Macro futures_util::select_biased
source · macro_rules! select_biased {
($($tokens:tt)*) => { ... };
}
Expand description
Polls multiple futures and streams simultaneously, executing the branch
for the future that finishes first. Unlike select!
, if multiple futures are ready,
one will be selected in order of declaration. Futures directly
passed to select_biased!
must be Unpin
and implement FusedFuture
.
If an expression which yields a Future
is passed to select_biased!
(e.g. an async fn
call) instead of a Future
by name the Unpin
requirement is relaxed, since the macro will pin the resulting Future
on the stack. However the Future
returned by the expression must
still implement FusedFuture
.
Futures and streams which are not already fused can be fused using the
.fuse()
method. Note, though, that fusing a future or stream directly
in the call to select_biased!
will not be enough to prevent it from being
polled after completion if the select_biased!
call is in a loop, so when
select_biased!
ing in a loop, users should take care to fuse()
outside of
the loop.
select_biased!
can be used as an expression and will return the return
value of the selected branch. For this reason the return type of every
branch in a select_biased!
must be the same.
This macro is only usable inside of async functions, closures, and blocks.
It is also gated behind the async-await
feature of this library, which is
activated by default.
Examples
use futures::future;
use futures::select_biased;
let mut a = future::ready(4);
let mut b = future::pending::<()>();
let res = select_biased! {
a_res = a => a_res + 1,
_ = b => 0,
};
assert_eq!(res, 5);
use futures::future;
use futures::stream::{self, StreamExt};
use futures::select_biased;
let mut st = stream::iter(vec![2]).fuse();
let mut fut = future::pending::<()>();
select_biased! {
x = st.next() => assert_eq!(Some(2), x),
_ = fut => panic!(),
};
As described earlier, select_biased
can directly select on expressions
which return Future
s - even if those do not implement Unpin
:
use futures::future::FutureExt;
use futures::select_biased;
// Calling the following async fn returns a Future which does not
// implement Unpin
async fn async_identity_fn(arg: usize) -> usize {
arg
}
let res = select_biased! {
a_res = async_identity_fn(62).fuse() => a_res + 1,
b_res = async_identity_fn(13).fuse() => b_res,
};
assert!(res == 63 || res == 12);
If a similar async function is called outside of select_biased
to produce
a Future
, the Future
must be pinned in order to be able to pass
it to select_biased
. This can be achieved via Box::pin
for pinning a
Future
on the heap or the pin_mut!
macro for pinning a Future
on the stack.
use futures::future::FutureExt;
use futures::select_biased;
use futures::pin_mut;
// Calling the following async fn returns a Future which does not
// implement Unpin
async fn async_identity_fn(arg: usize) -> usize {
arg
}
let fut_1 = async_identity_fn(1).fuse();
let fut_2 = async_identity_fn(2).fuse();
let mut fut_1 = Box::pin(fut_1); // Pins the Future on the heap
pin_mut!(fut_2); // Pins the Future on the stack
let res = select_biased! {
a_res = fut_1 => a_res,
b_res = fut_2 => b_res,
};
assert!(res == 1 || res == 2);
select_biased
also accepts a complete
branch and a default
branch.
complete
will run if all futures and streams have already been
exhausted. default
will run if no futures or streams are
immediately ready. complete
takes priority over default
in
the case where all futures have completed.
A motivating use-case for passing Future
s by name as well as for
complete
blocks is to call select_biased!
in a loop, which is
demonstrated in the following example:
use futures::future;
use futures::select_biased;
let mut a_fut = future::ready(4);
let mut b_fut = future::ready(6);
let mut total = 0;
loop {
select_biased! {
a = a_fut => total += a,
b = b_fut => total += b,
complete => break,
default => panic!(), // never runs (futures run first, then complete)
};
}
assert_eq!(total, 10);
Note that the futures that have been matched over can still be mutated
from inside the select_biased!
block’s branches. This can be used to implement
more complex behavior such as timer resets or writing into the head of
a stream.