abc — Abstract Base Classes¶

Source code:Lib/abc.py

This module provides the infrastructure for definingabstract baseclasses (ABCs) in Python, as outlined inPEP 3119;see the PEP for why this was added to Python. (See alsoPEP 3141 and thenumbers module regarding a type hierarchy for numbers based on ABCs.)

Thecollections module has some concrete classes that derive fromABCs; these can, of course, be further derived. In addition, thecollections.abc submodule has some ABCs that can be used to test whethera class or instance provides a particular interface, for example, if it ishashable or if it is amapping.

This module provides the metaclassABCMeta for defining ABCs anda helper classABC to alternatively define ABCs through inheritance:

classabc.ABC¶

A helper class that hasABCMeta as its metaclass. With this class,an abstract base class can be created by simply deriving fromABCavoiding sometimes confusing metaclass usage, for example:

fromabcimportABCclassMyABC(ABC):pass

Note that the type ofABC is stillABCMeta, thereforeinheriting fromABC requires the usual precautions regardingmetaclass usage, as multiple inheritance may lead to metaclass conflicts.One may also define an abstract base class by passing the metaclasskeyword and usingABCMeta directly, for example:

fromabcimportABCMetaclassMyABC(metaclass=ABCMeta):pass

Added in version 3.4.

classabc.ABCMeta¶

Metaclass for defining Abstract Base Classes (ABCs).

Use this metaclass to create an ABC. An ABC can be subclassed directly, andthen acts as a mix-in class. You can also register unrelated concreteclasses (even built-in classes) and unrelated ABCs as «virtual subclasses» –these and their descendants will be considered subclasses of the registeringABC by the built-inissubclass() function, but the registering ABCwon’t show up in their MRO (Method Resolution Order) nor will methodimplementations defined by the registering ABC be callable (not even viasuper()).[1]

Classes created with a metaclass ofABCMeta have the following method:

register(subclass)¶

Registersubclass as a «virtual subclass» of this ABC. Forexample:

fromabcimportABCclassMyABC(ABC):passMyABC.register(tuple)assertissubclass(tuple,MyABC)assertisinstance((),MyABC)

Άλλαξε στην έκδοση 3.3:Returns the registered subclass, to allow usage as a class decorator.

Άλλαξε στην έκδοση 3.4:To detect calls toregister(), you can use theget_cache_token() function.

You can also override this method in an abstract base class:

__subclasshook__(subclass)¶

(Must be defined as a class method.)

Check whethersubclass is considered a subclass of this ABC. This meansthat you can customize the behavior ofissubclass() further without theneed to callregister() on every class you want to consider asubclass of the ABC. (This class method is called from the__subclasscheck__() method of the ABC.)

This method should returnTrue,False orNotImplemented. Ifit returnsTrue, thesubclass is considered a subclass of this ABC.If it returnsFalse, thesubclass is not considered a subclass ofthis ABC, even if it would normally be one. If it returnsNotImplemented, the subclass check is continued with the usualmechanism.

For a demonstration of these concepts, look at this example ABC definition:

classFoo:def__getitem__(self,index):...def__len__(self):...defget_iterator(self):returniter(self)classMyIterable(ABC):@abstractmethoddef__iter__(self):whileFalse:yieldNonedefget_iterator(self):returnself.__iter__()@classmethoddef__subclasshook__(cls,C):ifclsisMyIterable:ifany("__iter__"inB.__dict__forBinC.__mro__):returnTruereturnNotImplementedMyIterable.register(Foo)

The ABCMyIterable defines the standard iterable method,__iter__(), as an abstract method. The implementation givenhere can still be called from subclasses. Theget_iterator() methodis also part of theMyIterable abstract base class, but it does not haveto be overridden in non-abstract derived classes.

The__subclasshook__() class method defined here says that any classthat has an__iter__() method in its__dict__ (or in that of one of its base classes, accessedvia the__mro__ list) is considered aMyIterable too.

Finally, the last line makesFoo a virtual subclass ofMyIterable,even though it does not define an__iter__() method (it usesthe old-style iterable protocol, defined in terms of__len__() and__getitem__()). Note that this will not makeget_iteratoravailable as a method ofFoo, so it is provided separately.

Theabc module also provides the following decorator:

@abc.abstractmethod¶

A decorator indicating abstract methods.

Using this decorator requires that the class’s metaclass isABCMetaor is derived from it. A class that has a metaclass derived fromABCMeta cannot be instantiated unless all of its abstract methodsand properties are overridden. The abstract methods can be called using anyof the normal “super” call mechanisms.abstractmethod() may be usedto declare abstract methods for properties and descriptors.

Dynamically adding abstract methods to a class, or attempting to modify theabstraction status of a method or class once it is created, are onlysupported using theupdate_abstractmethods() function. Theabstractmethod() only affects subclasses derived using regularinheritance; «virtual subclasses» registered with the ABC’sregister() method are not affected.

Whenabstractmethod() is applied in combination with other methoddescriptors, it should be applied as the innermost decorator, as shown inthe following usage examples:

classC(ABC):@abstractmethoddefmy_abstract_method(self,arg1):...@classmethod@abstractmethoddefmy_abstract_classmethod(cls,arg2):...@staticmethod@abstractmethoddefmy_abstract_staticmethod(arg3):...@property@abstractmethoddefmy_abstract_property(self):...@my_abstract_property.setter@abstractmethoddefmy_abstract_property(self,val):...@abstractmethoddef_get_x(self):...@abstractmethoddef_set_x(self,val):...x=property(_get_x,_set_x)

In order to correctly interoperate with the abstract base class machinery,the descriptor must identify itself as abstract using__isabstractmethod__. In general, this attribute should beTrueif any of the methods used to compose the descriptor are abstract. Forexample, Python’s built-inproperty does the equivalent of:

classDescriptor:...@propertydef__isabstractmethod__(self):returnany(getattr(f,'__isabstractmethod__',False)forfin(self._fget,self._fset,self._fdel))

Σημείωση

Unlike Java abstract methods, these abstractmethods may have an implementation. This implementation can becalled via thesuper() mechanism from the class thatoverrides it. This could be useful as an end-point for asuper-call in a framework that uses cooperativemultiple-inheritance.

Theabc module also supports the following legacy decorators:

@abc.abstractclassmethod¶

Added in version 3.2.

Αποσύρθηκε στην έκδοση 3.3:It is now possible to useclassmethod withabstractmethod(), making this decorator redundant.

A subclass of the built-inclassmethod(), indicating an abstractclassmethod. Otherwise it is similar toabstractmethod().

This special case is deprecated, as theclassmethod() decoratoris now correctly identified as abstract when applied to an abstractmethod:

classC(ABC):@classmethod@abstractmethoddefmy_abstract_classmethod(cls,arg):...

@abc.abstractstaticmethod¶

Added in version 3.2.

Αποσύρθηκε στην έκδοση 3.3:It is now possible to usestaticmethod withabstractmethod(), making this decorator redundant.

A subclass of the built-instaticmethod(), indicating an abstractstaticmethod. Otherwise it is similar toabstractmethod().

This special case is deprecated, as thestaticmethod() decoratoris now correctly identified as abstract when applied to an abstractmethod:

classC(ABC):@staticmethod@abstractmethoddefmy_abstract_staticmethod(arg):...

@abc.abstractproperty¶

Αποσύρθηκε στην έκδοση 3.3:It is now possible to useproperty,property.getter(),property.setter() andproperty.deleter() withabstractmethod(), making this decorator redundant.

A subclass of the built-inproperty(), indicating an abstractproperty.

This special case is deprecated, as theproperty() decoratoris now correctly identified as abstract when applied to an abstractmethod:

classC(ABC):@property@abstractmethoddefmy_abstract_property(self):...

The above example defines a read-only property; you can also define aread-write abstract property by appropriately marking one or more of theunderlying methods as abstract:

classC(ABC):@propertydefx(self):...@x.setter@abstractmethoddefx(self,val):...

If only some components are abstract, only those components need to beupdated to create a concrete property in a subclass:

classD(C):@C.x.setterdefx(self,val):...

Theabc module also provides the following functions:

abc.get_cache_token()¶

Returns the current abstract base class cache token.

The token is an opaque object (that supports equality testing) identifyingthe current version of the abstract base class cache for virtual subclasses.The token changes with every call toABCMeta.register() on any ABC.

Added in version 3.4.

abc.update_abstractmethods(cls)¶

A function to recalculate an abstract class’s abstraction status. Thisfunction should be called if a class’s abstract methods have beenimplemented or changed after it was created. Usually, this function shouldbe called from within a class decorator.

Returnscls, to allow usage as a class decorator.

Ifcls is not an instance ofABCMeta, does nothing.

Σημείωση

This function assumes thatcls’s superclasses are already updated.It does not update any subclasses.

Added in version 3.10.

Footnotes

C++ programmers should note that Python’s virtual base classconcept is not the same as C++”s.