Crate eventador[][src]

This crate provides a lock-free Pub/Sub event-bus inspired by the LMAX Disruptor.

Users can configure how publishers handle slow subscribers through WaitStrategy policies.

Both sync and async APIs are available.


Please use the provided example programs for a more thorough approach on how to use this crate.

Basic sync usage:

use eventador::Eventador;
let eventbus = Eventador::new(4).unwrap();
let subscriber = eventbus.subscribe::<usize>();

let i: usize = 1234;

let mut publisher = eventbus.publisher();
publisher.send(i + 1111);

let mut msg = subscriber.recv();
assert_eq!(i, *msg);

msg = subscriber.recv();
assert_eq!(i + 1111, *msg);

Basic async usage:

use eventador::{Eventador, SinkExt, StreamExt};
let eventbus = Eventador::new(4).unwrap();

let mut subscriber = eventbus.async_subscriber::<usize>();
let mut publisher = eventbus.async_publisher::<usize>(4);

let i: usize = 1234;

let msg =;
assert_eq!(i, *msg);


Event-buses ease the development burden of concurrent programs by enabling concurrent application subroutines to interact and affect other subroutines through events.

Eventador embraces the Rust model of Choose Your Guarantees ™ by offering different policies for publishing when subscribers are lagging. These are represented as WaitStrategies, with the default being to wait for all subscribers to read an event before it is overwritten.

Feature Flags

  • async: enables usage of async APIs

Design Considerations

A general overview of the architecture of the library can be found here.

Ring Buffer

Like Eventador, most event-bus implementations use some form of ring buffer for the underlying data structure to store published events. As such, an Eventador instance cannot indefinitely grow to accommodate events, unlike a Vec. Publishers require configurable policies to decide how and when to overwrite old data in the ring.

LMAX Disruptor

The LMAX Disruptor serves as a basis for a lot of event-bus implementations, though the contemporary architecture of the Disruptor looks very different from the one presented in the outdated LMAX white-paper. Eventador draws from the principles of the current Disruptor architecture, but the similarities stop there.

A sequencer atomically assigns an event to an index in the ring buffer on publishing of an event.

Subscribers internally have their own sequencer to determine their last read event in the ring buffer. On receiving a subscribed message, the sequencer is atomically updated to reflect that it can now receive the next event.


Eventador has the potential to be a high-contention (aka bottlenecking) structure to a given concurrent program, so the implementation needs to handle contention as effectively as possible. Atomic CAS operations are generally faster than locking, and is the preferred approach to handle contention.


This crate relies on the use of TypeId to determine what type an event is, and what types of events a subscriber is subscribed to.

Unfortunately, due to the limitations of Rust reflection tools, an Enum will have a different TypeId than an Enum variant. This means that a subscriber must subscribe to the Enum type and ignore any variants it's not interested in that it still receives. Likewise, the publisher must publish events as the Enum type and not the variant in order to maintain that consistency.



A handle to asynchronously publish to the event-bus.


A handle to subscribe to events and receive them asynchronously.


A wrapper that can be de-referenced to access and read the event.


A lock-free and thread-safe event-bus implementation.


An error thrown by the AsyncPublisher as part of the Sink trait implementation.


A handle to publish events to the event-bus.


A handle to receive events that were subscribed to from the event-bus.



Declares the policy for what producers should do when consumers are lagging.



An extension trait for Sinks that provides a variety of convenient combinator functions.


An extension trait for Streams that provides a variety of convenient combinator functions.