assert_eq!(rwlock.try_unwrap().unwrap(), 5);

pub fn try_read(&self) -> Result<RwLockReadGuard<T>, ()>

Attempts to acquire the RwLock nonexclusively.

If the operation would block, returns Err instead. Otherwise, returns a guard (not a Future).

Examples

let mut lock = RwLock::<u32>::new(5);
let r = match lock.try_read() {
    Ok(guard) => *guard,
    Err(()) => panic!("Better luck next time!")
};
assert_eq!(5, r);

pub fn try_write(&self) -> Result<RwLockWriteGuard<T>, ()>

Attempts to acquire the RwLock exclusively.

If the operation would block, returns Err instead. Otherwise, returns a guard (not a Future).

Examples

let mut lock = RwLock::<u32>::new(5);
match lock.try_write() {
    Ok(mut guard) => *guard += 5,
    Err(()) => panic!("Better luck next time!")
}
assert_eq!(10, lock.try_unwrap().unwrap());

impl<T: 'static + ?Sized> RwLock<T>

pub fn with_read<F, B, R, E>(
    &self,
    f: F
) -> Result<impl Future<Item = R, Error = E>, SpawnError> where
    F: FnOnce(RwLockReadGuard<T>) -> B + Send + 'static,
    B: IntoFuture<Item = R, Error = E> + 'static,
    <B as IntoFuture>::Future: Send,
    R: Send + 'static,
    E: Send + 'static,
    T: Send, 

This is supported on feature="tokio" only.

Acquires a RwLock nonexclusively and performs a computation on its guarded value in a separate task. Returns a Future containing the result of the computation.

When using Tokio, this method will often hold the RwLock for less time than chaining a computation to read. The reason is that Tokio polls all tasks promptly upon notification. However, Tokio does not guarantee that it will poll all futures promptly when their owning task gets notified. So it's best to hold RwLocks within their own tasks, lest their continuations get blocked by slow stacked combinators.

Examples

let rwlock = RwLock::<u32>::new(5);
let mut rt = Runtime::new().unwrap();
let r = rt.block_on(lazy(|| {
    rwlock.with_read(|mut guard| {
        Ok(*guard) as Result<u32, ()>
    }).unwrap()
}));
assert_eq!(r, Ok(5));

pub fn with_read_local<F, B, R, E>(
    &self,
    f: F
) -> Result<impl Future<Item = R, Error = E>, SpawnError> where
    F: FnOnce(RwLockReadGuard<T>) -> B + 'static,
    B: IntoFuture<Item = R, Error = E> + 'static,
    R: 'static,
    E: 'static, 

This is supported on feature="tokio" only.

Like with_read but for Futures that aren't Send. Spawns a new task on a single-threaded Runtime to complete the Future.

Examples

// Note: Rc is not `Send`
let rwlock = RwLock::<Rc<u32>>::new(Rc::new(5));
let mut rt = current_thread::Runtime::new().unwrap();
let r = rt.block_on(lazy(|| {
    rwlock.with_read_local(|mut guard| {
        Ok(**guard) as Result<u32, ()>
    }).unwrap()
}));
assert_eq!(r, Ok(5));

pub fn with_write<F, B, R, E>(
    &self,
    f: F
) -> Result<impl Future<Item = R, Error = E>, SpawnError> where
    F: FnOnce(RwLockWriteGuard<T>) -> B + Send + 'static,
    B: IntoFuture<Item = R, Error = E> + Send + 'static,
    <B as IntoFuture>::Future: Send,
    R: Send + 'static,
    E: Send + 'static,
    T: Send, 

This is supported on feature="tokio" only.

Acquires a RwLock exclusively and performs a computation on its guarded value in a separate task. Returns a Future containing the result of the computation.

When using Tokio, this method will often hold the RwLock for less time than chaining a computation to write. The reason is that Tokio polls all tasks promptly upon notification. However, Tokio does not guarantee that it will poll all futures promptly when their owning task gets notified. So it's best to hold RwLocks within their own tasks, lest their continuations get blocked by slow stacked combinators.

Examples

let rwlock = RwLock::<u32>::new(0);
let mut rt = Runtime::new().unwrap();
let r = rt.block_on(lazy(|| {
    rwlock.with_write(|mut guard| {
        *guard += 5;
        Ok(()) as Result<(), ()>
    }).unwrap()
}));
assert!(r.is_ok());
assert_eq!(rwlock.try_unwrap().unwrap(), 5);

pub fn with_write_local<F, B, R, E>(
    &self,
    f: F
) -> Result<impl Future<Item = R, Error = E>, SpawnError> where
    F: FnOnce(RwLockWriteGuard<T>) -> B + 'static,
    B: IntoFuture<Item = R, Error = E> + 'static,
    R: 'static,
    E: 'static, 

This is supported on feature="tokio" only.

Like with_write but for Futures that aren't Send. Spawns a new task on a single-threaded Runtime to complete the Future.

Examples

// Note: Rc is not `Send`
let rwlock = RwLock::<Rc<u32>>::new(Rc::new(0));
let mut rt = current_thread::Runtime::new().unwrap();
let r = rt.block_on(lazy(|| {
    rwlock.with_write_local(|mut guard| {
        *Rc::get_mut(&mut *guard).unwrap() += 5;
        Ok(()) as Result<(), ()>
    }).unwrap()
}));
assert!(r.is_ok());
assert_eq!(*rwlock.try_unwrap().unwrap(), 5);

Trait Implementations

impl<T: ?Sized> Clone for RwLock<T>

fn clone(&self) -> RwLock<T>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<T: ?Sized + Send> Send for RwLock<T>

impl<T: ?Sized + Send> Sync for RwLock<T>

impl<T: Debug + ?Sized> Debug for RwLock<T>

fn fmt(&self, f: &mut Formatter) -> Result

Formats the value using the given formatter.