|
1 :mod:`abc` --- Abstract Base Classes |
|
2 ==================================== |
|
3 |
|
4 .. module:: abc |
|
5 :synopsis: Abstract base classes according to PEP 3119. |
|
6 .. moduleauthor:: Guido van Rossum |
|
7 .. sectionauthor:: Georg Brandl |
|
8 .. much of the content adapted from docstrings |
|
9 |
|
10 .. versionadded:: 2.6 |
|
11 |
|
12 This module provides the infrastructure for defining an :term:`abstract base |
|
13 class` (ABCs) in Python, as outlined in :pep:`3119`; see the PEP for why this |
|
14 was added to Python. (See also :pep:`3141` and the :mod:`numbers` module |
|
15 regarding a type hierarchy for numbers based on ABCs.) |
|
16 |
|
17 The :mod:`collections` module has some concrete classes that derive from |
|
18 ABCs; these can, of course, be further derived. In addition the |
|
19 :mod:`collections` module has some ABCs that can be used to test whether |
|
20 a class or instance provides a particular interface, for example, is it |
|
21 hashable or a mapping. |
|
22 |
|
23 |
|
24 This module provides the following class: |
|
25 |
|
26 .. class:: ABCMeta |
|
27 |
|
28 Metaclass for defining Abstract Base Classes (ABCs). |
|
29 |
|
30 Use this metaclass to create an ABC. An ABC can be subclassed directly, and |
|
31 then acts as a mix-in class. You can also register unrelated concrete |
|
32 classes (even built-in classes) and unrelated ABCs as "virtual subclasses" -- |
|
33 these and their descendants will be considered subclasses of the registering |
|
34 ABC by the built-in :func:`issubclass` function, but the registering ABC |
|
35 won't show up in their MRO (Method Resolution Order) nor will method |
|
36 implementations defined by the registering ABC be callable (not even via |
|
37 :func:`super`). [#]_ |
|
38 |
|
39 Classes created with a metaclass of :class:`ABCMeta` have the following method: |
|
40 |
|
41 .. method:: register(subclass) |
|
42 |
|
43 Register *subclass* as a "virtual subclass" of this ABC. For |
|
44 example:: |
|
45 |
|
46 from abc import ABCMeta |
|
47 |
|
48 class MyABC: |
|
49 __metaclass__ = ABCMeta |
|
50 |
|
51 MyABC.register(tuple) |
|
52 |
|
53 assert issubclass(tuple, MyABC) |
|
54 assert isinstance((), MyABC) |
|
55 |
|
56 You can also override this method in an abstract base class: |
|
57 |
|
58 .. method:: __subclasshook__(subclass) |
|
59 |
|
60 (Must be defined as a class method.) |
|
61 |
|
62 Check whether *subclass* is considered a subclass of this ABC. This means |
|
63 that you can customize the behavior of ``issubclass`` further without the |
|
64 need to call :meth:`register` on every class you want to consider a |
|
65 subclass of the ABC. (This class method is called from the |
|
66 :meth:`__subclasscheck__` method of the ABC.) |
|
67 |
|
68 This method should return ``True``, ``False`` or ``NotImplemented``. If |
|
69 it returns ``True``, the *subclass* is considered a subclass of this ABC. |
|
70 If it returns ``False``, the *subclass* is not considered a subclass of |
|
71 this ABC, even if it would normally be one. If it returns |
|
72 ``NotImplemented``, the subclass check is continued with the usual |
|
73 mechanism. |
|
74 |
|
75 .. XXX explain the "usual mechanism" |
|
76 |
|
77 |
|
78 For a demonstration of these concepts, look at this example ABC definition:: |
|
79 |
|
80 class Foo(object): |
|
81 def __getitem__(self, index): |
|
82 ... |
|
83 def __len__(self): |
|
84 ... |
|
85 def get_iterator(self): |
|
86 return iter(self) |
|
87 |
|
88 class MyIterable: |
|
89 __metaclass__ = ABCMeta |
|
90 |
|
91 @abstractmethod |
|
92 def __iter__(self): |
|
93 while False: |
|
94 yield None |
|
95 |
|
96 def get_iterator(self): |
|
97 return self.__iter__() |
|
98 |
|
99 @classmethod |
|
100 def __subclasshook__(cls, C): |
|
101 if cls is MyIterable: |
|
102 if any("__iter__" in B.__dict__ for B in C.__mro__): |
|
103 return True |
|
104 return NotImplemented |
|
105 |
|
106 MyIterable.register(Foo) |
|
107 |
|
108 The ABC ``MyIterable`` defines the standard iterable method, |
|
109 :meth:`__iter__`, as an abstract method. The implementation given here can |
|
110 still be called from subclasses. The :meth:`get_iterator` method is also |
|
111 part of the ``MyIterable`` abstract base class, but it does not have to be |
|
112 overridden in non-abstract derived classes. |
|
113 |
|
114 The :meth:`__subclasshook__` class method defined here says that any class |
|
115 that has an :meth:`__iter__` method in its :attr:`__dict__` (or in that of |
|
116 one of its base classes, accessed via the :attr:`__mro__` list) is |
|
117 considered a ``MyIterable`` too. |
|
118 |
|
119 Finally, the last line makes ``Foo`` a virtual subclass of ``MyIterable``, |
|
120 even though it does not define an :meth:`__iter__` method (it uses the |
|
121 old-style iterable protocol, defined in terms of :meth:`__len__` and |
|
122 :meth:`__getitem__`). Note that this will not make ``get_iterator`` |
|
123 available as a method of ``Foo``, so it is provided separately. |
|
124 |
|
125 |
|
126 It also provides the following decorators: |
|
127 |
|
128 .. function:: abstractmethod(function) |
|
129 |
|
130 A decorator indicating abstract methods. |
|
131 |
|
132 Using this decorator requires that the class's metaclass is :class:`ABCMeta` or |
|
133 is derived from it. |
|
134 A class that has a metaclass derived from :class:`ABCMeta` |
|
135 cannot be instantiated unless all of its abstract methods and |
|
136 properties are overridden. |
|
137 The abstract methods can be called using any of the the normal 'super' call |
|
138 mechanisms. |
|
139 |
|
140 Dynamically adding abstract methods to a class, or attempting to modify the |
|
141 abstraction status of a method or class once it is created, are not |
|
142 supported. The :func:`abstractmethod` only affects subclasses derived using |
|
143 regular inheritance; "virtual subclasses" registered with the ABC's |
|
144 :meth:`register` method are not affected. |
|
145 |
|
146 Usage:: |
|
147 |
|
148 class C: |
|
149 __metaclass__ = ABCMeta |
|
150 @abstractmethod |
|
151 def my_abstract_method(self, ...): |
|
152 ... |
|
153 |
|
154 .. note:: |
|
155 |
|
156 Unlike C++'s pure virtual functions, or Java abstract methods, these abstract |
|
157 methods may have an implementation. This implementation can be |
|
158 called via the :func:`super` mechanism from the class that |
|
159 overrides it. This could be useful as an end-point for a |
|
160 super-call in a framework that uses cooperative |
|
161 multiple-inheritance. |
|
162 |
|
163 |
|
164 .. function:: abstractproperty(fget[, fset[, fdel[, doc]]]) |
|
165 |
|
166 A subclass of the built-in :func:`property`, indicating an abstract property. |
|
167 |
|
168 Using this function requires that the class's metaclass is :class:`ABCMeta` or |
|
169 is derived from it. |
|
170 A class that has a metaclass derived from :class:`ABCMeta` cannot be |
|
171 instantiated unless all of its abstract methods and properties are overridden. |
|
172 The abstract properties can be called using any of the normal |
|
173 'super' call mechanisms. |
|
174 |
|
175 Usage:: |
|
176 |
|
177 class C: |
|
178 __metaclass__ = ABCMeta |
|
179 @abstractproperty |
|
180 def my_abstract_property(self): |
|
181 ... |
|
182 |
|
183 This defines a read-only property; you can also define a read-write abstract |
|
184 property using the 'long' form of property declaration:: |
|
185 |
|
186 class C: |
|
187 __metaclass__ = ABCMeta |
|
188 def getx(self): ... |
|
189 def setx(self, value): ... |
|
190 x = abstractproperty(getx, setx) |
|
191 |
|
192 .. rubric:: Footnotes |
|
193 |
|
194 .. [#] C++ programmers should note that Python's virtual base class |
|
195 concept is not the same as C++'s. |