🗄️ purecache

Async-native in-memory cache with pluggable eviction backends — pure Python 3.12+, zero dependencies.

Just asyncio, collections.OrderedDict, and the irrational urge to understand what happens inside the black box.

Part of the pure-python-system-design project.

📦 Installation

pip install purecache

Or with uv:

uv add purecache

Python 3.12+ required.

⚡ Quick Start

Direct backend usage

from purecache.backends.lru import LRUCache

cache = LRUCache(capacity=128)

await cache.put("user:42", {"name": "Alice"})
value = await cache.get("user:42")  # {"name": "Alice"}
value = await cache.get("missing")  # None

Decorator

from purecache.decorators import cache
from purecache.backends.lru import LRUCache

@cache(backend=LRUCache, capacity=128)
async def get_user(user_id: str) -> dict:
    return await fetch_from_db(user_id)

# First call — executes get_user, caches result
user = await get_user("42")

# Second call — returns cached result, skips get_user
user = await get_user("42")

Cache keys are derived automatically from the function's arguments using pickle + SHA-256 — positional args keep their order, keyword args are sorted by name.

🧠 Backends

All backends implement the ICacheBackend protocol — swap them without touching your application code.

🔌 Framework Examples

The decorator integrates naturally with any async framework:

# FastAPI
from fastapi import FastAPI
from purecache.decorators import cache
from purecache.backends.lru import LRUCache

app = FastAPI()

@app.get("/user/{user_id}")
@cache(backend=LRUCache, capacity=512)
async def get_user(user_id: str):
    return await fetch_user_from_db(user_id)

TODO: Add more examples for aiohttp, Django, Flask, Litestar, and Sanic in examples/.

📐 Architecture

cache() decorator
  └── ICacheBackend (protocol)
        ├── LRUCache    — OrderedDict + move_to_end
        ├── LFUCache    — key_map + freq_map + min_freq pointer
        ├── TTLCache    — dict + expiry timestamps
        └── LRUTTLCache — LRU + TTL combined

The cache() decorator handles key generation and cache lookup. The backend handles storage and eviction. Swap the backend, keep everything else.

⚠️ Known Limitations

• Caching None: The decorator uses if cached_res is not None as the cache-hit check. Functions that legitimately return None will always miss — the value won't be cached. Use a sentinel-aware backend or wrap the return value if needed.

📋 Requirements

• Python 3.12+
• Courage

🧪 Development

uv sync
pre-commit install

uv run pytest
uv run ruff check .
uv run mypy src/
uv run mkdocs serve

📖 Documentation

Full docs at https://pure-python-system-design.github.io/purecache/

More designs to come, if the pizza supply holds.