Bases: object
Returns whether an abstract method is optional or not.
EXAMPLES:
sage: class AbstractClass:
... @abstract_method
... def required(): pass
...
... @abstract_method(optional = True)
... def optional(): pass
sage: AbstractClass.required.is_optional()
False
sage: AbstractClass.optional.is_optional()
True
Abstract methods
INPUT:
- f – a function
- optional – a boolean; defaults to False
The decorator abstract_method can be used to declare methods that should be implemented by all concrete derived classes. This declaration should typically include documentation for the specification for this method.
The purpose is to enforce a consistent and visual syntax for such declarations. It is used by the Sage categories for automated tests (see Sets.Parent.test_not_implemented).
EXAMPLES:
We create a class with an abstract method:
sage: class A(object):
...
... @abstract_method
... def my_method(self):
... '''
... The method :meth:`my_method` computes my_method
...
... EXAMPLES::
...
... '''
... pass
...
sage: A.my_method
<abstract method my_method at ...>
The current policy is that a NotImplementedError is raised when accessing the method through an instance, even before the method is called:
sage: x = A()
sage: x.my_method
...
NotImplementedError: <abstract method my_method at ...>
It is also possible to mark abstract methods as optional:
sage: class A(object):
...
... @abstract_method(optional = True)
... def my_method(self):
... '''
... The method :meth:`my_method` computes my_method
...
... EXAMPLES::
...
... '''
... pass
...
sage: A.my_method
<optional abstract method my_method at ...>
sage: x = A()
sage: x.my_method
NotImplemented
The official mantra for testing whether an optional abstract method is implemented is:
# Fixme: sage -t complains about indentation below
# sage: if x.my_method is not NotImplemented:
# ... x.my_method()
# ... else:
# ... print "x.my_method not available. Let's use some other trick."
# ...
# x.my_method not available. Let's use some other trick.
..topic: Discussion
The policy details are not yet fixed. The purpose of this first implementation is to let developers experiment with it and give feedback on what’s most practical.
The advantage of the current policy is that attempts at using a non implemented methods are caught as early as possible. On the other hand, one cannot use introspection directly to fetch the documentation:
sage: x.my_method? # todo: not implemented
Instead one needs to do:
sage: A._my_method? # todo: not implemented
This could probably be fixed in sage.misc.sageinspect.
TODO: what should be the recommended mantra for existence testing from the class?
TODO: should extra information appear in the output? The name of the class? That of the super class where the abstract method is defined?
TODO: look for similar decorators on the web, and merge
..topic: Implementation details
Technically, an abstract_method is a non-data descriptor (see Invoking Descriptors in the Python reference manual).
The syntax @abstract_method w.r.t. @abstract_method(optional = True) is achieved by a little trick which we test here:
sage: abstract_method(optional = True)
<function <lambda> at ...>
sage: abstract_method(optional = True)(banner)
<optional abstract method banner at ...>
sage: abstract_method(banner, optional = True)
<optional abstract method banner at ...>
Returns the required and optional abstract methods of the class
EXAMPLES:
sage: class AbstractClass:
... @abstract_method
... def required1(): pass
...
... @abstract_method(optional = True)
... def optional2(): pass
...
... @abstract_method(optional = True)
... def optional1(): pass
...
... @abstract_method
... def required2(): pass
...
sage: sage.misc.abstract_method.abstract_methods_of_class(AbstractClass)
{'required': ['required1', 'required2'], 'optional': ['optional1', 'optional2']}