pysergio / atomcache

Asynchronous cache manager designed for horizontally scaled web servers.

Introduction

Asynchronous cache manager designed for horizontally scaled web applications. NOTE: Currently has implementation only for FastAPI using Redis.

Requirements

Python 3.7+

• redis for cache implementation.
• FastAPI for the web parts.

Installation

$ pip install atomcache

---> 100%

Explanation schema

As UML

@startuml
    !theme materia
    participant Redis
    participant Instance_A as A
    participant Instance_B as B
    participant Instance_N as C


    B <-> Redis: GET: Cache=null & GET: Lock=null

    B <-> Redis: SET: Lock = true

    activate B #Red
    A <--> Redis: GET: Cache=null & GET: Lock=true
    activate A #Transparent
    C <--> Redis: GET: Cache=null & GET: Lock=true
    activate C #Transparent
    B <--> B: Do the computation
    B -> Redis: SET: Cache={...}
    deactivate B

    group Notify Cache SET
        Redis -> C
        Redis -> A
    end
    group GET Cache
        Redis <-> C
    deactivate C
        Redis <-> A
    deactivate A
    end
@enduml

Examples:

Usage as FastAPI Dependency

• Create a file events.py with:

from typing import Optional, Callable

from redis.asyncio import Redis
from fastapi import FastAPI, Depends
from atomcache import Cache


def create_start_app_handler(app: FastAPI) -> Callable:
    async def start_app() -> None:
        redis: Redis = await Redis.from_url(url="redis://localhost", encoding="utf-8")
        await Cache.init(app, redis)

    return start_app


def create_stop_app_handler(app: FastAPI) -> Callable:
    async def stop_app() -> None:
        await Cache.backend.close()

    return stop_app

• Create a file main.py with:

from typing import Optional

from fastapi import FastAPI, Depends
from atomcache import Cache

from .events import create_start_app_handler, create_stop_app_handler

app = FastAPI()

app.add_event_handler("startup", create_start_app_handler(app))
app.add_event_handler("shutdown", create_stop_app_handler(app))


@router.get("/resources", response_model=List[TheResponseModel], name="main:test-example")
async def resources(offset: int = 0, items: int = 10, cache: Cache = Depends(Cache(exp=600)):
    cache_id = f"{offset}-{items}"  # Build cache identifier
    await cache.raise_try(cache_id)  # Try to respond from cache
    response = await db.find(TheResponseModel, skip=offset, limit=items)
    await asyncio.sleep(10)  # Do some heavy work for 10 sec, see `lock_timeout`
    return cache.set(response, cache_id=cache_id)

Direct cache usage for avoiding repetitive calling on external resources:

from aiohttp import ClientSession
from atomcache import Cache

cache = Cache(exp=1200, namespace="my-namespace:")


async def requesting_helper(ref: str) -> List[dict]:
    cached_value = await cache.get(cache_id=ref)
    if cached_value is not None:
        return cached_value

    async with ClientSession() as session:
        async with session.get(f"https://external-api.io/{ref}") as response:
            if response.ok:
                cached_value = response.json()
                return cache.set(cached_value, cache_id=ref)
    return []

atomcache's Issues

Setup code unit-test pipeline

Finish and test switch to aioredis v2

Separate fastAPI logic from lock General porpouse cache implementation

Config KEA for redis container

Setup code linting pipline

Implement FastAPICache as extension of Cache