Objects are Python’s abstraction for data. All data in a Python programis represented by objects or by relations between objects. (In a sense, and inconformance to Von Neumann’s model of a “stored program computer,” code is alsorepresented by objects.)
Every object has an identity, a type and a value. An object’sidentity neverchanges once it has been created; you may think of it as the object’s address inmemory. The ‘is‘ operator compares the identity of two objects; theid() function returns an integer representing its identity (currentlyimplemented as its address). An object’stype is also unchangeable.[1]An object’s type determines the operations that the object supports (e.g., “doesit have a length?”) and also defines the possible values for objects of thattype. Thetype() function returns an object’s type (which is an objectitself). Thevalue of some objects can change. Objects whose value canchange are said to bemutable; objects whose value is unchangeable once theyare created are calledimmutable. (The value of an immutable container objectthat contains a reference to a mutable object can change when the latter’s valueis changed; however the container is still considered immutable, because thecollection of objects it contains cannot be changed. So, immutability is notstrictly the same as having an unchangeable value, it is more subtle.) Anobject’s mutability is determined by its type; for instance, numbers, stringsand tuples are immutable, while dictionaries and lists are mutable.
Objects are never explicitly destroyed; however, when they become unreachablethey may be garbage-collected. An implementation is allowed to postpone garbagecollection or omit it altogether — it is a matter of implementation qualityhow garbage collection is implemented, as long as no objects are collected thatare still reachable. (Implementation note: the current implementation uses areference-counting scheme with (optional) delayed detection of cyclically linkedgarbage, which collects most objects as soon as they become unreachable, but isnot guaranteed to collect garbage containing circular references. See thedocumentation of thegc module for information on controlling thecollection of cyclic garbage.)
Note that the use of the implementation’s tracing or debugging facilities maykeep objects alive that would normally be collectable. Also note that catchingan exception with a ‘try...except‘ statement may keepobjects alive.
Some objects contain references to “external” resources such as open files orwindows. It is understood that these resources are freed when the object isgarbage-collected, but since garbage collection is not guaranteed to happen,such objects also provide an explicit way to release the external resource,usually aclose() method. Programs are strongly recommended to explicitlyclose such objects. The ‘try...finally‘ statementand the ‘with‘ statement provide convenient ways to do this.
Some objects contain references to other objects; these are calledcontainers.Examples of containers are tuples, lists and dictionaries. The references arepart of a container’s value. In most cases, when we talk about the value of acontainer, we imply the values, not the identities of the contained objects;however, when we talk about the mutability of a container, only the identitiesof the immediately contained objects are implied. So, if an immutable container(like a tuple) contains a reference to a mutable object, its value changes ifthat mutable object is changed.
Types affect almost all aspects of object behavior. Even the importance ofobject identity is affected in some sense: for immutable types, operations thatcompute new values may actually return a reference to any existing object withthe same type and value, while for mutable objects this is not allowed. E.g.,aftera=1;b=1,a andb may or may not refer to the same objectwith the value one, depending on the implementation, but afterc=[];d=[],c andd are guaranteed to refer to two different, unique, newlycreated empty lists. (Note thatc=d=[] assigns the same object to bothc andd.)
Below is a list of the types that are built into Python. Extension modules(written in C, Java, or other languages, depending on the implementation) candefine additional types. Future versions of Python may add types to the typehierarchy (e.g., rational numbers, efficiently stored arrays of integers, etc.),although such additions will often be provided via the standard library instead.
Some of the type descriptions below contain a paragraph listing ‘specialattributes.’ These are attributes that provide access to the implementation andare not intended for general use. Their definition may change in the future.
This type has a single value. There is a single object with this value. Thisobject is accessed through the built-in nameNone. It is used to signify theabsence of a value in many situations, e.g., it is returned from functions thatdon’t explicitly return anything. Its truth value is false.
This type has a single value. There is a single object with this value. Thisobject is accessed through the built-in nameNotImplemented. Numeric methodsand rich comparison methods may return this value if they do not implement theoperation for the operands provided. (The interpreter will then try thereflected operation, or some other fallback, depending on the operator.) Itstruth value is true.
This type has a single value. There is a single object with this value. Thisobject is accessed through the literal... or the built-in nameEllipsis. Its truth value is true.
These are created by numeric literals and returned as results by arithmeticoperators and arithmetic built-in functions. Numeric objects are immutable;once created their value never changes. Python numbers are of course stronglyrelated to mathematical numbers, but subject to the limitations of numericalrepresentation in computers.
Python distinguishes between integers, floating point numbers, and complexnumbers:
These represent elements from the mathematical set of integers (positive andnegative).
There are two types of integers:
Integers (int)
These represent numbers in an unlimited range, subject to available (virtual)memory only. For the purpose of shift and mask operations, a binaryrepresentation is assumed, and negative numbers are represented in a variant of2’s complement which gives the illusion of an infinite string of sign bitsextending to the left.
These represent the truth values False and True. The two objects representingthe values False and True are the only Boolean objects. The Boolean type is asubtype of the integer type, and Boolean values behave like the values 0 and 1,respectively, in almost all contexts, the exception being that when converted toa string, the strings"False" or"True" are returned, respectively.
The rules for integer representation are intended to give the most meaningfulinterpretation of shift and mask operations involving negative integers.
These represent machine-level double precision floating point numbers. You areat the mercy of the underlying machine architecture (and C or Javaimplementation) for the accepted range and handling of overflow. Python does notsupport single-precision floating point numbers; the savings in processor andmemory usage that are usually the reason for using these is dwarfed by theoverhead of using objects in Python, so there is no reason to complicate thelanguage with two kinds of floating point numbers.
These represent complex numbers as a pair of machine-level double precisionfloating point numbers. The same caveats apply as for floating point numbers.The real and imaginary parts of a complex numberz can be retrieved throughthe read-only attributesz.real andz.imag.
These represent finite ordered sets indexed by non-negative numbers. Thebuilt-in functionlen() returns the number of items of a sequence. Whenthe length of a sequence isn, the index set contains the numbers 0, 1,...,n-1. Itemi of sequencea is selected bya[i].
Sequences also support slicing:a[i:j] selects all items with indexk suchthati<=k<j. When used as an expression, a slice is asequence of the same type. This implies that the index set is renumbered sothat it starts at 0.
Some sequences also support “extended slicing” with a third “step” parameter:a[i:j:k] selects all items ofa with indexx wherex=i+n*k,n>=0 andi<=x<j.
Sequences are distinguished according to their mutability:
An object of an immutable sequence type cannot change once it is created. (Ifthe object contains references to other objects, these other objects may bemutable and may be changed; however, the collection of objects directlyreferenced by an immutable object cannot change.)
The following types are immutable sequences:
The items of a string object are Unicode code units. A Unicode codeunit is represented by a string object of one item and can hold eithera 16-bit or 32-bit value representing a Unicode ordinal (the maximumvalue for the ordinal is given insys.maxunicode, and depends onhow Python is configured at compile time). Surrogate pairs may bepresent in the Unicode object, and will be reported as two separateitems. The built-in functionschr() andord() convertbetween code units and nonnegative integers representing the Unicodeordinals as defined in the Unicode Standard 3.0. Conversion from and toother encodings are possible through the string methodencode().
The items of a tuple are arbitrary Python objects. Tuples of two ormore items are formed by comma-separated lists of expressions. A tupleof one item (a ‘singleton’) can be formed by affixing a comma to anexpression (an expression by itself does not create a tuple, sinceparentheses must be usable for grouping of expressions). An emptytuple can be formed by an empty pair of parentheses.
A bytes object is an immutable array. The items are 8-bit bytes,represented by integers in the range 0 <= x < 256. Bytes literals(likeb'abc' and the built-in functionbytes() can be used toconstruct bytes objects. Also, bytes objects can be decoded to stringsvia thedecode() method.
Mutable sequences can be changed after they are created. The subscription andslicing notations can be used as the target of assignment anddel(delete) statements.
There are currently two intrinsic mutable sequence types:
The items of a list are arbitrary Python objects. Lists are formed byplacing a comma-separated list of expressions in square brackets. (Notethat there are no special cases needed to form lists of length 0 or 1.)
A bytearray object is a mutable array. They are created by the built-inbytearray() constructor. Aside from being mutable (and henceunhashable), byte arrays otherwise provide the same interface andfunctionality as immutable bytes objects.
The extension modulearray provides an additional example of amutable sequence type, as does thecollections module.
These represent unordered, finite sets of unique, immutable objects. As such,they cannot be indexed by any subscript. However, they can be iterated over, andthe built-in functionlen() returns the number of items in a set. Commonuses for sets are fast membership testing, removing duplicates from a sequence,and computing mathematical operations such as intersection, union, difference,and symmetric difference.
For set elements, the same immutability rules apply as for dictionary keys. Notethat numeric types obey the normal rules for numeric comparison: if two numberscompare equal (e.g.,1 and1.0), only one of them can be contained in aset.
There are currently two intrinsic set types:
These represent a mutable set. They are created by the built-inset()constructor and can be modified afterwards by several methods, such asadd().
These represent an immutable set. They are created by the built-infrozenset() constructor. As a frozenset is immutable andhashable, it can be used again as an element of another set, or asa dictionary key.
These represent finite sets of objects indexed by arbitrary index sets. Thesubscript notationa[k] selects the item indexed byk from the mappinga; this can be used in expressions and as the target of assignments ordel statements. The built-in functionlen() returns the numberof items in a mapping.
There is currently a single intrinsic mapping type:
These represent finite sets of objects indexed by nearly arbitrary values. Theonly types of values not acceptable as keys are values containing lists ordictionaries or other mutable types that are compared by value rather than byobject identity, the reason being that the efficient implementation ofdictionaries requires a key’s hash value to remain constant. Numeric types usedfor keys obey the normal rules for numeric comparison: if two numbers compareequal (e.g.,1 and1.0) then they can be used interchangeably to indexthe same dictionary entry.
Dictionaries are mutable; they can be created by the{...} notation (seesectionDictionary displays).
The extension modulesdbm.ndbm anddbm.gnu provideadditional examples of mapping types, as does thecollectionsmodule.
These are the types to which the function call operation (see sectionCalls) can be applied:
A user-defined function object is created by a function definition (seesectionFunction definitions). It should be called with an argument listcontaining the same number of items as the function’s formal parameterlist.
Special attributes:
| Attribute | Meaning | |
|---|---|---|
| __doc__ | The function’s documentationstring, orNone ifunavailable | Writable |
| __name__ | The function’s name | Writable |
| __module__ | The name of the module thefunction was defined in, orNone if unavailable. | Writable |
| __defaults__ | A tuple containing defaultargument values for thosearguments that have defaults,orNone if no argumentshave a default value | Writable |
| __code__ | The code object representingthe compiled function body. | Writable |
| __globals__ | A reference to the dictionarythat holds the function’sglobal variables — theglobal namespace of themodule in which the functionwas defined. | Read-only |
| __dict__ | The namespace supportingarbitrary functionattributes. | Writable |
| __closure__ | None or a tuple of cellsthat contain bindings for thefunction’s free variables. | Read-only |
| __annotations__ | A dict containing annotationsof parameters. The keys ofthe dict are the parameternames, or'return' forthe return annotation, ifprovided. | Writable |
| __kwdefaults__ | A dict containing defaultsfor keyword-only parameters. | Writable |
Most of the attributes labelled “Writable” check the type of the assigned value.
Function objects also support getting and setting arbitrary attributes, whichcan be used, for example, to attach metadata to functions. Regular attributedot-notation is used to get and set such attributes.Note that the currentimplementation only supports function attributes on user-defined functions.Function attributes on built-in functions may be supported in the future.
Additional information about a function’s definition can be retrieved from itscode object; see the description of internal types below.
An instance method object combines a class, a class instance and anycallable object (normally a user-defined function).
Special read-only attributes:__self__ is the class instance object,__func__ is the function object;__doc__ is the method’sdocumentation (same as__func__.__doc__);__name__ is themethod name (same as__func__.__name__);__module__ is thename of the module the method was defined in, orNone if unavailable.
Methods also support accessing (but not setting) the arbitrary functionattributes on the underlying function object.
User-defined method objects may be created when getting an attribute of aclass (perhaps via an instance of that class), if that attribute is auser-defined function object or a class method object.
When an instance method object is created by retrieving a user-definedfunction object from a class via one of its instances, its__self__ attribute is the instance, and the method object is saidto be bound. The new method’s__func__ attribute is the originalfunction object.
When a user-defined method object is created by retrieving another methodobject from a class or instance, the behaviour is the same as for afunction object, except that the__func__ attribute of the newinstance is not the original method object but its__func__attribute.
When an instance method object is created by retrieving a class methodobject from a class or instance, its__self__ attribute is theclass itself, and its__func__ attribute is the function objectunderlying the class method.
When an instance method object is called, the underlying function(__func__) is called, inserting the class instance(__self__) in front of the argument list. For instance, whenC is a class which contains a definition for a functionf(), andx is an instance ofC, callingx.f(1) isequivalent to callingC.f(x,1).
When an instance method object is derived from a class method object, the“class instance” stored in__self__ will actually be the classitself, so that calling eitherx.f(1) orC.f(1) is equivalent tocallingf(C,1) wheref is the underlying function.
Note that the transformation from function object to instance methodobject happens each time the attribute is retrieved from the instance. Insome cases, a fruitful optimization is to assign the attribute to a localvariable and call that local variable. Also notice that thistransformation only happens for user-defined functions; other callableobjects (and all non-callable objects) are retrieved withouttransformation. It is also important to note that user-defined functionswhich are attributes of a class instance are not converted to boundmethods; thisonly happens when the function is an attribute of theclass.
A function or method which uses theyield statement (see sectionThe yield statement) is called agenerator function. Such a function, whencalled, always returns an iterator object which can be used to execute thebody of the function: calling the iterator’s__next__() method willcause the function to execute until it provides a value using theyield statement. When the function executes areturn statement or falls off the end, aStopIterationexception is raised and the iterator will have reached the end of the set ofvalues to be returned.
A built-in function object is a wrapper around a C function. Examples ofbuilt-in functions arelen() andmath.sin() (math is astandard built-in module). The number and type of the arguments aredetermined by the C function. Special read-only attributes:__doc__ is the function’s documentation string, orNone ifunavailable;__name__ is the function’s name;__self__ isset toNone (but see the next item);__module__ is the name ofthe module the function was defined in orNone if unavailable.
This is really a different disguise of a built-in function, this time containingan object passed to the C function as an implicit extra argument. An example ofa built-in method isalist.append(), assumingalist is a list object. Inthis case, the special read-only attribute__self__ is set to the objectdenoted bylist.
Modules are imported by theimport statement (see sectionThe import statement). A module object has anamespace implemented by a dictionary object (this is the dictionary referencedby the __globals__ attribute of functions defined in the module). Attributereferences are translated to lookups in this dictionary, e.g.,m.x isequivalent tom.__dict__["x"]. A module object does not contain the codeobject used to initialize the module (since it isn’t needed once theinitialization is done).
Attribute assignment updates the module’s namespace dictionary, e.g.,m.x=1 is equivalent tom.__dict__["x"]=1.
Special read-only attribute:__dict__ is the module’s namespace as adictionary object.
Predefined (writable) attributes:__name__ is the module’s name;__doc__ is the module’s documentation string, orNone ifunavailable;__file__ is the pathname of the file from which the modulewas loaded, if it was loaded from a file. The__file__ attribute is notpresent for C modules that are statically linked into the interpreter; forextension modules loaded dynamically from a shared library, it is the pathnameof the shared library file.
Custon class types are typically created by class definitions (see sectionClass definitions). A class has a namespace implemented by a dictionary object.Class attribute references are translated to lookups in this dictionary, e.g.,C.x is translated toC.__dict__["x"] (although there are a number ofhooks which allow for other means of locating attributes). When the attributename is not found there, the attribute search continues in the base classes.This search of the base classes uses the C3 method resolution order whichbehaves correctly even in the presence of ‘diamond’ inheritance structureswhere there are multiple inheritance paths leading back to a common ancestor.Additional details on the C3 MRO used by Python can be found in thedocumentation accompanying the 2.3 release athttp://www.python.org/download/releases/2.3/mro/.
When a class attribute reference (for classC, say) would yield aclass method object, it is transformed into an instance method object whose__self__ attributes isC. When it would yield a staticmethod object, it is transformed into the object wrapped by the static methodobject. See sectionImplementing Descriptors for another way in which attributesretrieved from a class may differ from those actually contained in its__dict__.
Class attribute assignments update the class’s dictionary, never the dictionaryof a base class.
A class object can be called (see above) to yield a class instance (see below).
Special attributes:__name__ is the class name;__module__ isthe module name in which the class was defined;__dict__ is thedictionary containing the class’s namespace;__bases__ is a tuple(possibly empty or a singleton) containing the base classes, in the order oftheir occurrence in the base class list;__doc__ is the class’sdocumentation string, or None if undefined.
A class instance is created by calling a class object (see above). A classinstance has a namespace implemented as a dictionary which is the first placein which attribute references are searched. When an attribute is not foundthere, and the instance’s class has an attribute by that name, the searchcontinues with the class attributes. If a class attribute is found that is auser-defined function object, it is transformed into an instance methodobject whose__self__ attribute is the instance. Static method andclass method objects are also transformed; see above under “Classes”. SeesectionImplementing Descriptors for another way in which attributes of a classretrieved via its instances may differ from the objects actually stored inthe class’s__dict__. If no class attribute is found, and theobject’s class has a__getattr__() method, that is called to satisfythe lookup.
Attribute assignments and deletions update the instance’s dictionary, never aclass’s dictionary. If the class has a__setattr__() or__delattr__() method, this is called instead of updating the instancedictionary directly.
Class instances can pretend to be numbers, sequences, or mappings if they havemethods with certain special names. See sectionSpecial method names.
Special attributes:__dict__ is the attribute dictionary;__class__ is the instance’s class.
A file object represents an open file. File objects are created by theopen() built-in function, and also byos.popen(),os.fdopen(), and themakefile() method of socket objects (andperhaps by other functions or methods provided by extension modules). Theobjectssys.stdin,sys.stdout andsys.stderr are initialized tofile objects corresponding to the interpreter’s standard input, output anderror streams. SeeFile Objects for complete documentation offile objects.
A few types used internally by the interpreter are exposed to the user. Theirdefinitions may change with future versions of the interpreter, but they arementioned here for completeness.
Code objects representbyte-compiled executable Python code, orbytecode.The difference between a code object and a function object is that the functionobject contains an explicit reference to the function’s globals (the module inwhich it was defined), while a code object contains no context; also the defaultargument values are stored in the function object, not in the code object(because they represent values calculated at run-time). Unlike functionobjects, code objects are immutable and contain no references (directly orindirectly) to mutable objects.
Special read-only attributes:co_name gives the function name;co_argcount is the number of positional arguments (including argumentswith default values);co_nlocals is the number of local variables usedby the function (including arguments);co_varnames is a tuple containingthe names of the local variables (starting with the argument names);co_cellvars is a tuple containing the names of local variables that arereferenced by nested functions;co_freevars is a tuple containing thenames of free variables;co_code is a string representing the sequenceof bytecode instructions;co_consts is a tuple containing the literalsused by the bytecode;co_names is a tuple containing the names used bythe bytecode;co_filename is the filename from which the code wascompiled;co_firstlineno is the first line number of the function;co_lnotab is a string encoding the mapping from bytecode offsets toline numbers (for details see the source code of the interpreter);co_stacksize is the required stack size (including local variables);co_flags is an integer encoding a number of flags for the interpreter.
The following flag bits are defined forco_flags: bit0x04 is set ifthe function uses the*arguments syntax to accept an arbitrary number ofpositional arguments; bit0x08 is set if the function uses the**keywords syntax to accept arbitrary keyword arguments; bit0x20 is setif the function is a generator.
Future feature declarations (from__future__importdivision) also use bitsinco_flags to indicate whether a code object was compiled with aparticular feature enabled: bit0x2000 is set if the function was compiledwith future division enabled; bits0x10 and0x1000 were used in earlierversions of Python.
Other bits inco_flags are reserved for internal use.
If a code object represents a function, the first item inco_consts isthe documentation string of the function, orNone if undefined.
Frame objects represent execution frames. They may occur in traceback objects(see below).
Special read-only attributes:f_back is to the previous stack frame(towards the caller), orNone if this is the bottom stack frame;f_code is the code object being executed in this frame;f_localsis the dictionary used to look up local variables;f_globals is used forglobal variables;f_builtins is used for built-in (intrinsic) names;f_lasti gives the precise instruction (this is an index into thebytecode string of the code object).
Special writable attributes:f_trace, if notNone, is a functioncalled at the start of each source code line (this is used by the debugger);f_lineno is the current line number of the frame — writing to thisfrom within a trace function jumps to the given line (only for the bottom-mostframe). A debugger can implement a Jump command (aka Set Next Statement)by writing to f_lineno.
Traceback objects represent a stack trace of an exception. A traceback objectis created when an exception occurs. When the search for an exception handlerunwinds the execution stack, at each unwound level a traceback object isinserted in front of the current traceback. When an exception handler isentered, the stack trace is made available to the program. (See sectionThe try statement.) It is accessible as the third item of thetuple returned bysys.exc_info(). When the program contains no suitablehandler, the stack trace is written (nicely formatted) to the standard errorstream; if the interpreter is interactive, it is also made available to the userassys.last_traceback.
Special read-only attributes:tb_next is the next level in the stacktrace (towards the frame where the exception occurred), orNone if there isno next level;tb_frame points to the execution frame of the currentlevel;tb_lineno gives the line number where the exception occurred;tb_lasti indicates the precise instruction. The line number and lastinstruction in the traceback may differ from the line number of its frame objectif the exception occurred in atry statement with no matching exceptclause or with a finally clause.
Slice objects are used to represent slices for__getitem__()methods. They are also created by the built-inslice() function.
Special read-only attributes:start is the lower bound;stop isthe upper bound;step is the step value; each isNone if omitted.These attributes can have any type.
Slice objects support one method:
A class can implement certain operations that are invoked by special syntax(such as arithmetic operations or subscripting and slicing) by defining methodswith special names. This is Python’s approach tooperator overloading,allowing classes to define their own behavior with respect to languageoperators. For instance, if a class defines a method named__getitem__(),andx is an instance of this class, thenx[i] is roughly equivalenttotype(x).__getitem__(x,i). Except where mentioned, attempts to execute anoperation raise an exception when no appropriate method is defined (typicallyAttributeError orTypeError).
When implementing a class that emulates any built-in type, it is important thatthe emulation only be implemented to the degree that it makes sense for theobject being modelled. For example, some sequences may work well with retrievalof individual elements, but extracting a slice may not make sense. (One exampleof this is theNodeList interface in the W3C’s Document Object Model.)
Called to create a new instance of classcls.__new__() is a staticmethod (special-cased so you need not declare it as such) that takes the classof which an instance was requested as its first argument. The remainingarguments are those passed to the object constructor expression (the call to theclass). The return value of__new__() should be the new object instance(usually an instance ofcls).
Typical implementations create a new instance of the class by invoking thesuperclass’s__new__() method usingsuper(currentclass,cls).__new__(cls[,...]) with appropriate arguments and then modifying thenewly-created instance as necessary before returning it.
If__new__() returns an instance ofcls, then the new instance’s__init__() method will be invoked like__init__(self[,...]), whereself is the new instance and the remaining arguments are the same as werepassed to__new__().
If__new__() does not return an instance ofcls, then the new instance’s__init__() method will not be invoked.
__new__() is intended mainly to allow subclasses of immutable types (likeint, str, or tuple) to customize instance creation. It is also commonlyoverridden in custom metaclasses in order to customize class creation.
Called when the instance is created. The arguments are those passed to theclass constructor expression. If a base class has an__init__() method,the derived class’s__init__() method, if any, must explicitly call it toensure proper initialization of the base class part of the instance; forexample:BaseClass.__init__(self,[args...]). As a special constraint onconstructors, no value may be returned; doing so will cause aTypeErrorto be raised at runtime.
Called when the instance is about to be destroyed. This is also called adestructor. If a base class has a__del__() method, the derived class’s__del__() method, if any, must explicitly call it to ensure properdeletion of the base class part of the instance. Note that it is possible(though not recommended!) for the__del__() method to postpone destructionof the instance by creating a new reference to it. It may then be called at alater time when this new reference is deleted. It is not guaranteed that__del__() methods are called for objects that still exist when theinterpreter exits.
Note
delx doesn’t directly callx.__del__() — the former decrementsthe reference count forx by one, and the latter is only called whenx‘s reference count reaches zero. Some common situations that mayprevent the reference count of an object from going to zero include:circular references between objects (e.g., a doubly-linked list or a treedata structure with parent and child pointers); a reference to the objecton the stack frame of a function that caught an exception (the tracebackstored insys.exc_info()[2] keeps the stack frame alive); or areference to the object on the stack frame that raised an unhandledexception in interactive mode (the traceback stored insys.last_traceback keeps the stack frame alive). The first situationcan only be remedied by explicitly breaking the cycles; the latter twosituations can be resolved by storingNone insys.last_traceback.Circular references which are garbage are detected when the option cycledetector is enabled (it’s on by default), but can only be cleaned up ifthere are no Python- level__del__() methods involved. Refer to thedocumentation for thegc module for more information about how__del__() methods are handled by the cycle detector, particularlythe description of thegarbage value.
Warning
Due to the precarious circumstances under which__del__() methods areinvoked, exceptions that occur during their execution are ignored, and a warningis printed tosys.stderr instead. Also, when__del__() is invoked inresponse to a module being deleted (e.g., when execution of the program isdone), other globals referenced by the__del__() method may already havebeen deleted or in the process of being torn down (e.g. the importmachinery shutting down). For this reason,__del__() methodsshould do the absoluteminimum needed to maintain external invariants. Starting with version 1.5,Python guarantees that globals whose name begins with a single underscore aredeleted from their module before other globals are deleted; if no otherreferences to such globals exist, this may help in assuring that importedmodules are still available at the time when the__del__() method iscalled.
Called by therepr() built-in function to compute the “official” stringrepresentation of an object. If at all possible, this should look like avalid Python expression that could be used to recreate an object with thesame value (given an appropriate environment). If this is not possible, astring of the form<...someusefuldescription...> should be returned.The return value must be a string object. If a class defines__repr__()but not__str__(), then__repr__() is also used when an“informal” string representation of instances of that class is required.
This is typically used for debugging, so it is important that the representationis information-rich and unambiguous.
Called by thestr() built-in function and by theprint() functionto compute the “informal” string representation of an object. This differsfrom__repr__() in that it does not have to be a valid Pythonexpression: a more convenient or concise representation may be used instead.The return value must be a string object.
Called by theformat() built-in function (and by extension, theformat() method of classstr) to produce a “formatted”string representation of an object. Theformat_spec argument isa string that contains a description of the formatting options desired.The interpretation of theformat_spec argument is up to the typeimplementing__format__(), however most classes will eitherdelegate formatting to one of the built-in types, or use a similarformatting option syntax.
SeeFormat Specification Mini-Language for a description of the standard formatting syntax.
The return value must be a string object.
These are the so-called “rich comparison” methods. The correspondence betweenoperator symbols and method names is as follows:x<y callsx.__lt__(y),x<=y callsx.__le__(y),x==y callsx.__eq__(y),x!=y callsx.__ne__(y),x>y callsx.__gt__(y), andx>=y callsx.__ge__(y).
A rich comparison method may return the singletonNotImplemented if it doesnot implement the operation for a given pair of arguments. By convention,False andTrue are returned for a successful comparison. However, thesemethods can return any value, so if the comparison operator is used in a Booleancontext (e.g., in the condition of anif statement), Python will callbool() on the value to determine if the result is true or false.
There are no implied relationships among the comparison operators. The truthofx==y does not imply thatx!=y is false. Accordingly, whendefining__eq__(), one should also define__ne__() so that theoperators will behave as expected. See the paragraph on__hash__() forsome important notes on creatinghashable objects which supportcustom comparison operations and are usable as dictionary keys.
There are no swapped-argument versions of these methods (to be used when theleft argument does not support the operation but the right argument does);rather,__lt__() and__gt__() are each other’s reflection,__le__() and__ge__() are each other’s reflection, and__eq__() and__ne__() are their own reflection.
Arguments to rich comparison methods are never coerced.
Called by built-in functionhash() and for operations on members ofhashed collections includingset,frozenset, anddict.__hash__() should return an integer. The only requiredproperty is that objects which compare equal have the same hash value; it isadvised to somehow mix together (e.g. using exclusive or) the hash values forthe components of the object that also play a part in comparison of objects.
If a class does not define an__eq__() method it should not define a__hash__() operation either; if it defines__eq__() but not__hash__(), its instances will not be usable as items in hashablecollections. If a class defines mutable objects and implements an__eq__() method, it should not implement__hash__(), since theimplementation of hashable collections requires that a key’s hash value isimmutable (if the object’s hash value changes, it will be in the wrong hashbucket).
User-defined classes have__eq__() and__hash__() methodsby default; with them, all objects compare unequal (except with themselves)andx.__hash__() returnsid(x).
Classes which inherit a__hash__() method from a parent class butchange the meaning of__eq__() such that the hash value returned is nolonger appropriate (e.g. by switching to a value-based concept of equalityinstead of the default identity based equality) can explicitly flagthemselves as being unhashable by setting__hash__=None in the classdefinition. Doing so means that not only will instances of the class raise anappropriateTypeError when a program attempts to retrieve their hashvalue, but they will also be correctly identified as unhashable when checkingisinstance(obj,collections.Hashable) (unlike classes which define theirown__hash__() to explicitly raiseTypeError).
If a class that overrrides__eq__() needs to retain the implementationof__hash__() from a parent class, the interpreter must be told thisexplicitly by setting__hash__=<ParentClass>.__hash__. Otherwise theinheritance of__hash__() will be blocked, just as if__hash__had been explicitly set toNone.
Called to implement truth value testing, and the built-in operationbool();should returnFalse orTrue. When this method is not defined,__len__() is called, if it is defined (see below) andTrue is returnedwhen the length is not zero. If a class defines neither__len__() nor__bool__(), all its instances are considered true.
The following methods can be defined to customize the meaning of attributeaccess (use of, assignment to, or deletion ofx.name) for class instances.
Called when an attribute lookup has not found the attribute in the usual places(i.e. it is not an instance attribute nor is it found in the class tree forself).name is the attribute name. This method should return the(computed) attribute value or raise anAttributeError exception.
Note that if the attribute is found through the normal mechanism,__getattr__() is not called. (This is an intentional asymmetry between__getattr__() and__setattr__().) This is done both for efficiencyreasons and because otherwise__getattr__() would have no way to accessother attributes of the instance. Note that at least for instance variables,you can fake total control by not inserting any values in the instance attributedictionary (but instead inserting them in another object). See the__getattribute__() method below for a way to actually get total controlover attribute access.
Called unconditionally to implement attribute accesses for instances of theclass. If the class also defines__getattr__(), the latter will not becalled unless__getattribute__() either calls it explicitly or raises anAttributeError. This method should return the (computed) attribute valueor raise anAttributeError exception. In order to avoid infiniterecursion in this method, its implementation should always call the base classmethod with the same name to access any attributes it needs, for example,object.__getattribute__(self,name).
Note
This method may still be bypassed when looking up special methods as theresult of implicit invocation via language syntax or builtin functions.SeeSpecial method lookup.
Called when an attribute assignment is attempted. This is called instead ofthe normal mechanism (i.e. store the value in the instance dictionary).name is the attribute name,value is the value to be assigned to it.
If__setattr__() wants to assign to an instance attribute, it shouldcall the base class method with the same name, for example,object.__setattr__(self,name,value).
The following methods only apply when an instance of the class containing themethod (a so-calleddescriptor class) appears in the class dictionary ofanother class, known as theowner class. In the examples below, “theattribute” refers to the attribute whose name is the key of the property in theowner class’__dict__.
In general, a descriptor is an object attribute with “binding behavior”, onewhose attribute access has been overridden by methods in the descriptorprotocol:__get__(),__set__(), and__delete__(). If any ofthose methods are defined for an object, it is said to be a descriptor.
The default behavior for attribute access is to get, set, or delete theattribute from an object’s dictionary. For instance,a.x has a lookup chainstarting witha.__dict__['x'], thentype(a).__dict__['x'], andcontinuing through the base classes oftype(a) excluding metaclasses.
However, if the looked-up value is an object defining one of the descriptormethods, then Python may override the default behavior and invoke the descriptormethod instead. Where this occurs in the precedence chain depends on whichdescriptor methods were defined and how they were called.
The starting point for descriptor invocation is a binding,a.x. How thearguments are assembled depends ona:
For instance bindings, the precedence of descriptor invocation depends on thewhich descriptor methods are defined. Normally, data descriptors define both__get__() and__set__(), while non-data descriptors have just the__get__() method. Data descriptors always override a redefinition in aninstance dictionary. In contrast, non-data descriptors can be overridden byinstances.[2]
Python methods (includingstaticmethod() andclassmethod()) areimplemented as non-data descriptors. Accordingly, instances can redefine andoverride methods. This allows individual instances to acquire behaviors thatdiffer from other instances of the same class.
Theproperty() function is implemented as a data descriptor. Accordingly,instances cannot override the behavior of a property.
By default, instances of classes have a dictionary for attribute storage. Thiswastes space for objects having very few instance variables. The spaceconsumption can become acute when creating large numbers of instances.
The default can be overridden by defining__slots__ in a class definition.The__slots__ declaration takes a sequence of instance variables and reservesjust enough space in each instance to hold a value for each variable. Space issaved because__dict__ is not created for each instance.
By default, classes are constructed usingtype(). A class definition isread into a separate namespace and the value of class name is bound to theresult oftype(name,bases,dict).
When the class definition is read, if a callablemetaclass keyword argumentis passed after the bases in the class definition, the callable given will becalled instead oftype(). If other keyword arguments are passed, theywill also be passed to the metaclass. This allows classes or functions to bewritten which monitor or alter the class creation process:
These steps will have to be performed in the metaclass’s__new__() method–type.__new__() can then be called from this method to create a classwith different properties. This example adds a new element to the classdictionary before creating the class:
class metacls(type): def __new__(mcs, name, bases, dict): dict['foo'] = 'metacls was here' return type.__new__(mcs, name, bases, dict)
You can of course also override other class methods (or add new methods); forexample defining a custom__call__() method in the metaclass allows custombehavior when the class is called, e.g. not always creating a new instance.
If the metaclass has a__prepare__() attribute (usually implemented as aclass or static method), it is called before the class body is evaluated withthe name of the class and a tuple of its bases for arguments. It should returnan object that supports the mapping interface that will be used to store thenamespace of the class. The default is a plain dictionary. This could be used,for example, to keep track of the order that class attributes are declared in byreturning an ordered dictionary.
The appropriate metaclass is determined by the following precedence rules:
The potential uses for metaclasses are boundless. Some ideas that have beenexplored including logging, interface checking, automatic delegation, automaticproperty creation, proxies, frameworks, and automatic resourcelocking/synchronization.
Called when the instance is “called” as a function; if this method is defined,x(arg1,arg2,...) is a shorthand forx.__call__(arg1,arg2,...).
The following methods can be defined to implement container objects. Containersusually are sequences (such as lists or tuples) or mappings (like dictionaries),but can represent other containers as well. The first set of methods is usedeither to emulate a sequence or to emulate a mapping; the difference is that fora sequence, the allowable keys should be the integersk for which0<=k<N whereN is the length of the sequence, or slice objects, which define arange of items. It is also recommended that mappings provide the methodskeys(),values(),items(),get(),clear(),setdefault(),pop(),popitem(),copy(), andupdate() behaving similar to those for Python’s standard dictionaryobjects. Thecollections module provides aMutableMappingabstract base class to help create those methods from a base set of__getitem__(),__setitem__(),__delitem__(), andkeys().Mutable sequences should provide methodsappend(),count(),index(),extend(),insert(),pop(),remove(),reverse() andsort(), like Python standard list objects. Finally,sequence types should implement addition (meaning concatenation) andmultiplication (meaning repetition) by defining the methods__add__(),__radd__(),__iadd__(),__mul__(),__rmul__() and__imul__() described below; they should not define other numericaloperators. It is recommended that both mappings and sequences implement the__contains__() method to allow efficient use of thein operator; formappings,in should search the mapping’s keys; for sequences, it shouldsearch through the values. It is further recommended that both mappings andsequences implement the__iter__() method to allow efficient iterationthrough the container; for mappings,__iter__() should be the same askeys(); for sequences, it should iterate through the values.
Called to implement the built-in functionlen(). Should return the lengthof the object, an integer>= 0. Also, an object that doesn’t define a__bool__() method and whose__len__() method returns zero isconsidered to be false in a Boolean context.
Note
Slicing is done exclusively with the following three methods. A call like
a[1:2]=b
is translated to
a[slice(1,2,None)]=b
and so forth. Missing slice items are always filled in withNone.
Called to implement evaluation ofself[key]. For sequence types, theaccepted keys should be integers and slice objects. Note that the specialinterpretation of negative indexes (if the class wishes to emulate a sequencetype) is up to the__getitem__() method. Ifkey is of an inappropriatetype,TypeError may be raised; if of a value outside the set of indexesfor the sequence (after any special interpretation of negative values),IndexError should be raised. For mapping types, ifkey is missing (notin the container),KeyError should be raised.
Note
for loops expect that anIndexError will be raised for illegalindexes to allow proper detection of the end of the sequence.
This method is called when an iterator is required for a container. This methodshould return a new iterator object that can iterate over all the objects in thecontainer. For mappings, it should iterate over the keys of the container, andshould also be made available as the methodkeys().
Iterator objects also need to implement this method; they are required to returnthemselves. For more information on iterator objects, seeIterator Types.
Called (if present) by thereversed() builtin to implementreverse iteration. It should return a new iterator object that iteratesover all the objects in the container in reverse order.
If the__reversed__() method is not provided, thereversed() builtin will fall back to using the sequence protocol(__len__() and__getitem__()). Objects should normallyonly provide__reversed__() if they do not support the sequenceprotocol and an efficient implementation of reverse iteration is possible.
The membership test operators (in andnotin) are normallyimplemented as an iteration through a sequence. However, container objects cansupply the following special method with a more efficient implementation, whichalso does not require the object be a sequence.
The following methods can be defined to emulate numeric objects. Methodscorresponding to operations that are not supported by the particular kind ofnumber implemented (e.g., bitwise operations for non-integral numbers) should beleft undefined.
These methods are called to implement the binary arithmetic operations (+,-,*,/,//,%,divmod(),pow(),**,<<,>>,&,^,|). For instance, to evaluate the expressionx+y, wherex is an instance of a class that has an__add__()method,x.__add__(y) is called. The__divmod__() method should be theequivalent to using__floordiv__() and__mod__(); it should not berelated to__truediv__(). Note that__pow__() should be definedto accept an optional third argument if the ternary version of the built-inpow() function is to be supported.
If one of those methods does not support the operation with the suppliedarguments, it should returnNotImplemented.
These methods are called to implement the binary arithmetic operations (+,-,*,/,//,%,divmod(),pow(),**,<<,>>,&,^,|) with reflected (swapped) operands.These functions are only called if the left operand does not support thecorresponding operation and the operands are of different types.[3] Forinstance, to evaluate the expressionx-y, wherey is an instance ofa class that has an__rsub__() method,y.__rsub__(x) is called ifx.__sub__(y) returnsNotImplemented.
Note that ternarypow() will not try calling__rpow__() (thecoercion rules would become too complicated).
Note
If the right operand’s type is a subclass of the left operand’s type and thatsubclass provides the reflected method for the operation, this method will becalled before the left operand’s non-reflected method. This behavior allowssubclasses to override their ancestors’ operations.
Called to implement the unary arithmetic operations (-,+,abs()and~).
Called to implement the built-in functionscomplex(),int(),float() andround(). Should return a valueof the appropriate type.
Acontext manager is an object that defines the runtime context to beestablished when executing awith statement. The context managerhandles the entry into, and the exit from, the desired runtime context for theexecution of the block of code. Context managers are normally invoked using thewith statement (described in sectionThe with statement), but can also beused by directly invoking their methods.
Typical uses of context managers include saving and restoring various kinds ofglobal state, locking and unlocking resources, closing opened files, etc.
For more information on context managers, seeContext Manager Types.
Exit the runtime context related to this object. The parameters describe theexception that caused the context to be exited. If the context was exitedwithout an exception, all three arguments will beNone.
If an exception is supplied, and the method wishes to suppress the exception(i.e., prevent it from being propagated), it should return a true value.Otherwise, the exception will be processed normally upon exit from this method.
Note that__exit__() methods should not reraise the passed-in exception;this is the caller’s responsibility.
For custom classes, implicit invocations of special methods are only guaranteedto work correctly if defined on an object’s type, not in the object’s instancedictionary. That behaviour is the reason why the following code raises anexception:
>>>classC(object):...pass...>>>c=C()>>>c.__len__=lambda:5>>>len(c)Traceback (most recent call last): File"<stdin>", line1, in<module>TypeError:object of type 'C' has no len()
The rationale behind this behaviour lies with a number of special methods suchas__hash__() and__repr__() that are implemented by all objects,including type objects. If the implicit lookup of these methods used theconventional lookup process, they would fail when invoked on the type objectitself:
>>>1.__hash__()==hash(1)True>>>int.__hash__()==hash(int)Traceback (most recent call last): File"<stdin>", line1, in<module>TypeError:descriptor '__hash__' of 'int' object needs an argument
Incorrectly attempting to invoke an unbound method of a class in this way issometimes referred to as ‘metaclass confusion’, and is avoided by bypassingthe instance when looking up special methods:
>>>type(1).__hash__(1)==hash(1)True>>>type(int).__hash__(int)==hash(int)True
In addition to bypassing any instance attributes in the interest ofcorrectness, implicit special method lookup generally also bypasses the__getattribute__() method even of the object’s metaclass:
>>>classMeta(type):...def__getattribute__(*args):...print("Metaclass getattribute invoked")...returntype.__getattribute__(*args)...>>>classC(object,metaclass=Meta):...def__len__(self):...return10...def__getattribute__(*args):...print("Class getattribute invoked")...returnobject.__getattribute__(*args)...>>>c=C()>>>c.__len__()# Explicit lookup via instanceClass getattribute invoked10>>>type(c).__len__(c)# Explicit lookup via typeMetaclass getattribute invoked10>>>len(c)# Implicit lookup10
Bypassing the__getattribute__() machinery in this fashionprovides significant scope for speed optimisations within theinterpreter, at the cost of some flexibility in the handling ofspecial methods (the special methodmust be set on the classobject itself in order to be consistently invoked by the interpreter).
Footnotes
| [1] | Itis possible in some cases to change an object’s type, under certaincontrolled conditions. It generally isn’t a good idea though, since it canlead to some very strange behaviour if it is handled incorrectly. |
| [2] | A descriptor can define any combination of__get__(),__set__() and__delete__(). If it does not define__get__(),then accessing the attribute even on an instance will return the descriptorobject itself. If the descriptor defines__set__() and/or__delete__(), it is a data descriptor; if it defines neither, it is anon-data descriptor. |
| [3] | For operands of the same type, it is assumed that if the non-reflected method(such as__add__()) fails the operation is not supported, which is why thereflected method is not called. |
Enter search terms or a module, class or function name.
