Use the enum library for the retry strategy enumerations

Instead of having a set of string constants that are the
retry strategy/decision result have it instead be an enumerator
that can be more clearly associated in docs and in code that
these values are explicitly the strategies that the retry
code can use (making it more obvious to users of these constants
what they are for).

This also updates the retry docs to use this new linkable
class when describing the various strategies that can be
used/returned.

Change-Id: I944e521562be26f5315d7553da8d5820fc284c49
This commit is contained in:
Joshua Harlow
2015-02-17 21:39:48 -08:00
committed by Joshua Harlow
parent f874b396b7
commit 37b72ec7b5
5 changed files with 70 additions and 26 deletions

View File

@@ -58,20 +58,11 @@ associated `scope`_).
Currently derivatives of the :py:class:`retry <taskflow.retry.Retry>` base
class must provide a :py:func:`~taskflow.retry.Retry.on_failure` method to
determine how a failure should be handled.
determine how a failure should be handled. The current enumeration(s) that can
be returned from the :py:func:`~taskflow.retry.Retry.on_failure` method
are defined in an enumeration class described here:
The current enumeration set that can be returned from this method is:
* ``RETRY`` - retries the surrounding subflow (a retry object is associated
with a flow, which is typically converted into a graph hierarchy at
compilation time) again.
* ``REVERT`` - reverts only the surrounding subflow but *consult* the
parent atom before doing this to determine if the parent retry object
provides a different reconciliation strategy (retry atoms can be nested, this
is possible since flows themselves can be nested).
* ``REVERT_ALL`` - completely reverts a whole flow.
.. autoclass:: taskflow.retry.Decision
To aid in the reconciliation process the
:py:class:`retry <taskflow.retry.Retry>` base class also mandates
@@ -177,7 +168,13 @@ Interfaces
==========
.. automodule:: taskflow.task
.. automodule:: taskflow.retry
.. autoclass:: taskflow.retry.Retry
.. autoclass:: taskflow.retry.History
.. autoclass:: taskflow.retry.AlwaysRevert
.. autoclass:: taskflow.retry.AlwaysRevertAll
.. autoclass:: taskflow.retry.Times
.. autoclass:: taskflow.retry.ForEach
.. autoclass:: taskflow.retry.ParameterizedForEach
Hierarchy
=========
@@ -185,5 +182,10 @@ Hierarchy
.. inheritance-diagram::
taskflow.atom
taskflow.task
taskflow.retry
taskflow.retry.Retry
taskflow.retry.AlwaysRevert
taskflow.retry.AlwaysRevertAll
taskflow.retry.Times
taskflow.retry.ForEach
taskflow.retry.ParameterizedForEach
:parts: 1

View File

@@ -13,6 +13,9 @@ ordereddict
# Python 2->3 compatibility library.
six>=1.7.0
# Enum library made for <= python 3.3
enum34
# Very nice graph library
networkx>=1.8

View File

@@ -10,6 +10,9 @@ pbr>=0.6,!=0.7,<1.0
# Python 2->3 compatibility library.
six>=1.7.0
# Enum library made for <= python 3.3
enum34
# Very nice graph library
networkx>=1.8

View File

@@ -17,16 +17,38 @@
import abc
import enum
import six
from taskflow import atom
from taskflow import exceptions as exc
from taskflow.utils import misc
# Decision results.
REVERT = "REVERT"
REVERT_ALL = "REVERT_ALL"
RETRY = "RETRY"
@enum.unique
class Decision(misc.StrEnum):
"""Decision results/strategy enumeration."""
REVERT = "REVERT"
"""Reverts only the surrounding/associated subflow.
This strategy first consults the parent atom before reverting the
associated subflow to determine if the parent retry object provides a
different reconciliation strategy (if no parent retry object exists
then reverting will proceed, if one does exist the parent retry may
override this reconciliation strategy with its own).
"""
#: Completely reverts the whole flow.
REVERT_ALL = "REVERT_ALL"
#: Retries the surrounding/associated subflow again.
RETRY = "RETRY"
# Retain these aliases for a number of releases...
REVERT = Decision.REVERT
REVERT_ALL = Decision.REVERT_ALL
RETRY = Decision.RETRY
# Constants passed into revert/execute kwargs.
#
@@ -108,14 +130,16 @@ class Retry(atom.Atom):
This abstract base class is used to inherit from and provide different
strategies that will be activated upon execution failures. Since a retry
object is an atom it may also provide :meth:`.execute` and
:meth:`.revert` methods to alter the inputs of connected atoms (depending
on the desired strategy to be used this can be quite useful).
object is an atom it may also provide :meth:`~taskflow.retry.Retry.execute`
and :meth:`~taskflow.retry.Retry.revert` methods to alter the inputs of
connected atoms (depending on the desired strategy to be used this can be
quite useful).
NOTE(harlowja): the :meth:`.execute` and :meth:`.revert` and
:meth:`.on_failure` will automatically be given a ``history`` parameter,
which contains information about the past decisions and outcomes
that have occurred (if available).
NOTE(harlowja): the :meth:`~taskflow.retry.Retry.execute` and
:meth:`~taskflow.retry.Retry.revert` and
:meth:`~taskflow.retry.Retry.on_failure` will automatically be given
a ``history`` parameter, which contains information about the past
decisions and outcomes that have occurred (if available).
"""
default_provides = None

View File

@@ -26,6 +26,7 @@ import threading
import time
import types
import enum
from oslo_serialization import jsonutils
from oslo_utils import importutils
from oslo_utils import netutils
@@ -57,6 +58,17 @@ _MONOTONIC_LOCATIONS = tuple([
])
class StrEnum(str, enum.Enum):
"""An enumeration that is also a string and can be compared to strings."""
def __new__(cls, *args, **kwargs):
for a in args:
if not isinstance(a, str):
raise TypeError("Enumeration '%s' (%s) is not"
" a string" % (a, type(a).__name__))
return super(StrEnum, cls).__new__(cls, *args, **kwargs)
def find_monotonic(allow_time_time=False):
"""Tries to find a monotonic time providing function (and returns it)."""
for import_str in _MONOTONIC_LOCATIONS: