repdeco 0.1.3

Retry, Cache, Circuit Breaker decorator for Python (sync + async)

🛡️ Repdeco

Repdeco (Resilience & Protection Decorator) is a powerful, all-in-one Python decorator designed to make your functions bulletproof. Whether you are dealing with flaky APIs, unstable database connections, or heavy computations, repdeco has you covered.

It provides a unified, elegant interface for Retries, Exponential Backoff, Timeouts, Circuit Breaking, and Memory Caching — seamlessly supporting both sync and async functions.

---

🚀 Features

• 🔄 Retries & Backoff — Automatically retry failing operations with optional exponential backoff.
• ⏱️ Timeouts — Stop execution if a function takes too long (safely works for both async and sync functions using ThreadPool).
• 🔌 Circuit Breaker — Prevent cascading failures by failing fast when a downstream system is down.
• 💾 Caching — In-memory caching with Time-To-Live (TTL) to speed up repeated calls.
• ⚡ Sync & Async Ready — Use the exact same @repdeco decorator for standard def and async def.
• 📝 Metadata Preserved — Keeps your original function names and docstrings intact (@functools.wraps support).

---

📦 Installation

pip install repdeco

---

💡 Quick Start & Usage

Import the decorator into your project:

import time
import asyncio
from repdeco.core import repdeco

---

1. Retries with Exponential Backoff

If the function raises an exception, it will retry up to retry times. The backoff=1 means it will wait 1s, 2s, and 4s between attempts.

@repdeco(retry=3, backoff=1)
def fetch_data():
    print("Fetching data...")
    raise ConnectionError("Network unstable")

# Will try 4 times total (1 initial + 3 retries) before raising the error.

---

2. Timeouts (Sync & Async)

Enforce a strict time limit. If the function exceeds timeout seconds, it will raise a TimeoutError.

@repdeco(timeout=2)
def slow_sync_task():
    time.sleep(5)
    return "Done"

@repdeco(timeout=2)
async def slow_async_task():
    await asyncio.sleep(5)
    return "Done"

---

3. Circuit Breaker

If a function fails cb_threshold times in a row, the circuit opens. It immediately blocks further calls for cb_timeout seconds, raising an Exception instantly without even trying to run the failing function.

@repdeco(cb_threshold=5, cb_timeout=60)
def fragile_api_call():
    # If this fails 5 times, it gets blocked for 60 seconds
    pass

Circuit states:

CLOSED (normal) → [N failures] → OPEN (blocked) → [timeout expires] → HALF-OPEN → [success] → CLOSED

---

4. Caching with TTL (Time-To-Live)

Cache the results of expensive function calls based on their arguments.

@repdeco(cache_ttl=300)  # Cache for 5 minutes
def heavy_computation(x, y):
    print("Calculating...")
    time.sleep(3)
    return x * y

print(heavy_computation(5, 5))  # Takes 3 seconds → returns 25
print(heavy_computation(5, 5))  # Instant! (Returns cached 25)

---

5. The "All-in-One" Combo

You can combine all features to create highly resilient functions. The decorator evaluates logic in this order:

Cache → Circuit Breaker → Timeout → Retries

@repdeco(
    retry=2,
    backoff=2,
    timeout=5,
    cache_ttl=60,
    cb_threshold=3,
    cb_timeout=30
)
async def resilient_web_scraper(url):
    # Your robust scraping logic here
    pass

---

⚙️ Configuration Parameters

Parameter	Type	Default	Description
retry	int	0	Number of times to retry the function after an initial failure.
retry_on	tuple	None	Specific exception classes to catch and retry on (e.g., (ConnectionError,)). If None, retries on all exceptions.
backoff	int/float	0	Base multiplier for exponential backoff sleep time (backoff * 2^attempt). 0 means retry immediately.
timeout	int/float	None	Maximum seconds the function is allowed to run before raising a TimeoutError.
cache_ttl	int/float	0	Time-To-Live in seconds for the in-memory cache. 0 disables caching.
cb_threshold	int	5	Number of consecutive failures before the Circuit Breaker opens.
cb_timeout	int/float	10	Seconds the Circuit Breaker remains open before attempting a half-open retry.

---

📜 License

This project is licensed under the MIT License. See the LICENSE file for details.