A ringbuffer that implements AsyncRead/AsyncWrite.
It can be used for testing async network crates cross-platform without having to make TCP connections. The crate provides a type Endpoint
which allows creating both ends of a fake network stream with a ringbuffer in each direction.
It facilitates testing more complex situations like back pressure.
It can also be used as an in memory buffer for communicating between async tasks. I haven't done benchmarks yet.
There are currently 2 versions of the AsyncRead/Write traits. The futures-rs version and the tokio version. This crate implements the futures version. You can get the tokio version by using tokio_util::compat
.
Data in transit is held in an internal RingBuffer from the ringbuf crate.
When the sketchy
feature is enabled, a type [Sketchy
] is available that randomizes the behavior of the in memory buffers which would otherwise always be ready which isn't very realistic for testing code that will run against actual network connections later. This will randomly return pending and fill only partial buffers.
With cargo add:
cargo add futures_ringbuf
With cargo yaml:
dependencies:
futures_ringbuf: ^0.4
With raw Cargo.toml
[dependencies]
futures_ringbuf = "^0.4"
Please check out the changelog when upgrading.
This crate has few dependencies. Cargo will automatically handle its dependencies for you.
The sketchy
feature will turn on the Sketchy
type which allows randomly changing the behavior of an async stream to enable testing situations that occur on an actual network like timing out, processing only partial buffers, pending, ...
This crate uses #![ forbid(unsafe_code) ]
, but it's dependencies use quite some unsafe. On first sight the unsafe usage in ringbuf
looks sound, but I haven't scrutinized every detail of it and it's not documented.
A lot of unsafe code is present in the futures library, which I haven't reviewed.
The crate provides a RingBuffer<T>
struct which implements AsyncRead
/AsyncWrite
from the futures library
when T
is u8. You can now call split
provided by AsyncRead
and treat them as both ends of a network connection.
The reader will return Poll::Pending
when the buffer is empty, and the writer when the buffer is full. They will
wake each other up when new data/space is available.
If you want to play with std::io::Read
/std::io::Write
, check out the ringbuf
crate directly, as it's Producer
and
Consumer
types implement these traits, so I didn't include them here.
I haven't yet included Stream<T>
, Sink<T>
, because on u8
that doesn't make much sense, but if there is demand,
it can definitely be added.
The requirements on T
are T: Sized + Copy
.
If you want to seed the buffer before using it with futures_ringbuf, you can use the Producer
and Consumer
types of ringbuf. futures_ringbuf::RingBuffer
implements From< (Producer<T>, Consumer<T>) >
.
This crate works on WASM. See the integration test for some code.
//! Frame a RingBuf with futures_codec. This example shows how the sending task will
//! block when the buffer is full. When a reader consumes the buffer, the sender is woken up.
//!
//! Run with `cargo run --example basic`.
//
use
{
futures_ringbuf :: { * } ,
futures :: { SinkExt, StreamExt, executor::block_on, join } ,
asynchronous_codec :: { Framed, LinesCodec } ,
};
#[ async_std::main ]
//
async fn main()
{
let mock = RingBuffer::new( 13 );
let (mut writer, mut reader) = Framed::new( mock, LinesCodec{} ).split();
let send_task = async move
{
writer.send( "Hello World\n".to_string() ).await.expect( "send" );
println!( "sent first line" );
writer.send( "Second line\n".to_string() ).await.expect( "send" );
println!( "sent second line" );
writer.close().await.expect( "close sender" );
println!( "sink closed" );
};
let receive_task = async move
{
// If we would return here, the second line will never get sent
// because the buffer is full.
//
// return;
while let Some(msg) = reader.next().await.transpose().expect( "receive message" )
{
println!( "Received: {:#?}", msg );
}
};
// Poll them in concurrently
//
join!( send_task, receive_task );
}
When using one ringbuffer, we get both ends of one connection. If we want a more realistic duplex connection, we
need two ringbuffers, with one endpoint reading from the ringbuffer the other endpoint is writing to. Tasks need
to be woken up correctly when new data or space becomes available... To facilitate this, an Endpoint
type is provided which will take care of this setup for you.
use
{
futures_ringbuf :: { * } ,
futures :: { AsyncWriteExt, AsyncReadExt, executor::block_on } ,
};
#[ async_std::main ]
//
async fn main()
{
// Buffer of 10 bytes in each direction. The buffer size always refers to the writing side, so here
// the first 10 means the server can write 10 bytes before it's buffer is full.
// When it's full it will return pending on writing and when it's empty it returns
// pending on reading.
//
let (mut server, mut client) = Endpoint::pair( 10, 10 );
let data = vec![ 1,2,3 ];
let mut read = [0u8;3];
server.write( &data ).await.expect( "write" );
let n = client.read( &mut read ).await.expect( "read" );
assert_eq!( n , 3 );
assert_eq!( read, vec![ 1,2,3 ][..] );
}
API documentation can be found on docs.rs.
Please check out the contribution guidelines.
cargo test
On WASM, after installing wasm-pack:
wasm-pack test --firefox --headless
or
wasm-pack test --chrome --headless
Any of the behaviors described in point 4 "Unacceptable Behavior" of the Citizens Code of Conduct are not welcome here and might get you banned. If anyone including maintainers and moderators of the project fail to respect these/your limits, you are entitled to call them out.