riprova

riprova (meaning retry in Italian) is a small, general-purpose and versatile Python library that provides retry mechanisms with multiple backoff strategies for any sort of failed operations.

It's domain agnostic, highly customizable, extensible and provides a minimal API that's easy to instrument in any code base via decorators, context managers or raw API consumption.

For a brief introduction about backoff mechanisms for potential failed operations, read this article.

Features

• Retry decorator for simple and idiomatic consumption.
• Simple Pythonic programmatic interface.
• Maximum retry timeout support.
• Supports error whitelisting and blacklisting.
• Supports custom error evaluation retry logic (useful to retry only in specific cases).
• Automatically retry operations on raised exceptions.
• Supports asynchronous coroutines with both async/await and yield from syntax.
• Configurable maximum number of retry attempts.
• Highly configurable supporting max retries, timeouts or retry notifier callback.
• Built-in backoff strategies: constant, fibonacci and exponential backoffs.
• Supports sync/async context managers.
• Pluggable custom backoff strategies.
• Lightweight library with almost zero embedding cost.
• Works with Python +2.6, 3.0+ and PyPy.

Backoff strategies

List of built-in backoff strategies.

• Constant backoff
• Fibonacci backoff
• Exponential backoff

You can also implement your own one easily. See ConstantBackoff for an implementation reference.

Installation

Using pip package manager (requires pip 1.9+. Upgrade it running: pip install -U pip):

pip install -U riprova

Or install the latest sources from Github:

pip install -e git+git://github.com/h2non/riprova.git#egg=riprova

API

• riprova.retry
• riprova.Retrier
• riprova.AsyncRetrier
• riprova.Backoff
• riprova.ConstantBackoff
• riprova.FibonacciBackoff
• riprova.ExponentialBackoff
• riprova.ErrorWhitelist
• riprova.ErrorBlacklist
• riprova.add_whitelist_error
• riprova.RetryError
• riprova.RetryTimeoutError
• riprova.MaxRetriesExceeded
• riprova.NotRetriableError

Examples

You can see more featured examples from the documentation site.

Basic usage examples:

import riprova

@riprova.retry
def task():
    """Retry operation if it fails with constant backoff (default)"""

@riprova.retry(backoff=riprova.ConstantBackoff(retries=5))
def task():
    """Retry operation if it fails with custom max number of retry attempts"""

@riprova.retry(backoff=riprova.ExponentialBackOff(factor=0.5))
def task():
    """Retry operation if it fails using exponential backoff"""

@riprova.retry(timeout=10)
def task():
    """Raises a TimeoutError if the retry loop exceeds from 10 seconds"""

def on_retry(err, next_try):
    print('Operation error: {}'.format(err))
    print('Next try in: {}ms'.format(next_try))

@riprova.retry(on_retry=on_retry)
def task():
    """Subscribe via function callback to every retry attempt"""

def evaluator(response):
    # Force retry operation if not a valid response
    if response.status >= 400:
        raise RuntimeError('invalid response status')  # or simple return True
    # Otherwise return False, meaning no retry
    return False

@riprova.retry(evaluator=evaluator)
def task():
    """Use a custom evaluator function to determine if the operation failed or not"""

@riprova.retry
async def task():
    """Asynchronous coroutines are also supported :)"""

Retry failed HTTP requests:

import pook
import requests
from riprova import retry

# Define HTTP mocks to simulate failed requests
pook.get('server.com').times(3).reply(503)
pook.get('server.com').times(1).reply(200).json({'hello': 'world'})


# Retry evaluator function used to determine if the operated failed or not
def evaluator(response):
    if response != 200:
        return Exception('failed request')  # you can also simply return True
    return False


# On retry even subscriptor
def on_retry(err, next_try):
    print('Operation error {}'.format(err))
    print('Next try in {}ms'.format(next_try))


# Register retriable operation
@retry(evaluator=evaluator, on_retry=on_retry)
def fetch(url):
    return requests.get(url)


# Run task that might fail
fetch('http://server.com')

License

MIT - Tomas Aparicio