picocache 0.2.0

persistent decorators mirroring functools.lru_cache

PicoCache

Persistent, datastore‑backed lru_cache for Python.
PicoCache gives you the ergonomics of functools.lru_cache while keeping your cached values safe across process restarts and even across machines.
Two back‑ends are provided out of the box:

• SQLAlchemyCache – persists to any RDBMS supported by SQLAlchemy (SQLite, Postgres, MySQL, …).
• RedisCache – stores values in Redis, perfect for distributed deployments.

---

Why PicoCache?

• Familiar API – decorators feel identical to functools.lru_cache.
• Durable – survive restarts, scale horizontally.
• Introspectable – cache_info() and cache_clear() just like the standard library.
• Zero boilerplate – pass a connection URL and start decorating.

---

Installation

pip install picocache

---

Quick‑start

1. SQL (SQLite example)

from picocache import SQLAlchemyCache

# Create the decorator bound to an SQLite file
sql_cache = SQLAlchemyCache("sqlite:///cache.db")

@sql_cache(maxsize=256)        # feels just like functools.lru_cache
def fib(n: int) -> int:
    return n if n < 2 else fib(n - 1) + fib(n - 2)

2. Redis

from picocache import RedisCache

redis_cache = RedisCache("redis://localhost:6379/0")

@redis_cache(maxsize=128, typed=True)
def slow_add(a: int, b: int) -> int:
    print("Executing body…")
    return a + b

On the second call with the same arguments, slow_add() returns instantly and “Executing body…” is not printed – the result came from Redis.

---

API

Each decorator object is initialised with connection details and called with the same signature as functools.lru_cache:

SQLAlchemyCache(url_or_engine, *, key_serializer=None, value_serializer=None, ...)
RedisCache(url_or_params, *, key_serializer=None, value_serializer=None, ...)

__call__(maxsize=128, typed=False)

Returns a decorator that memoises the target function.

Param	Type	Default	Meaning
maxsize	int/None	128	Per‑function entry limit (None → no limit).
typed	bool	False	Treat arguments with different types as distinct (same as stdlib).

The wrapped function gains:

• .cache_info() → namedtuple(hits, misses, currsize, maxsize)
• .cache_clear() → empties the persistent store for that function.

---

Running the tests

uv sync
just test

• SQL tests run against an in‑memory SQLite DB (no external services).
• Redis tests are skipped automatically unless a Redis server is available on localhost:6379.

---

License

MIT – see LICENSE for details.