https://github.com/zweifisch/sieve

a simple yet efficient cache
https://github.com/zweifisch/sieve

Last synced: 4 months ago
JSON representation

a simple yet efficient cache

• Host: GitHub
• URL: https://github.com/zweifisch/sieve
• Owner: zweifisch
• License: mit
• Created: 2024-02-15T12:11:28.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-02-19T12:43:30.000Z (about 1 year ago)
• Last Synced: 2025-01-08T11:52:23.589Z (4 months ago)
• Language: TypeScript
• Size: 862 KB
• Stars: 6
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# sieve

a simple yet efficient cache, [original introduction](https://cachemon.github.io/SIEVE-website/blog/2023/12/17/sieve-is-simpler-than-lru/)

![algo animation](https://cachemon.github.io/SIEVE-website/blog/assets/sieve/sieve_diagram_animation.gif)

## Usage

for Node.js, install via npm: `npm install @zf/sieve`

```typescript

import { SieveCache, LRUCache, LFUCache } from '@zf/sieve'

const cache = new SieveCache(3 /* capacity */)

cache.set('key', 'value')

cache.get('key')

```

for Deno

```typescript

import { SieveCache, LRUCache, LFUCache } from "https://deno.land/x/sieve/mod.ts"

```

## Benchmark

[Benchmark](/benchmark.ts) reading 1 million normally distributed items through a cache with a capacity of 100 compared with the [LRU](https://deno.land/x/[email protected]) package 

showing SIEVE is more performant, while the cache hit/miss ratio is about the same:

![chart](/chart.png)

It turned out that the LRU package's implementation is not very efficient, so I wrote my own [LRU](/lru.ts), and it is actually better than SIEVE:

![chart](/lru.png)

Anyway, the cache hit/miss ratio is of much greater importance, and it is determined by the data distribution.

Added [LFU](/lfu.ts), which has the best hit rate for normally distributed data:

![chart](lfu.png)

## Dev

```sh

deno test

```

```sh

deno run --allow-all build_npm.ts 1.0.0

```