Recollection is a state recalling system. It allows for state of objects to be snap-shotted exposing functionality to then 'rollback' to any previously stored state.
Recollection gives two distinctly different exposures of this mechanism. If performance is not the most critical concern and your focus is on code simplicity then the inheritence approach is possibly your best option providing you have ownership of the classes which need historical state.
Note: This is currently pre-release
You can install this using pip:
pip install recollection
Alternatively you can get the source from: https://github.com/mikemalinowski/recollection
This example shows how to setup a class using inheritence to automatically handle state storing. As you can see from the example, there is no need to explicitly ask recollection to store at any time as it is handled entirely for you. This example specifically shows attribute storing :
import recollection
# -- Inference is the recollection class designed specifically
# -- for inheritance situations.
class Foo(recollection.Inference):
def __init__(self):
super(Foo, self).__init__()
self.number = 10
# -- Register 'number' as a property to monitor
self.memento.register('number')
# -- Instance our object
foo = Foo()
# -- Here we demonstrate directly changing properties
# -- which are registered
foo.number = 5
foo.number = 99
print(foo.number == 99)
# -- Now restore back one step
foo.recollection.restore(1)
print(foo.number == 5)
Whilst this example shows how decorators can be utilised to store method getters and setters:
import recollection
# -- Inference is the recollection class designed specifically
# -- for inheritence situations.
class Foo(recollection.Inference):
def __init__(self):
super(Foo, self).__init__()
self._number = 10
# -- Declare that this is a recollection getter
@recollection.infer.get('number')
def number(self):
return self._number
# -- Declare this as a recollection setter
@recollection.infer.store('number')
def set_number(self, value):
self._number = value
# -- Instance our object
foo = Foo()
# -- Update our variable using our accessor
for number in range(10):
foo.set_number(number)
# -- Demonstrate that the current state is 'e'
print(foo.number()) # -- Prints 9
# -- Roll back one step in the memento history
foo.recollection.restore(1)
print(foo.number()) # -- Prints 8
However, we do not always have the luxury of changing class inheritance or you may specifically want to keep the recollection state management out of your actual inheritance hierarchy. The following examples all demonstrate how this can be achieved.
In this example we have a class with two properties. We the instance a Memento class targeting our foo instance. Each time we call the store method within Memento we are taking a snapshot of the values returned by the registered properties/functions
import recollection
class Foo(object):
def __init__(self):
self.number = 0
# -- Instance our object
foo = Foo()
# -- Instance a memento object pointing at foo
memento = recollection.Memento(foo)
memento.register('number')
# -- Start changing some values on foo, and
# -- ask our stack to store those changes
for number in range(11):
foo.number = number
# -- Ask the memento object to store the state
memento.store()
# -- Printing i, shows us 10
print(foo.number)
# -- But lets say we roll back to the state 5 versions
# -- ago
memento.restore(5)
# -- Now we can see i is at the version it was when
# -- it was stored 5 versions back
print(foo.number)
It also allows multiple Memento objects to be put into a lock-step, such that whenever one memento object is storing or restoring then all other memento objects in that sync group will also store or restore.
import recollection
class Foo(object):
def __init__(self):
self.number = 0
# -- This time we instance two completely seperate
# -- foo objects
foo_a = Foo()
foo_b = Foo()
# -- Instance a memento stack for each
memento_a = recollection.Memento(foo_a)
memento_b = recollection.Memento(foo_b)
memento_a.register('number')
memento_b.register('number')
# -- Now we will put our stacks into a state of lock-step
# -- which means whenever one of them is stored or restored
# -- all others in the lock-step group will have the same
# -- action performed
memento_a.group(memento_b)
# -- Increment some values on both objects
for i in range(11):
foo_a.i = i
foo_b.i = i
# -- Trigger a store on only one stack
memento_a.store()
# -- We can see that both A and B have a value of 10
print(foo_a.i == 10 and foo_b.i == 10)
# -- Now we rollback - knowing that this action will occur
# -- across all grouped memento objects
memento_a.restore(5)
# -- Now we can see i is at the version it was when
# -- it was stored 5 versions back
print(foo_a.i == 5 and foo_b.i == 5)
Serialisers can also be registered against memento instances allowing the stored state of a memento object to be serialised into a persistent state.
This example shows how we might define a user preferences class, and within that class we define a memento object to store the preference state. By registering a serialiser the preferences state will be written to disk whenever the 'store' is called.
Notice that in this example we're also choosing not to store private member variables, but instead we're harnessing the public api of the class as getters and setters.
class UserPreferences(object):
def __init__(self):
# -- Visual preferences
self._theme = 'default'
# -- Define our memento, which we utilise specifically to
# -- store our preferences to a persistent location
self._memento = recollection.Memento(self)
# -- We will utilise the JSON Appdata serialiser, which
# -- writes our memento information to the app data system
# -- location
self._memento.register_serialiser(
serialiser=recollection.JsonAppSerialiser,
identifier='memento/demos/userprefs/UserPreferenceA',
)
# -- Register which properties we want the store to focus on
self._memento.register(
label='theme',
getter=self.get_theme,
setter=self.set_theme,
)
# -- Finally, we deserialise - which will update this class
# -- with any previously stored state
self._memento.deserialise()
# --------------------------------------------------------------
def get_theme(self):
return self._theme
# --------------------------------------------------------------
def set_theme(self, theme):
self._theme = theme
self._memento.store(serialise=True)
Equally, if we want to make it a little more obvious at the class level which functions are storing we could opt to utilise the Memento decorator, which stores and serialises:
class UserPreferences(object):
@recollection.serialise_after('theme')
def set_theme(self, theme):
self._theme = theme
These mechanics are all demonstrated in the example modules, specifically:
This demo shows a user preferences object being interacted which which works in the same way as the example above, where the settings are stored as changed come in - allowing the preferences to be 'undone'.
# -- This demo shows a user preferences object being interacted
# -- which which works in the same way as the example above.
from recollection.examples.userprefs.demo import demo
demo()
This demo is identical to the demo above in terms of output but is handled through decorators.
from recollection.examples.userprefs.demo import demo2
demo2()
This demo utilises a 'boardgame' style scenario where we're given two players and the desire to 'undo' the results of turns if they are not desirable!
from recollection.examples.boardgame.game import demo
demo()
This demo shows the utilisation of a setter which is actually setting multiple recorded attributes but wants to have a single restore step.
from recollection.examples.pins.demo import demo
demo()
This demo shows how recollection can be used to store state of a functional object which is represented as a visual tool (utilising PySide2).
from recollection.examples.renamer.demo import demo
demo()
There are currently unittests which cover most of Memento's core, but it is not yet exhaustive.
This has been tested under Python 2.7.13 and Python 3.6.6 on both Ubuntu and Windows.
If you would like to contribute thoughts, ideas, fixes or features please get in touch! [email protected]