|Build Status| |Coverage Status|

tryagain

A lightweight and pythonic retry helper.

tryagain aims to simplify working with unstable functions. Whether you have networking code that sometimes raises timeout exceptions or you are controlling devices which only seem to listen on the second try - tryagain makes it easier to repeat the call.

tryagain offers you hooks to clean up after a failed attempt or to prepare for the next call. You can set a waittime between retries or specify your own waittime function to realize exponential waittimes etc.

tryagain is lightweight, fully tested, MIT licensed and comes as a single python file with no dependencies. It supports Python 2.6+ and 3.2+.

To install, run pip install tryagain.

Basic syntax

Using the tryagain function call:

.. code:: python

import tryagain

def unstable_function():
    # Attention: This function sometimes fails!
    ...

result = tryagain.call(unstable_function,
                       max_attempts=None, exceptions=Exception, wait=0.0,
                       cleanup_hook=None, pre_retry_hook=None)

Using the tryagain decorator retries:

.. code:: python

from tryagain import retries

@retries(max_attempts=3)
def unstable_funcation(arg1, arg2):
    # Attention: This function sometimes fails!
    ...

result = unstable_function('foo', arg2='bar')

Parameters

-  ``func``: The unstable function to call
-  ``max_attemps``: Any integer number to limit the maximum number of
   attempts. Set to None for unlimited retries. (Default = None)
-  ``exceptions``: An iterable of exceptions that should result in a
   retry. (Default = ``Exception``)
-  ``wait``: Can be an integer or float value (to specify a waittime in seconds) or a custom function (see Waittime documentation) (Default = 0.0)
-  ``cleanup_hook``: Can be set to a callable and will be called after
   an exception is raised from calling ``func``. (Default = None)
-  ``pre_retry_hook``: Can be set to any callable that will be called
   before ``func`` is called. (Default = None)


Result
~~~~~~

``tryagain.call`` will return whatever the unstable function would
return. ``tryagain.call`` (and the decorator ``tryagain.retries``) reraises
any exception which is:

-  not in the given ``exceptions``
-  raised in the ``pre_retry_hook`` or in ``cleanup_hook``
-  raised in the last attempt at calling the unstable function.


Quickstart
----------

Retry calling an unstable function

.. code:: python

import tryagain

def unstable():
    ...

# retry calling 'unstable' until it returns without raising an exception
result = tryagain.call(unstable)

# limit to maximum 5 attempts
result = tryagain.call(unstable, max_attempts=5)

# only retry after specific exceptions
result = tryagain.call(unstable, exceptions=(ValueError, TypeError))

Waittimes

The tryagain library allows fixed wait values as well as custom waittime
functions.

.. code:: python

    # wait one second before trying again
    tryagain.call(unstable, wait=1.0)

    # waittime rises linearly (n is the number of attempts)
    # (will wait 1s, 2s, 3s, ...)
    tryagain.call(unstable, wait=lambda n: n)

    # waittime rises exponentially with each attempt
    # (will wait 2s, 4s, 8s, ...)
    tryagain.call(unstable, wait=lambda n: 2 ** n)

    # exponentially rising waittime with maximum
    # (will wait 2s, 4s, 5s, 5s, ..., 5s)
    tryagain.call(unstable, wait=lambda n: min(n ** 2, 5))

    # no waiting time before second attempt, 1.0s afterwards
    def no_first_wait(attempt):
        if attempt == 2:
            return 0
        else:
            return 1.0
    tryagain.call(unstable, wait=no_first_wait)


Retry calling a function with parameters

The tryagain.call-function only supports a function reference as the func parameter. To pass arguments to the unstable function you have to use one of the following idioms:

.. code:: python

# using a lambda
tryagain.call(lambda: unstable('message', some_arg=True), wait=1.0)

# using a partial
from functools import partial
tryagain.call(partial(unstable, 'message', some_arg=True), wait=1.0)

# using a separate function
def call_unstable_function():
    msg = 'message'
    return unstable(msg, some_arg=True)
tryagain.call(call_unstable_function, wait=1.0)

But it is much nicer to wrap your unstable function in the @retries decorator. This way you can call your unstable function with parameters easily:

Function decorator

Instead of using the ``tryagain.call`` function, you can use the ``retries``
decorator.

.. code:: python

    from tryagain import retries
    @retries(max_attempts=3, exceptions=(TypeError, ValueError))
    def unstable(arg1, arg2):
        # your unstable function here

    result = unstable('foo', arg2='bar')

The decorator takes the same arguments as the ``call``-function
except the ``func`` parameter.


Hooks
~~~~~

The tryagain library features two hooks that can be used,
``cleanup_hook`` and ``pre_retry_hook``.

.. code:: python


    def unstable():
        print('Calling unstable function')
        print('Exception!')
        raise Exception

    tryagain.call(unstable, max_attempts=2,
                  wait=lambda n: print('waiting'),
                  cleanup_hook=lambda: print('cleaning up'),
                  pre_retry_hook=lambda: print('do preparations'))
    'Calling unstable function'
    'Exception!'
    'cleaning up'
    'waiting'
    'do preparations'
    'Calling unstable function'
    'Exception!'
    'cleaning up'
    Error: Exception raised...


.. |Build Status| image:: https://travis-ci.org/tfeldmann/tryagain.svg?branch=master
   :target: https://travis-ci.org/tfeldmann/tryagain
.. |Coverage Status| image:: https://coveralls.io/repos/github/tfeldmann/tryagain/badge.svg?branch=master
   :target: https://coveralls.io/github/tfeldmann/tryagain?branch=master