onetime¶

References¶

[1] How-To Guide for Descriptors, Raymond Hettinger. http://users.rcn.com/python/download/Descriptor.htm

[2] Python data model, http://docs.python.org/reference/datamodel.html

OneTimeProperty¶

class popeye.onetime.OneTimeProperty(func)¶

Bases: object

A descriptor to make special properties that become normal attributes.

This is meant to be used mostly by the auto_attr decorator in this module.

__init__(func)¶

Create a OneTimeProperty instance.

func : method

The method that will be called the first time to compute a value. Afterwards, the method’s name will be a standard attribute holding the value of this computation.

ResetMixin¶

class popeye.onetime.ResetMixin¶

Bases: object

A Mixin class to add a .reset() method to users of OneTimeProperty.

By default, auto attributes once computed, become static. If they happen to depend on other parts of an object and those parts change, their values may now be invalid.

This class offers a .reset() method that users can call explicitly when they know the state of their objects may have changed and they want to ensure that all their special attributes should be invalidated. Once reset() is called, all their auto attributes are reset to their OneTimeProperty descriptors, and their accessor functions will be triggered again.

Warning

If a class has a set of attributes that are OneTimeProperty, but that can be initialized from any one of them, do NOT use this mixin! For instance, UniformTimeSeries can be initialized with only sampling_rate and t0, sampling_interval and time are auto-computed. But if you were to reset() a UniformTimeSeries, it would lose all 4, and there would be then no way to break the circular dependency chains.

If this becomes a problem in practice (for our analyzer objects it isn’t, as they don’t have the above pattern), we can extend reset() to check for a _no_reset set of names in the instance which are meant to be kept protected. But for now this is NOT done, so caveat emptor.

>>> class A(ResetMixin):
...     def __init__(self,x=1.0):
...         self.x = x
...
...     @auto_attr
...     def y(self):
...         print('*** y computation executed ***')
...         return self.x / 2.0
...

>>> a = A(10)

About to access y twice, the second time no computation is done: >>> a.y * y computation executed * 5.0 >>> a.y 5.0

Changing x >>> a.x = 20

a.y doesn’t change to 10, since it is a static attribute: >>> a.y 5.0

We now reset a, and this will then force all auto attributes to recompute the next time we access them: >>> a.reset()

About to access y twice again after reset(): >>> a.y * y computation executed * 10.0 >>> a.y 10.0

__init__()¶

x.__init__(...) initializes x; see help(type(x)) for signature

reset()¶

Reset all OneTimeProperty attributes that may have fired already.

auto_attr¶

popeye.onetime.auto_attr(func)¶

Decorator to create OneTimeProperty attributes.

func : method

The method that will be called the first time to compute a value. Afterwards, the method’s name will be a standard attribute holding the value of this computation.

>>> class MagicProp(object):
...     @auto_attr
...     def a(self):
...         return 99
...
>>> x = MagicProp()
>>> 'a' in x.__dict__
False
>>> x.a
99
>>> 'a' in x.__dict__
True

setattr_on_read¶

popeye.onetime.setattr_on_read(func)¶

Decorator to create OneTimeProperty attributes.

func : method

The method that will be called the first time to compute a value. Afterwards, the method’s name will be a standard attribute holding the value of this computation.

>>> class MagicProp(object):
...     @auto_attr
...     def a(self):
...         return 99
...
>>> x = MagicProp()
>>> 'a' in x.__dict__
False
>>> x.a
99
>>> 'a' in x.__dict__
True