Modern Python Caching

Cachex is a modern caching library for Python 3 built from the ground up to support type hints. It has built in Redis, MongoDB and Memcached support and provides both sync and async APIs for all storage backends.

Installation

pip install cachex

Redis Dependencies:

pip install cachex[redis]

MongoDB Dependencies:

pip install cachex[mongo]

Memcached Dependencies:

No additional dependencies are required for Memcached support. Unlike Redis and MongoDB, there is not a single client library supported by the core dev team and most libraries do not support both sync and async API's. For that reason, the Memcached storage backend is implemented as an interface based API. You bring your own client, and as long as that client adheres to the MemcachedClient or AsyncMemcachedClient interfaces it will work with Cachex.

Typing Support

Cachex was built to work with type declarations. This means it offers great editor support for decorated functions. It also means that Cachex plays nicely with other popular libraries that rely heavily on typing. FastAPI and Litestar are two popular ASGI web frameworks that offer powerful dependency injection systems which rely on type hints (via Pydantic). Cachex can be inserted anywhere in a dedendency chain and it will "just work". This offers tremendous flexibility to the developer when designing applications.

Basic Example

The example below is very simple (and mostly useless) web server intended to give you a feel for how Cachex works and how it's native typing support can be leveraged in a dependency chain.

from enum import Enum
from typing import Annotated

import pandas as pd
from cachex import cache_value
from fastapi import Depends, FastAPI, Response
from pydantic import AnyHttpUrl
from pydantic_core import Url


app = FastAPI()

class DataSet(Enum):
    UBER1 = "uber1"
    UBER2 = "uber2"
    UBER3 = "uber3"

URLS: dict[DataSet, AnyHttpUrl] = {
    DataSet.UBER1: AnyHttpUrl("https://github.com/plotly/datasets/raw/master/uber-rides-data1.csv"),
    DataSet.UBER2: AnyHttpUrl("https://github.com/plotly/datasets/raw/master/uber-rides-data2.csv"),
    DataSet.UBER3: AnyHttpUrl("https://github.com/plotly/datasets/raw/master/uber-rides-data3.csv"),
}


def get_url(dataset: DataSet) -> AnyHttpUrl:
    """Get the dataset URL from the global mapping."""
    return URLS[dataset]


# We want to cache the result of the download and use that as a dependency.
# Notice, AnyHttpUrl is not natively hashable by cachex so we need to provide a type encoder
@cache_value(type_encoders={Url: lambda t: t.unicode_string()})
def download_csv_data(url: Annotated[AnyHttpUrl, Depends(get_url)]) -> pd.DataFrame:
    """Download a CSV file from the given URL and convert it to a DataFrame."""
    return pd.read_csv(url.unicode_string())


@app.get("/datasets/{dataset}")
def get_dataset(df: Annotated[pd.DataFrame, Depends(download_csv_data)]):
    """Download a CSV dataset from the web and return the data in JSON form."""
    data = df.iloc[0:1000, :].to_json(orient="records", indent=2)
    return Response(content=data, media_type="application/json")

Save the file, call it main.py. Run the app with uvicorn main:app and head to the docs (/docs). The swagger docs show the correct API arguments and types through the dependency chain...

Next, try the /datasets/uber1 endpoint. Depending on your internet connection, this may take 2-30 seconds to run. After it runs once, run it again, the response should load almost instantly! You can then repeat this with the /datasets/uber2 and datasets/uber3 endpoints if you'd like.

Documentation

• Data Caching
• Object Caching
• Type Encoders
• Storage Factories and Factory Keys