A python retry wrapper/decorator

A simple way to auto retry a funciton that has the possibility of raising an error. You can either call retry directly or decorate a function to get it to retry. Intelligent matching logic allows you to retry certain exception while raising other exceptions.

Call a function directly:

from retrypy import retry, check

def dummy_func():
    print "dummy_func called..."
    raise Exception("House")

retry.call(
    dummy_func,
    times=2
)

dummy_func called...
dummy_func called...
Exception: House

# usign a check method
retry.call(
    dummy_func,
    check = check.message_equals("foobar")
)
dummy_func called...
Exception: House

Decorating a function:

# Only retry IOErrors
@retry.decorate(IOError, times=2)
def dummy_func():
    print "dummy_func called..."
    raise IOError("House")
dummy_func()

dummy_func called...
dummy_func called...
IOError: House

# Retry any Exception, use a custom wait function
@retry.decorate(times=2, wait=lambda n: 2*n)
def dummy_func():
    print "dummy_func called..."
    raise Exception("House")
dummy_func()

dummy_func called...
dummy_func called...
Exception: House

Wrap a function and return a new callable:

def dummy_func():
    print "dummy_func called..."
    raise Exception("House")

func = retry.wrap(
    dummy_func,
    times=2
)
func()

dummy_func called...
dummy_func called...
Exception: House

Delay Helpers:

Functions to implement some common backoff strategies: random, exponential, and incremental.

Random:

import time
from retrypy import retry, delay

def dummy_func():
    print time.time()
    raise Exception("House")

retry.call(dummy_func, wait=delay.random(
    min_seconds=1,
    max_seconds=5,
))

Output (wait times: .66s, .84s, 3.42s, .33s):

1423297137.49
1423297138.15
1423297138.99
1423297142.41
1423297142.74
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "retrypy/retry.py", line 63, in call
    wait,
  File "retrypy/retry.py", line 29, in _retry
    raise previous_exception
Exception: House

Exponential:

import time
from retrypy import retry, delay

def dummy_func():
    print time.time()
    raise Exception("House")

retry.call(dummy_func, wait=delay.exponential(
    start_at=1,
))

Output (wait times: 1s, 2s, 4s, 8s):

1423297238.49
1423297239.49
1423297241.49
1423297245.5
1423297253.5
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "retrypy/retry.py", line 63, in call
    wait,
  File "retrypy/retry.py", line 29, in _retry
    raise previous_exception
Exception: House

Incremental:

import time
from retrypy import retry, delay

def dummy_func():
    print time.time()
    raise Exception("House")

retry.call(dummy_func, wait=delay.incremental(
  start_at=1,
  step=1,
))

Output (wait times: 1s, 2s, 3s, 4s):

1423297301.64
1423297302.64
1423297304.64
1423297307.65
1423297311.65
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "retrypy/retry.py", line 63, in call
    wait,
  File "retrypy/retry.py", line 29, in _retry
    raise previous_exception
Exception: House

Custom Delay Functions:

You can write your own delay functions, their only requirements are that they take an Integer and return a Number of seconds to wait.

def custom_delay(call_count):
    if call_count == 1:
        # don't wait at all the first time
        return 0

    # wait 4, 8, 16, 32
    return 2 ** call_count

Builtin Exception Checkers:

Exception Checkers can be used to check if you want to retry a specific exception. If the check function returns true then the exception is retryable otherwise we will not catch the exception and retry. The available checkers are: message_equals, message_contains, and message_matches.

message_equals, will match any Exception with a message that is identical to the string provided.

message_contains, will match any Exception with a message that contains the string provided.

message_matches, will match any Exception with a message that matches the regex provided. The regex may be passed as a string or a compiled regex pattern.

Custom Exception Checkers:

You can write your own exception checkers, their only requirements are that they: take an Exception and an Integer as parameters. They should return True if the exception is retryable otherwise False.

def custom_matcher(e, call_count):
    # never fail their first time no matter what
    if call_count == 1:
        return True

    # only retry errors with Bob Barker in the message.
    return "Bob Barker" in str(e):

Installation:

>> pip install retrypy

Development:

>> git clone https://github.com/toddsifleet/retrypy
>> cd retrypy
>> make bootstrap
>> make

License:

See LICENSE

thibault-ml/retrypy