2 releases

Uses old Rust 2015

0.3.1 May 4, 2019
0.3.0 Apr 21, 2019

#1351 in Asynchronous

MIT license

51KB
573 lines

Reactive Streams for Rust.

Reactive Programming the Rust way

This library provides 2 basic reactive programming primitives. Stream represents a stateless pipe of data that can be subscribed to and handled asynchronously. ReactiveValue represents a piece of data whose value can change over time (similar to an Atomic, but backed by streams so that dependents be alerted when its value changes). These two primitives loosely correspond to 'Stream' and 'BehaviorSubject' respectively in the Rx family of libraries.

One unique feature of this library is that stream subscriptions only last as long as the subscription object stays in scope, preventing a many of the memory leaks and zombie callback problems common in reactive code.

let stream_host: epoxy::Sink<i32> = epoxy::Sink::new();
let stream = stream_host.get_stream();
{
    let _sub = stream.subscribe(|val| println!("Emitted {}", val));
    stream_host.emit(1); // 'Emitted 1' is printed
    assert_eq!(stream.count_subscribers(), 1);
}
stream_host.emit(2); // Nothing is printed
assert_eq!(stream.count_subscribers(), 0);

Streams can be manipulated using a library of built-in functions based on Rust's set of iterator operations. Currently these operations include:

Operation Property of returned stream
map(fn) Runs all values from the input stream through a mapper function
map_rc(fn) Same as map() but the mapper function takes and returns an Arc
flat_map(fn) Similar to map() but iterates through the result of the mapper function
filter(fn) Returns only input values that pass the given filter function
inspect(method) Passes through the original stream, calls a method for each item
scan(fn, default) Similar to reduce(), but returns the value after each iteration
count_values() Returns the number of times the stream has emitted
buffer(size) Collects emitted values into vectors of length size

ReactiveValues have their own set of operators, although it is also possible to get a reference to the underlying stream of a ReactiveValue with .as_stream() and use any of the above operations as well.

Operation Property of returned reactive value
map(fn) Runs all values from the input stream through a mapper function
sanitize(fn, default) Does not change the value if the input does not pass a test fn
fallback(fn, fallback) Changes the value to fallback if the input does not pass a test fn

However, this library also ships with a computed! macro that makes dealing with ReactiveValue just as easy as dealing with any other Rust variable.

# #[macro_use] extern crate epoxy;
use epoxy::ReactiveValue;

let points = epoxy::ReactiveValue::new(4);
let multiplier = epoxy::ReactiveValue::new(1);
let score = computed!(points * multiplier);

assert_eq!(*score.get(), 4);

multiplier.set(2);
assert_eq!(*score.get(), 8);

Comparisons to other FRP Libraries

Carboxyl / Frappe

Carboxyl and Frappe are the two most common FRP libraries in Rust right now (I have combined them here because they are structured very similarly). This library was inspired by ReactiveX rather than the FRP paper that Carboxyl and Frappe used, so some of the terminology here is different. There are also a number of significant API differences:

  • Epoxy Subscription <-> Carboxyl Observers
    • Subscriptions in Epoxy are unsubscribed when they go out of scope
    • Carboxyl Observers are unsubscribed when the observer function returns False-y
  • Epoxy ReactiveValue <-> Carboxyl Signal
    • Carboxyl Signals cannot be subscribed to (observed), which is a problem for UI frameworks
    • Epoxy ReactiveValues are push, rather than pull, making them less efficient in some cases
  • Epoxy computed! <-> Carboxyl lift!
    • The Epoxy computed! macro extracts variables from the function def, making it more readable
    • Carboxyl's lift! macro has explicitly defined inputs, making it less error-prone

As you can see, there are tradeoffs to both of these frameworks. Epoxy was designed to optimize for frontend use cases, making it easier to integrate with things like DOM libraries.

ReactiveX

These streams are intended to be substantially simpler than those in the ReactiveX family of libraries. The most significant difference is that this library has no concept of a 'cold' stream, meaning no streams will ever emit a value immediately upon subscription. Streams in this library also never close, as they are intended to model long-term asynchronous data flows (of course it is possible to make a stream 'closeable' by making a stream of Option enums and unsubscribing on None, that just isn't built in to the library). Finally, where Rx subscriptions live until explicitly unsubscribed, Rust Reactive subscriptions only live as long as they are in scope.

Status

This crate is under active development and is probably not ready for production use yet.

Dependencies

~2MB
~48K SLoC