High Performance Middleware Streams 🏎️
Asyncrush is a high-performance stream processing library that leverages the middleware pattern inspired by RxJS. Unlike traditional stream libraries, Asyncrush connects middleware using pure functions instead of operators, offering a flexible, efficient, and scalable approach to stream processing. It dramatically improves performance while minimizing memory usage, making it the ideal choice for high-throughput applications.
- Blazing Fast Performance: Achieves up to +89% faster performance compared to traditional stream libraries.
- Middleware-based Stream Control: Process streams with middleware instead of operators, allowing for dynamic error handling, custom retry policies, and advanced state management.
- Memory Efficiency: Minimized object creation and small code size lead to significant reductions in memory consumption.
- Developer-Friendly API: The pure function-based API ensures that the library is easy to understand, use, and maintain.
- Built-in Error Handling & Backoff: Automatic error handling and backoff strategies allow for stable and resilient stream processing, even in the face of network or data failures.
Unpacked + Full feature code size : 110 kB
npm install asyncrushyarn add asyncrushpnpm add asyncrushbun install asyncrushconst stream = new RushStream<number>((observer) => {
observer.next(1);
});
stream.use(
(v: number) => v + 1,
(v: number) => v * 2
).listen({
next: (value) => {
console.log(value);
},
complete: () => { },
});(10,000,000 events vs RxJS)
| Transforms | ops/sec | How much faster |
|---|---|---|
| 200 | 612,033 | +49% ( > 409,546) |
| 150 | 818,150 | +52% ( > 538,197) |
| 100 | 1,205,331 | +89% ( > 636,078) |
| 50 | 2,335,202 | +60% ( > 1,463,261) |
| 25 | 4,535,603 | +28% ( > 3,547,165) |
| 10 | 12,049,110 | +32% ( > 9,138,667) |
| 5 | 20,226,773 | +16% ( > 17,410,811) |
(1,000,000,000 events vs RxJS)
| Transforms | ops/sec | How much faster |
|---|---|---|
| 200 | 2,614,347 | +67% ( > 1,565,985) |
| 150 | 3,346,151 | +53% ( > 2,179,874) |
| 100 | 4,988,732 | +47% ( > 3,401,912) |
| 50 | 9,333,233 | +33% ( > 7,029,017) |
| 25 | 16,665,887 | +23% ( > 13,503,015) |
| 10 | 40,911,683 | +18% ( > 34,761,997) |
| 5 | 76,309,147 | +14% ( > 66,908,963) |
Tested with randomly selected operators/middleware processes,
processing 1,000,000 batches from a total of 1 billion events, 10 million events
with operator selection based on Math.random values.
For reference, running the same operators/middleware repeatedly can achieve up to 8x better performance due to JIT optimizations. However, this benchmark intentionally minimizes JIT optimizations to provide a more realistic, production-like performance measurement.