1 2 .. _datamodel: 3 4 ********** 5 Data model 6 ********** 7 8 9 .. _objects: 10 11 Objects, values and types 12 ========================= 13 14 .. index:: 15 single: object 16 single: data 17 18 :dfn:`Objects` are Python's abstraction for data. All data in a Python program 19 is represented by objects or by relations between objects. (In a sense, and in 20 conformance to Von Neumann's model of a "stored program computer," code is also 21 represented by objects.) 22 23 .. index:: 24 builtin: id 25 builtin: type 26 single: identity of an object 27 single: value of an object 28 single: type of an object 29 single: mutable object 30 single: immutable object 31 32 .. XXX it *is* now possible in some cases to change an object's 33 type, under certain controlled conditions 34 35 Every object has an identity, a type and a value. An object's *identity* never 36 changes once it has been created; you may think of it as the object's address in 37 memory. The ':keyword:`is`' operator compares the identity of two objects; the 38 :func:`id` function returns an integer representing its identity. 39 40 .. impl-detail:: 41 42 For CPython, ``id(x)`` is the memory address where ``x`` is stored. 43 44 An object's type determines the operations that the object supports (e.g., "does 45 it have a length?") and also defines the possible values for objects of that 46 type. The :func:`type` function returns an object's type (which is an object 47 itself). Like its identity, an object's :dfn:`type` is also unchangeable. 48 [#]_ 49 50 The *value* of some objects can change. Objects whose value can 51 change are said to be *mutable*; objects whose value is unchangeable once they 52 are created are called *immutable*. (The value of an immutable container object 53 that contains a reference to a mutable object can change when the latter's value 54 is changed; however the container is still considered immutable, because the 55 collection of objects it contains cannot be changed. So, immutability is not 56 strictly the same as having an unchangeable value, it is more subtle.) An 57 object's mutability is determined by its type; for instance, numbers, strings 58 and tuples are immutable, while dictionaries and lists are mutable. 59 60 .. index:: 61 single: garbage collection 62 single: reference counting 63 single: unreachable object 64 65 Objects are never explicitly destroyed; however, when they become unreachable 66 they may be garbage-collected. An implementation is allowed to postpone garbage 67 collection or omit it altogether --- it is a matter of implementation quality 68 how garbage collection is implemented, as long as no objects are collected that 69 are still reachable. 70 71 .. impl-detail:: 72 73 CPython currently uses a reference-counting scheme with (optional) delayed 74 detection of cyclically linked garbage, which collects most objects as soon 75 as they become unreachable, but is not guaranteed to collect garbage 76 containing circular references. See the documentation of the :mod:`gc` 77 module for information on controlling the collection of cyclic garbage. 78 Other implementations act differently and CPython may change. 79 Do not depend on immediate finalization of objects when they become 80 unreachable (so you should always close files explicitly). 81 82 Note that the use of the implementation's tracing or debugging facilities may 83 keep objects alive that would normally be collectable. Also note that catching 84 an exception with a ':keyword:`try`...\ :keyword:`except`' statement may keep 85 objects alive. 86 87 Some objects contain references to "external" resources such as open files or 88 windows. It is understood that these resources are freed when the object is 89 garbage-collected, but since garbage collection is not guaranteed to happen, 90 such objects also provide an explicit way to release the external resource, 91 usually a :meth:`close` method. Programs are strongly recommended to explicitly 92 close such objects. The ':keyword:`try`...\ :keyword:`finally`' statement 93 and the ':keyword:`with`' statement provide convenient ways to do this. 94 95 .. index:: single: container 96 97 Some objects contain references to other objects; these are called *containers*. 98 Examples of containers are tuples, lists and dictionaries. The references are 99 part of a container's value. In most cases, when we talk about the value of a 100 container, we imply the values, not the identities of the contained objects; 101 however, when we talk about the mutability of a container, only the identities 102 of the immediately contained objects are implied. So, if an immutable container 103 (like a tuple) contains a reference to a mutable object, its value changes if 104 that mutable object is changed. 105 106 Types affect almost all aspects of object behavior. Even the importance of 107 object identity is affected in some sense: for immutable types, operations that 108 compute new values may actually return a reference to any existing object with 109 the same type and value, while for mutable objects this is not allowed. E.g., 110 after ``a = 1; b = 1``, ``a`` and ``b`` may or may not refer to the same object 111 with the value one, depending on the implementation, but after ``c = []; d = 112 []``, ``c`` and ``d`` are guaranteed to refer to two different, unique, newly 113 created empty lists. (Note that ``c = d = []`` assigns the same object to both 114 ``c`` and ``d``.) 115 116 117 .. _types: 118 119 The standard type hierarchy 120 =========================== 121 122 .. index:: 123 single: type 124 pair: data; type 125 pair: type; hierarchy 126 pair: extension; module 127 pair: C; language 128 129 Below is a list of the types that are built into Python. Extension modules 130 (written in C, Java, or other languages, depending on the implementation) can 131 define additional types. Future versions of Python may add types to the type 132 hierarchy (e.g., rational numbers, efficiently stored arrays of integers, etc.), 133 although such additions will often be provided via the standard library instead. 134 135 .. index:: 136 single: attribute 137 pair: special; attribute 138 triple: generic; special; attribute 139 140 Some of the type descriptions below contain a paragraph listing 'special 141 attributes.' These are attributes that provide access to the implementation and 142 are not intended for general use. Their definition may change in the future. 143 144 None 145 .. index:: object: None 146 147 This type has a single value. There is a single object with this value. This 148 object is accessed through the built-in name ``None``. It is used to signify the 149 absence of a value in many situations, e.g., it is returned from functions that 150 don't explicitly return anything. Its truth value is false. 151 152 NotImplemented 153 .. index:: object: NotImplemented 154 155 This type has a single value. There is a single object with this value. This 156 object is accessed through the built-in name ``NotImplemented``. Numeric methods 157 and rich comparison methods should return this value if they do not implement the 158 operation for the operands provided. (The interpreter will then try the 159 reflected operation, or some other fallback, depending on the operator.) Its 160 truth value is true. 161 162 See 163 :ref:`implementing-the-arithmetic-operations` 164 for more details. 165 166 167 Ellipsis 168 .. index:: object: Ellipsis 169 170 This type has a single value. There is a single object with this value. This 171 object is accessed through the literal ``...`` or the built-in name 172 ``Ellipsis``. Its truth value is true. 173 174 :class:`numbers.Number` 175 .. index:: object: numeric 176 177 These are created by numeric literals and returned as results by arithmetic 178 operators and arithmetic built-in functions. Numeric objects are immutable; 179 once created their value never changes. Python numbers are of course strongly 180 related to mathematical numbers, but subject to the limitations of numerical 181 representation in computers. 182 183 Python distinguishes between integers, floating point numbers, and complex 184 numbers: 185 186 :class:`numbers.Integral` 187 .. index:: object: integer 188 189 These represent elements from the mathematical set of integers (positive and 190 negative). 191 192 There are two types of integers: 193 194 Integers (:class:`int`) 195 196 These represent numbers in an unlimited range, subject to available (virtual) 197 memory only. For the purpose of shift and mask operations, a binary 198 representation is assumed, and negative numbers are represented in a variant of 199 2's complement which gives the illusion of an infinite string of sign bits 200 extending to the left. 201 202 Booleans (:class:`bool`) 203 .. index:: 204 object: Boolean 205 single: False 206 single: True 207 208 These represent the truth values False and True. The two objects representing 209 the values ``False`` and ``True`` are the only Boolean objects. The Boolean type is a 210 subtype of the integer type, and Boolean values behave like the values 0 and 1, 211 respectively, in almost all contexts, the exception being that when converted to 212 a string, the strings ``"False"`` or ``"True"`` are returned, respectively. 213 214 .. index:: pair: integer; representation 215 216 The rules for integer representation are intended to give the most meaningful 217 interpretation of shift and mask operations involving negative integers. 218 219 :class:`numbers.Real` (:class:`float`) 220 .. index:: 221 object: floating point 222 pair: floating point; number 223 pair: C; language 224 pair: Java; language 225 226 These represent machine-level double precision floating point numbers. You are 227 at the mercy of the underlying machine architecture (and C or Java 228 implementation) for the accepted range and handling of overflow. Python does not 229 support single-precision floating point numbers; the savings in processor and 230 memory usage that are usually the reason for using these are dwarfed by the 231 overhead of using objects in Python, so there is no reason to complicate the 232 language with two kinds of floating point numbers. 233 234 :class:`numbers.Complex` (:class:`complex`) 235 .. index:: 236 object: complex 237 pair: complex; number 238 239 These represent complex numbers as a pair of machine-level double precision 240 floating point numbers. The same caveats apply as for floating point numbers. 241 The real and imaginary parts of a complex number ``z`` can be retrieved through 242 the read-only attributes ``z.real`` and ``z.imag``. 243 244 Sequences 245 .. index:: 246 builtin: len 247 object: sequence 248 single: index operation 249 single: item selection 250 single: subscription 251 252 These represent finite ordered sets indexed by non-negative numbers. The 253 built-in function :func:`len` returns the number of items of a sequence. When 254 the length of a sequence is *n*, the index set contains the numbers 0, 1, 255 ..., *n*-1. Item *i* of sequence *a* is selected by ``a[i]``. 256 257 .. index:: single: slicing 258 259 Sequences also support slicing: ``a[i:j]`` selects all items with index *k* such 260 that *i* ``<=`` *k* ``<`` *j*. When used as an expression, a slice is a 261 sequence of the same type. This implies that the index set is renumbered so 262 that it starts at 0. 263 264 Some sequences also support "extended slicing" with a third "step" parameter: 265 ``a[i:j:k]`` selects all items of *a* with index *x* where ``x = i + n*k``, *n* 266 ``>=`` ``0`` and *i* ``<=`` *x* ``<`` *j*. 267 268 Sequences are distinguished according to their mutability: 269 270 Immutable sequences 271 .. index:: 272 object: immutable sequence 273 object: immutable 274 275 An object of an immutable sequence type cannot change once it is created. (If 276 the object contains references to other objects, these other objects may be 277 mutable and may be changed; however, the collection of objects directly 278 referenced by an immutable object cannot change.) 279 280 The following types are immutable sequences: 281 282 .. index:: 283 single: string; immutable sequences 284 285 Strings 286 .. index:: 287 builtin: chr 288 builtin: ord 289 single: character 290 single: integer 291 single: Unicode 292 293 A string is a sequence of values that represent Unicode code points. 294 All the code points in the range ``U+0000 - U+10FFFF`` can be 295 represented in a string. Python doesn't have a :c:type:`char` type; 296 instead, every code point in the string is represented as a string 297 object with length ``1``. The built-in function :func:`ord` 298 converts a code point from its string form to an integer in the 299 range ``0 - 10FFFF``; :func:`chr` converts an integer in the range 300 ``0 - 10FFFF`` to the corresponding length ``1`` string object. 301 :meth:`str.encode` can be used to convert a :class:`str` to 302 :class:`bytes` using the given text encoding, and 303 :meth:`bytes.decode` can be used to achieve the opposite. 304 305 Tuples 306 .. index:: 307 object: tuple 308 pair: singleton; tuple 309 pair: empty; tuple 310 311 The items of a tuple are arbitrary Python objects. Tuples of two or 312 more items are formed by comma-separated lists of expressions. A tuple 313 of one item (a 'singleton') can be formed by affixing a comma to an 314 expression (an expression by itself does not create a tuple, since 315 parentheses must be usable for grouping of expressions). An empty 316 tuple can be formed by an empty pair of parentheses. 317 318 Bytes 319 .. index:: bytes, byte 320 321 A bytes object is an immutable array. The items are 8-bit bytes, 322 represented by integers in the range 0 <= x < 256. Bytes literals 323 (like ``b'abc'``) and the built-in function :func:`bytes` can be used to 324 construct bytes objects. Also, bytes objects can be decoded to strings 325 via the :meth:`~bytes.decode` method. 326 327 Mutable sequences 328 .. index:: 329 object: mutable sequence 330 object: mutable 331 pair: assignment; statement 332 single: subscription 333 single: slicing 334 335 Mutable sequences can be changed after they are created. The subscription and 336 slicing notations can be used as the target of assignment and :keyword:`del` 337 (delete) statements. 338 339 There are currently two intrinsic mutable sequence types: 340 341 Lists 342 .. index:: object: list 343 344 The items of a list are arbitrary Python objects. Lists are formed by 345 placing a comma-separated list of expressions in square brackets. (Note 346 that there are no special cases needed to form lists of length 0 or 1.) 347 348 Byte Arrays 349 .. index:: bytearray 350 351 A bytearray object is a mutable array. They are created by the built-in 352 :func:`bytearray` constructor. Aside from being mutable (and hence 353 unhashable), byte arrays otherwise provide the same interface and 354 functionality as immutable bytes objects. 355 356 .. index:: module: array 357 358 The extension module :mod:`array` provides an additional example of a 359 mutable sequence type, as does the :mod:`collections` module. 360 361 Set types 362 .. index:: 363 builtin: len 364 object: set type 365 366 These represent unordered, finite sets of unique, immutable objects. As such, 367 they cannot be indexed by any subscript. However, they can be iterated over, and 368 the built-in function :func:`len` returns the number of items in a set. Common 369 uses for sets are fast membership testing, removing duplicates from a sequence, 370 and computing mathematical operations such as intersection, union, difference, 371 and symmetric difference. 372 373 For set elements, the same immutability rules apply as for dictionary keys. Note 374 that numeric types obey the normal rules for numeric comparison: if two numbers 375 compare equal (e.g., ``1`` and ``1.0``), only one of them can be contained in a 376 set. 377 378 There are currently two intrinsic set types: 379 380 Sets 381 .. index:: object: set 382 383 These represent a mutable set. They are created by the built-in :func:`set` 384 constructor and can be modified afterwards by several methods, such as 385 :meth:`~set.add`. 386 387 Frozen sets 388 .. index:: object: frozenset 389 390 These represent an immutable set. They are created by the built-in 391 :func:`frozenset` constructor. As a frozenset is immutable and 392 :term:`hashable`, it can be used again as an element of another set, or as 393 a dictionary key. 394 395 Mappings 396 .. index:: 397 builtin: len 398 single: subscription 399 object: mapping 400 401 These represent finite sets of objects indexed by arbitrary index sets. The 402 subscript notation ``a[k]`` selects the item indexed by ``k`` from the mapping 403 ``a``; this can be used in expressions and as the target of assignments or 404 :keyword:`del` statements. The built-in function :func:`len` returns the number 405 of items in a mapping. 406 407 There is currently a single intrinsic mapping type: 408 409 Dictionaries 410 .. index:: object: dictionary 411 412 These represent finite sets of objects indexed by nearly arbitrary values. The 413 only types of values not acceptable as keys are values containing lists or 414 dictionaries or other mutable types that are compared by value rather than by 415 object identity, the reason being that the efficient implementation of 416 dictionaries requires a key's hash value to remain constant. Numeric types used 417 for keys obey the normal rules for numeric comparison: if two numbers compare 418 equal (e.g., ``1`` and ``1.0``) then they can be used interchangeably to index 419 the same dictionary entry. 420 421 Dictionaries are mutable; they can be created by the ``{...}`` notation (see 422 section :ref:`dict`). 423 424 .. index:: 425 module: dbm.ndbm 426 module: dbm.gnu 427 428 The extension modules :mod:`dbm.ndbm` and :mod:`dbm.gnu` provide 429 additional examples of mapping types, as does the :mod:`collections` 430 module. 431 432 Callable types 433 .. index:: 434 object: callable 435 pair: function; call 436 single: invocation 437 pair: function; argument 438 439 These are the types to which the function call operation (see section 440 :ref:`calls`) can be applied: 441 442 User-defined functions 443 .. index:: 444 pair: user-defined; function 445 object: function 446 object: user-defined function 447 448 A user-defined function object is created by a function definition (see 449 section :ref:`function`). It should be called with an argument list 450 containing the same number of items as the function's formal parameter 451 list. 452 453 Special attributes: 454 455 .. tabularcolumns:: |l|L|l| 456 457 .. index:: 458 single: __doc__ (function attribute) 459 single: __name__ (function attribute) 460 single: __module__ (function attribute) 461 single: __dict__ (function attribute) 462 single: __defaults__ (function attribute) 463 single: __closure__ (function attribute) 464 single: __code__ (function attribute) 465 single: __globals__ (function attribute) 466 single: __annotations__ (function attribute) 467 single: __kwdefaults__ (function attribute) 468 pair: global; namespace 469 470 +-------------------------+-------------------------------+-----------+ 471 | Attribute | Meaning | | 472 +=========================+===============================+===========+ 473 | :attr:`__doc__` | The function's documentation | Writable | 474 | | string, or ``None`` if | | 475 | | unavailable; not inherited by | | 476 | | subclasses | | 477 +-------------------------+-------------------------------+-----------+ 478 | :attr:`~definition.\ | The function's name | Writable | 479 | __name__` | | | 480 +-------------------------+-------------------------------+-----------+ 481 | :attr:`~definition.\ | The function's | Writable | 482 | __qualname__` | :term:`qualified name` | | 483 | | | | 484 | | .. versionadded:: 3.3 | | 485 +-------------------------+-------------------------------+-----------+ 486 | :attr:`__module__` | The name of the module the | Writable | 487 | | function was defined in, or | | 488 | | ``None`` if unavailable. | | 489 +-------------------------+-------------------------------+-----------+ 490 | :attr:`__defaults__` | A tuple containing default | Writable | 491 | | argument values for those | | 492 | | arguments that have defaults, | | 493 | | or ``None`` if no arguments | | 494 | | have a default value | | 495 +-------------------------+-------------------------------+-----------+ 496 | :attr:`__code__` | The code object representing | Writable | 497 | | the compiled function body. | | 498 +-------------------------+-------------------------------+-----------+ 499 | :attr:`__globals__` | A reference to the dictionary | Read-only | 500 | | that holds the function's | | 501 | | global variables --- the | | 502 | | global namespace of the | | 503 | | module in which the function | | 504 | | was defined. | | 505 +-------------------------+-------------------------------+-----------+ 506 | :attr:`~object.__dict__`| The namespace supporting | Writable | 507 | | arbitrary function | | 508 | | attributes. | | 509 +-------------------------+-------------------------------+-----------+ 510 | :attr:`__closure__` | ``None`` or a tuple of cells | Read-only | 511 | | that contain bindings for the | | 512 | | function's free variables. | | 513 +-------------------------+-------------------------------+-----------+ 514 | :attr:`__annotations__` | A dict containing annotations | Writable | 515 | | of parameters. The keys of | | 516 | | the dict are the parameter | | 517 | | names, and ``'return'`` for | | 518 | | the return annotation, if | | 519 | | provided. | | 520 +-------------------------+-------------------------------+-----------+ 521 | :attr:`__kwdefaults__` | A dict containing defaults | Writable | 522 | | for keyword-only parameters. | | 523 +-------------------------+-------------------------------+-----------+ 524 525 Most of the attributes labelled "Writable" check the type of the assigned value. 526 527 Function objects also support getting and setting arbitrary attributes, which 528 can be used, for example, to attach metadata to functions. Regular attribute 529 dot-notation is used to get and set such attributes. *Note that the current 530 implementation only supports function attributes on user-defined functions. 531 Function attributes on built-in functions may be supported in the future.* 532 533 Additional information about a function's definition can be retrieved from its 534 code object; see the description of internal types below. 535 536 Instance methods 537 .. index:: 538 object: method 539 object: user-defined method 540 pair: user-defined; method 541 542 An instance method object combines a class, a class instance and any 543 callable object (normally a user-defined function). 544 545 .. index:: 546 single: __func__ (method attribute) 547 single: __self__ (method attribute) 548 single: __doc__ (method attribute) 549 single: __name__ (method attribute) 550 single: __module__ (method attribute) 551 552 Special read-only attributes: :attr:`__self__` is the class instance object, 553 :attr:`__func__` is the function object; :attr:`__doc__` is the method's 554 documentation (same as ``__func__.__doc__``); :attr:`~definition.__name__` is the 555 method name (same as ``__func__.__name__``); :attr:`__module__` is the 556 name of the module the method was defined in, or ``None`` if unavailable. 557 558 Methods also support accessing (but not setting) the arbitrary function 559 attributes on the underlying function object. 560 561 User-defined method objects may be created when getting an attribute of a 562 class (perhaps via an instance of that class), if that attribute is a 563 user-defined function object or a class method object. 564 565 When an instance method object is created by retrieving a user-defined 566 function object from a class via one of its instances, its 567 :attr:`__self__` attribute is the instance, and the method object is said 568 to be bound. The new method's :attr:`__func__` attribute is the original 569 function object. 570 571 When a user-defined method object is created by retrieving another method 572 object from a class or instance, the behaviour is the same as for a 573 function object, except that the :attr:`__func__` attribute of the new 574 instance is not the original method object but its :attr:`__func__` 575 attribute. 576 577 When an instance method object is created by retrieving a class method 578 object from a class or instance, its :attr:`__self__` attribute is the 579 class itself, and its :attr:`__func__` attribute is the function object 580 underlying the class method. 581 582 When an instance method object is called, the underlying function 583 (:attr:`__func__`) is called, inserting the class instance 584 (:attr:`__self__`) in front of the argument list. For instance, when 585 :class:`C` is a class which contains a definition for a function 586 :meth:`f`, and ``x`` is an instance of :class:`C`, calling ``x.f(1)`` is 587 equivalent to calling ``C.f(x, 1)``. 588 589 When an instance method object is derived from a class method object, the 590 "class instance" stored in :attr:`__self__` will actually be the class 591 itself, so that calling either ``x.f(1)`` or ``C.f(1)`` is equivalent to 592 calling ``f(C,1)`` where ``f`` is the underlying function. 593 594 Note that the transformation from function object to instance method 595 object happens each time the attribute is retrieved from the instance. In 596 some cases, a fruitful optimization is to assign the attribute to a local 597 variable and call that local variable. Also notice that this 598 transformation only happens for user-defined functions; other callable 599 objects (and all non-callable objects) are retrieved without 600 transformation. It is also important to note that user-defined functions 601 which are attributes of a class instance are not converted to bound 602 methods; this *only* happens when the function is an attribute of the 603 class. 604 605 Generator functions 606 .. index:: 607 single: generator; function 608 single: generator; iterator 609 610 A function or method which uses the :keyword:`yield` statement (see section 611 :ref:`yield`) is called a :dfn:`generator function`. Such a function, when 612 called, always returns an iterator object which can be used to execute the 613 body of the function: calling the iterator's :meth:`iterator.__next__` 614 method will cause the function to execute until it provides a value 615 using the :keyword:`yield` statement. When the function executes a 616 :keyword:`return` statement or falls off the end, a :exc:`StopIteration` 617 exception is raised and the iterator will have reached the end of the set of 618 values to be returned. 619 620 Coroutine functions 621 .. index:: 622 single: coroutine; function 623 624 A function or method which is defined using :keyword:`async def` is called 625 a :dfn:`coroutine function`. Such a function, when called, returns a 626 :term:`coroutine` object. It may contain :keyword:`await` expressions, 627 as well as :keyword:`async with` and :keyword:`async for` statements. See 628 also the :ref:`coroutine-objects` section. 629 630 Asynchronous generator functions 631 .. index:: 632 single: asynchronous generator; function 633 single: asynchronous generator; asynchronous iterator 634 635 A function or method which is defined using :keyword:`async def` and 636 which uses the :keyword:`yield` statement is called a 637 :dfn:`asynchronous generator function`. Such a function, when called, 638 returns an asynchronous iterator object which can be used in an 639 :keyword:`async for` statement to execute the body of the function. 640 641 Calling the asynchronous iterator's :meth:`aiterator.__anext__` method 642 will return an :term:`awaitable` which when awaited 643 will execute until it provides a value using the :keyword:`yield` 644 expression. When the function executes an empty :keyword:`return` 645 statement or falls off the end, a :exc:`StopAsyncIteration` exception 646 is raised and the asynchronous iterator will have reached the end of 647 the set of values to be yielded. 648 649 Built-in functions 650 .. index:: 651 object: built-in function 652 object: function 653 pair: C; language 654 655 A built-in function object is a wrapper around a C function. Examples of 656 built-in functions are :func:`len` and :func:`math.sin` (:mod:`math` is a 657 standard built-in module). The number and type of the arguments are 658 determined by the C function. Special read-only attributes: 659 :attr:`__doc__` is the function's documentation string, or ``None`` if 660 unavailable; :attr:`~definition.__name__` is the function's name; :attr:`__self__` is 661 set to ``None`` (but see the next item); :attr:`__module__` is the name of 662 the module the function was defined in or ``None`` if unavailable. 663 664 Built-in methods 665 .. index:: 666 object: built-in method 667 object: method 668 pair: built-in; method 669 670 This is really a different disguise of a built-in function, this time containing 671 an object passed to the C function as an implicit extra argument. An example of 672 a built-in method is ``alist.append()``, assuming *alist* is a list object. In 673 this case, the special read-only attribute :attr:`__self__` is set to the object 674 denoted by *alist*. 675 676 Classes 677 Classes are callable. These objects normally act as factories for new 678 instances of themselves, but variations are possible for class types that 679 override :meth:`__new__`. The arguments of the call are passed to 680 :meth:`__new__` and, in the typical case, to :meth:`__init__` to 681 initialize the new instance. 682 683 Class Instances 684 Instances of arbitrary classes can be made callable by defining a 685 :meth:`__call__` method in their class. 686 687 688 Modules 689 .. index:: 690 statement: import 691 object: module 692 693 Modules are a basic organizational unit of Python code, and are created by 694 the :ref:`import system <importsystem>` as invoked either by the 695 :keyword:`import` statement (see :keyword:`import`), or by calling 696 functions such as :func:`importlib.import_module` and built-in 697 :func:`__import__`. A module object has a namespace implemented by a 698 dictionary object (this is the dictionary referenced by the ``__globals__`` 699 attribute of functions defined in the module). Attribute references are 700 translated to lookups in this dictionary, e.g., ``m.x`` is equivalent to 701 ``m.__dict__["x"]``. A module object does not contain the code object used 702 to initialize the module (since it isn't needed once the initialization is 703 done). 704 705 Attribute assignment updates the module's namespace dictionary, e.g., 706 ``m.x = 1`` is equivalent to ``m.__dict__["x"] = 1``. 707 708 .. index:: 709 single: __name__ (module attribute) 710 single: __doc__ (module attribute) 711 single: __file__ (module attribute) 712 single: __annotations__ (module attribute) 713 pair: module; namespace 714 715 Predefined (writable) attributes: :attr:`__name__` is the module's name; 716 :attr:`__doc__` is the module's documentation string, or ``None`` if 717 unavailable; :attr:`__annotations__` (optional) is a dictionary containing 718 :term:`variable annotations <variable annotation>` collected during module 719 body execution; :attr:`__file__` is the pathname of the file from which the 720 module was loaded, if it was loaded from a file. The :attr:`__file__` 721 attribute may be missing for certain types of modules, such as C modules 722 that are statically linked into the interpreter; for extension modules 723 loaded dynamically from a shared library, it is the pathname of the shared 724 library file. 725 726 .. index:: single: __dict__ (module attribute) 727 728 Special read-only attribute: :attr:`~object.__dict__` is the module's 729 namespace as a dictionary object. 730 731 .. impl-detail:: 732 733 Because of the way CPython clears module dictionaries, the module 734 dictionary will be cleared when the module falls out of scope even if the 735 dictionary still has live references. To avoid this, copy the dictionary 736 or keep the module around while using its dictionary directly. 737 738 Custom classes 739 Custom class types are typically created by class definitions (see section 740 :ref:`class`). A class has a namespace implemented by a dictionary object. 741 Class attribute references are translated to lookups in this dictionary, e.g., 742 ``C.x`` is translated to ``C.__dict__["x"]`` (although there are a number of 743 hooks which allow for other means of locating attributes). When the attribute 744 name is not found there, the attribute search continues in the base classes. 745 This search of the base classes uses the C3 method resolution order which 746 behaves correctly even in the presence of 'diamond' inheritance structures 747 where there are multiple inheritance paths leading back to a common ancestor. 748 Additional details on the C3 MRO used by Python can be found in the 749 documentation accompanying the 2.3 release at 750 https://www.python.org/download/releases/2.3/mro/. 751 752 .. XXX: Could we add that MRO doc as an appendix to the language ref? 753 754 .. index:: 755 object: class 756 object: class instance 757 object: instance 758 pair: class object; call 759 single: container 760 object: dictionary 761 pair: class; attribute 762 763 When a class attribute reference (for class :class:`C`, say) would yield a 764 class method object, it is transformed into an instance method object whose 765 :attr:`__self__` attributes is :class:`C`. When it would yield a static 766 method object, it is transformed into the object wrapped by the static method 767 object. See section :ref:`descriptors` for another way in which attributes 768 retrieved from a class may differ from those actually contained in its 769 :attr:`~object.__dict__`. 770 771 .. index:: triple: class; attribute; assignment 772 773 Class attribute assignments update the class's dictionary, never the dictionary 774 of a base class. 775 776 .. index:: pair: class object; call 777 778 A class object can be called (see above) to yield a class instance (see below). 779 780 .. index:: 781 single: __name__ (class attribute) 782 single: __module__ (class attribute) 783 single: __dict__ (class attribute) 784 single: __bases__ (class attribute) 785 single: __doc__ (class attribute) 786 single: __annotations__ (class attribute) 787 788 Special attributes: :attr:`~definition.__name__` is the class name; :attr:`__module__` is 789 the module name in which the class was defined; :attr:`~object.__dict__` is the 790 dictionary containing the class's namespace; :attr:`~class.__bases__` is a 791 tuple containing the base classes, in the order of their occurrence in the 792 base class list; :attr:`__doc__` is the class's documentation string, 793 or ``None`` if undefined; :attr:`__annotations__` (optional) is a dictionary 794 containing :term:`variable annotations <variable annotation>` collected during 795 class body execution. 796 797 Class instances 798 .. index:: 799 object: class instance 800 object: instance 801 pair: class; instance 802 pair: class instance; attribute 803 804 A class instance is created by calling a class object (see above). A class 805 instance has a namespace implemented as a dictionary which is the first place 806 in which attribute references are searched. When an attribute is not found 807 there, and the instance's class has an attribute by that name, the search 808 continues with the class attributes. If a class attribute is found that is a 809 user-defined function object, it is transformed into an instance method 810 object whose :attr:`__self__` attribute is the instance. Static method and 811 class method objects are also transformed; see above under "Classes". See 812 section :ref:`descriptors` for another way in which attributes of a class 813 retrieved via its instances may differ from the objects actually stored in 814 the class's :attr:`~object.__dict__`. If no class attribute is found, and the 815 object's class has a :meth:`__getattr__` method, that is called to satisfy 816 the lookup. 817 818 .. index:: triple: class instance; attribute; assignment 819 820 Attribute assignments and deletions update the instance's dictionary, never a 821 class's dictionary. If the class has a :meth:`__setattr__` or 822 :meth:`__delattr__` method, this is called instead of updating the instance 823 dictionary directly. 824 825 .. index:: 826 object: numeric 827 object: sequence 828 object: mapping 829 830 Class instances can pretend to be numbers, sequences, or mappings if they have 831 methods with certain special names. See section :ref:`specialnames`. 832 833 .. index:: 834 single: __dict__ (instance attribute) 835 single: __class__ (instance attribute) 836 837 Special attributes: :attr:`~object.__dict__` is the attribute dictionary; 838 :attr:`~instance.__class__` is the instance's class. 839 840 I/O objects (also known as file objects) 841 .. index:: 842 builtin: open 843 module: io 844 single: popen() (in module os) 845 single: makefile() (socket method) 846 single: sys.stdin 847 single: sys.stdout 848 single: sys.stderr 849 single: stdio 850 single: stdin (in module sys) 851 single: stdout (in module sys) 852 single: stderr (in module sys) 853 854 A :term:`file object` represents an open file. Various shortcuts are 855 available to create file objects: the :func:`open` built-in function, and 856 also :func:`os.popen`, :func:`os.fdopen`, and the 857 :meth:`~socket.socket.makefile` method of socket objects (and perhaps by 858 other functions or methods provided by extension modules). 859 860 The objects ``sys.stdin``, ``sys.stdout`` and ``sys.stderr`` are 861 initialized to file objects corresponding to the interpreter's standard 862 input, output and error streams; they are all open in text mode and 863 therefore follow the interface defined by the :class:`io.TextIOBase` 864 abstract class. 865 866 Internal types 867 .. index:: 868 single: internal type 869 single: types, internal 870 871 A few types used internally by the interpreter are exposed to the user. Their 872 definitions may change with future versions of the interpreter, but they are 873 mentioned here for completeness. 874 875 .. index:: bytecode, object; code, code object 876 877 Code objects 878 Code objects represent *byte-compiled* executable Python code, or :term:`bytecode`. 879 The difference between a code object and a function object is that the function 880 object contains an explicit reference to the function's globals (the module in 881 which it was defined), while a code object contains no context; also the default 882 argument values are stored in the function object, not in the code object 883 (because they represent values calculated at run-time). Unlike function 884 objects, code objects are immutable and contain no references (directly or 885 indirectly) to mutable objects. 886 887 .. index:: 888 single: co_argcount (code object attribute) 889 single: co_code (code object attribute) 890 single: co_consts (code object attribute) 891 single: co_filename (code object attribute) 892 single: co_firstlineno (code object attribute) 893 single: co_flags (code object attribute) 894 single: co_lnotab (code object attribute) 895 single: co_name (code object attribute) 896 single: co_names (code object attribute) 897 single: co_nlocals (code object attribute) 898 single: co_stacksize (code object attribute) 899 single: co_varnames (code object attribute) 900 single: co_cellvars (code object attribute) 901 single: co_freevars (code object attribute) 902 903 Special read-only attributes: :attr:`co_name` gives the function name; 904 :attr:`co_argcount` is the number of positional arguments (including arguments 905 with default values); :attr:`co_nlocals` is the number of local variables used 906 by the function (including arguments); :attr:`co_varnames` is a tuple containing 907 the names of the local variables (starting with the argument names); 908 :attr:`co_cellvars` is a tuple containing the names of local variables that are 909 referenced by nested functions; :attr:`co_freevars` is a tuple containing the 910 names of free variables; :attr:`co_code` is a string representing the sequence 911 of bytecode instructions; :attr:`co_consts` is a tuple containing the literals 912 used by the bytecode; :attr:`co_names` is a tuple containing the names used by 913 the bytecode; :attr:`co_filename` is the filename from which the code was 914 compiled; :attr:`co_firstlineno` is the first line number of the function; 915 :attr:`co_lnotab` is a string encoding the mapping from bytecode offsets to 916 line numbers (for details see the source code of the interpreter); 917 :attr:`co_stacksize` is the required stack size (including local variables); 918 :attr:`co_flags` is an integer encoding a number of flags for the interpreter. 919 920 .. index:: object: generator 921 922 The following flag bits are defined for :attr:`co_flags`: bit ``0x04`` is set if 923 the function uses the ``*arguments`` syntax to accept an arbitrary number of 924 positional arguments; bit ``0x08`` is set if the function uses the 925 ``**keywords`` syntax to accept arbitrary keyword arguments; bit ``0x20`` is set 926 if the function is a generator. 927 928 Future feature declarations (``from __future__ import division``) also use bits 929 in :attr:`co_flags` to indicate whether a code object was compiled with a 930 particular feature enabled: bit ``0x2000`` is set if the function was compiled 931 with future division enabled; bits ``0x10`` and ``0x1000`` were used in earlier 932 versions of Python. 933 934 Other bits in :attr:`co_flags` are reserved for internal use. 935 936 .. index:: single: documentation string 937 938 If a code object represents a function, the first item in :attr:`co_consts` is 939 the documentation string of the function, or ``None`` if undefined. 940 941 .. _frame-objects: 942 943 Frame objects 944 .. index:: object: frame 945 946 Frame objects represent execution frames. They may occur in traceback objects 947 (see below). 948 949 .. index:: 950 single: f_back (frame attribute) 951 single: f_code (frame attribute) 952 single: f_globals (frame attribute) 953 single: f_locals (frame attribute) 954 single: f_lasti (frame attribute) 955 single: f_builtins (frame attribute) 956 957 Special read-only attributes: :attr:`f_back` is to the previous stack frame 958 (towards the caller), or ``None`` if this is the bottom stack frame; 959 :attr:`f_code` is the code object being executed in this frame; :attr:`f_locals` 960 is the dictionary used to look up local variables; :attr:`f_globals` is used for 961 global variables; :attr:`f_builtins` is used for built-in (intrinsic) names; 962 :attr:`f_lasti` gives the precise instruction (this is an index into the 963 bytecode string of the code object). 964 965 .. index:: 966 single: f_trace (frame attribute) 967 single: f_lineno (frame attribute) 968 969 Special writable attributes: :attr:`f_trace`, if not ``None``, is a function 970 called at the start of each source code line (this is used by the debugger); 971 :attr:`f_lineno` is the current line number of the frame --- writing to this 972 from within a trace function jumps to the given line (only for the bottom-most 973 frame). A debugger can implement a Jump command (aka Set Next Statement) 974 by writing to f_lineno. 975 976 Frame objects support one method: 977 978 .. method:: frame.clear() 979 980 This method clears all references to local variables held by the 981 frame. Also, if the frame belonged to a generator, the generator 982 is finalized. This helps break reference cycles involving frame 983 objects (for example when catching an exception and storing its 984 traceback for later use). 985 986 :exc:`RuntimeError` is raised if the frame is currently executing. 987 988 .. versionadded:: 3.4 989 990 Traceback objects 991 .. index:: 992 object: traceback 993 pair: stack; trace 994 pair: exception; handler 995 pair: execution; stack 996 single: exc_info (in module sys) 997 single: last_traceback (in module sys) 998 single: sys.exc_info 999 single: sys.last_traceback 1000 1001 Traceback objects represent a stack trace of an exception. A traceback object 1002 is created when an exception occurs. When the search for an exception handler 1003 unwinds the execution stack, at each unwound level a traceback object is 1004 inserted in front of the current traceback. When an exception handler is 1005 entered, the stack trace is made available to the program. (See section 1006 :ref:`try`.) It is accessible as the third item of the 1007 tuple returned by ``sys.exc_info()``. When the program contains no suitable 1008 handler, the stack trace is written (nicely formatted) to the standard error 1009 stream; if the interpreter is interactive, it is also made available to the user 1010 as ``sys.last_traceback``. 1011 1012 .. index:: 1013 single: tb_next (traceback attribute) 1014 single: tb_frame (traceback attribute) 1015 single: tb_lineno (traceback attribute) 1016 single: tb_lasti (traceback attribute) 1017 statement: try 1018 1019 Special read-only attributes: :attr:`tb_next` is the next level in the stack 1020 trace (towards the frame where the exception occurred), or ``None`` if there is 1021 no next level; :attr:`tb_frame` points to the execution frame of the current 1022 level; :attr:`tb_lineno` gives the line number where the exception occurred; 1023 :attr:`tb_lasti` indicates the precise instruction. The line number and last 1024 instruction in the traceback may differ from the line number of its frame object 1025 if the exception occurred in a :keyword:`try` statement with no matching except 1026 clause or with a finally clause. 1027 1028 Slice objects 1029 .. index:: builtin: slice 1030 1031 Slice objects are used to represent slices for :meth:`__getitem__` 1032 methods. They are also created by the built-in :func:`slice` function. 1033 1034 .. index:: 1035 single: start (slice object attribute) 1036 single: stop (slice object attribute) 1037 single: step (slice object attribute) 1038 1039 Special read-only attributes: :attr:`~slice.start` is the lower bound; 1040 :attr:`~slice.stop` is the upper bound; :attr:`~slice.step` is the step 1041 value; each is ``None`` if omitted. These attributes can have any type. 1042 1043 Slice objects support one method: 1044 1045 .. method:: slice.indices(self, length) 1046 1047 This method takes a single integer argument *length* and computes 1048 information about the slice that the slice object would describe if 1049 applied to a sequence of *length* items. It returns a tuple of three 1050 integers; respectively these are the *start* and *stop* indices and the 1051 *step* or stride length of the slice. Missing or out-of-bounds indices 1052 are handled in a manner consistent with regular slices. 1053 1054 Static method objects 1055 Static method objects provide a way of defeating the transformation of function 1056 objects to method objects described above. A static method object is a wrapper 1057 around any other object, usually a user-defined method object. When a static 1058 method object is retrieved from a class or a class instance, the object actually 1059 returned is the wrapped object, which is not subject to any further 1060 transformation. Static method objects are not themselves callable, although the 1061 objects they wrap usually are. Static method objects are created by the built-in 1062 :func:`staticmethod` constructor. 1063 1064 Class method objects 1065 A class method object, like a static method object, is a wrapper around another 1066 object that alters the way in which that object is retrieved from classes and 1067 class instances. The behaviour of class method objects upon such retrieval is 1068 described above, under "User-defined methods". Class method objects are created 1069 by the built-in :func:`classmethod` constructor. 1070 1071 1072 .. _specialnames: 1073 1074 Special method names 1075 ==================== 1076 1077 .. index:: 1078 pair: operator; overloading 1079 single: __getitem__() (mapping object method) 1080 1081 A class can implement certain operations that are invoked by special syntax 1082 (such as arithmetic operations or subscripting and slicing) by defining methods 1083 with special names. This is Python's approach to :dfn:`operator overloading`, 1084 allowing classes to define their own behavior with respect to language 1085 operators. For instance, if a class defines a method named :meth:`__getitem__`, 1086 and ``x`` is an instance of this class, then ``x[i]`` is roughly equivalent 1087 to ``type(x).__getitem__(x, i)``. Except where mentioned, attempts to execute an 1088 operation raise an exception when no appropriate method is defined (typically 1089 :exc:`AttributeError` or :exc:`TypeError`). 1090 1091 Setting a special method to ``None`` indicates that the corresponding 1092 operation is not available. For example, if a class sets 1093 :meth:`__iter__` to ``None``, the class is not iterable, so calling 1094 :func:`iter` on its instances will raise a :exc:`TypeError` (without 1095 falling back to :meth:`__getitem__`). [#]_ 1096 1097 When implementing a class that emulates any built-in type, it is important that 1098 the emulation only be implemented to the degree that it makes sense for the 1099 object being modelled. For example, some sequences may work well with retrieval 1100 of individual elements, but extracting a slice may not make sense. (One example 1101 of this is the :class:`~xml.dom.NodeList` interface in the W3C's Document 1102 Object Model.) 1103 1104 1105 .. _customization: 1106 1107 Basic customization 1108 ------------------- 1109 1110 .. method:: object.__new__(cls[, ...]) 1111 1112 .. index:: pair: subclassing; immutable types 1113 1114 Called to create a new instance of class *cls*. :meth:`__new__` is a static 1115 method (special-cased so you need not declare it as such) that takes the class 1116 of which an instance was requested as its first argument. The remaining 1117 arguments are those passed to the object constructor expression (the call to the 1118 class). The return value of :meth:`__new__` should be the new object instance 1119 (usually an instance of *cls*). 1120 1121 Typical implementations create a new instance of the class by invoking the 1122 superclass's :meth:`__new__` method using ``super(currentclass, 1123 cls).__new__(cls[, ...])`` with appropriate arguments and then modifying the 1124 newly-created instance as necessary before returning it. 1125 1126 If :meth:`__new__` returns an instance of *cls*, then the new instance's 1127 :meth:`__init__` method will be invoked like ``__init__(self[, ...])``, where 1128 *self* is the new instance and the remaining arguments are the same as were 1129 passed to :meth:`__new__`. 1130 1131 If :meth:`__new__` does not return an instance of *cls*, then the new instance's 1132 :meth:`__init__` method will not be invoked. 1133 1134 :meth:`__new__` is intended mainly to allow subclasses of immutable types (like 1135 int, str, or tuple) to customize instance creation. It is also commonly 1136 overridden in custom metaclasses in order to customize class creation. 1137 1138 1139 .. method:: object.__init__(self[, ...]) 1140 1141 .. index:: pair: class; constructor 1142 1143 Called after the instance has been created (by :meth:`__new__`), but before 1144 it is returned to the caller. The arguments are those passed to the 1145 class constructor expression. If a base class has an :meth:`__init__` 1146 method, the derived class's :meth:`__init__` method, if any, must explicitly 1147 call it to ensure proper initialization of the base class part of the 1148 instance; for example: ``BaseClass.__init__(self, [args...])``. 1149 1150 Because :meth:`__new__` and :meth:`__init__` work together in constructing 1151 objects (:meth:`__new__` to create it, and :meth:`__init__` to customize it), 1152 no non-``None`` value may be returned by :meth:`__init__`; doing so will 1153 cause a :exc:`TypeError` to be raised at runtime. 1154 1155 1156 .. method:: object.__del__(self) 1157 1158 .. index:: 1159 single: destructor 1160 statement: del 1161 1162 Called when the instance is about to be destroyed. This is also called a 1163 destructor. If a base class has a :meth:`__del__` method, the derived class's 1164 :meth:`__del__` method, if any, must explicitly call it to ensure proper 1165 deletion of the base class part of the instance. Note that it is possible 1166 (though not recommended!) for the :meth:`__del__` method to postpone destruction 1167 of the instance by creating a new reference to it. It may then be called at a 1168 later time when this new reference is deleted. It is not guaranteed that 1169 :meth:`__del__` methods are called for objects that still exist when the 1170 interpreter exits. 1171 1172 .. note:: 1173 1174 ``del x`` doesn't directly call ``x.__del__()`` --- the former decrements 1175 the reference count for ``x`` by one, and the latter is only called when 1176 ``x``'s reference count reaches zero. Some common situations that may 1177 prevent the reference count of an object from going to zero include: 1178 circular references between objects (e.g., a doubly-linked list or a tree 1179 data structure with parent and child pointers); a reference to the object 1180 on the stack frame of a function that caught an exception (the traceback 1181 stored in ``sys.exc_info()[2]`` keeps the stack frame alive); or a 1182 reference to the object on the stack frame that raised an unhandled 1183 exception in interactive mode (the traceback stored in 1184 ``sys.last_traceback`` keeps the stack frame alive). The first situation 1185 can only be remedied by explicitly breaking the cycles; the second can be 1186 resolved by freeing the reference to the traceback object when it is no 1187 longer useful, and the third can be resolved by storing ``None`` in 1188 ``sys.last_traceback``. 1189 Circular references which are garbage are detected and cleaned up when 1190 the cyclic garbage collector is enabled (it's on by default). Refer to the 1191 documentation for the :mod:`gc` module for more information about this 1192 topic. 1193 1194 .. warning:: 1195 1196 Due to the precarious circumstances under which :meth:`__del__` methods are 1197 invoked, exceptions that occur during their execution are ignored, and a warning 1198 is printed to ``sys.stderr`` instead. Also, when :meth:`__del__` is invoked in 1199 response to a module being deleted (e.g., when execution of the program is 1200 done), other globals referenced by the :meth:`__del__` method may already have 1201 been deleted or in the process of being torn down (e.g. the import 1202 machinery shutting down). For this reason, :meth:`__del__` methods 1203 should do the absolute 1204 minimum needed to maintain external invariants. Starting with version 1.5, 1205 Python guarantees that globals whose name begins with a single underscore are 1206 deleted from their module before other globals are deleted; if no other 1207 references to such globals exist, this may help in assuring that imported 1208 modules are still available at the time when the :meth:`__del__` method is 1209 called. 1210 1211 .. index:: 1212 single: repr() (built-in function); __repr__() (object method) 1213 1214 1215 .. method:: object.__repr__(self) 1216 1217 Called by the :func:`repr` built-in function to compute the "official" string 1218 representation of an object. If at all possible, this should look like a 1219 valid Python expression that could be used to recreate an object with the 1220 same value (given an appropriate environment). If this is not possible, a 1221 string of the form ``<...some useful description...>`` should be returned. 1222 The return value must be a string object. If a class defines :meth:`__repr__` 1223 but not :meth:`__str__`, then :meth:`__repr__` is also used when an 1224 "informal" string representation of instances of that class is required. 1225 1226 This is typically used for debugging, so it is important that the representation 1227 is information-rich and unambiguous. 1228 1229 .. index:: 1230 single: string; __str__() (object method) 1231 single: format() (built-in function); __str__() (object method) 1232 single: print() (built-in function); __str__() (object method) 1233 1234 1235 .. method:: object.__str__(self) 1236 1237 Called by :func:`str(object) <str>` and the built-in functions 1238 :func:`format` and :func:`print` to compute the "informal" or nicely 1239 printable string representation of an object. The return value must be a 1240 :ref:`string <textseq>` object. 1241 1242 This method differs from :meth:`object.__repr__` in that there is no 1243 expectation that :meth:`__str__` return a valid Python expression: a more 1244 convenient or concise representation can be used. 1245 1246 The default implementation defined by the built-in type :class:`object` 1247 calls :meth:`object.__repr__`. 1248 1249 .. XXX what about subclasses of string? 1250 1251 1252 .. method:: object.__bytes__(self) 1253 1254 .. index:: builtin: bytes 1255 1256 Called by :func:`bytes` to compute a byte-string representation of an 1257 object. This should return a ``bytes`` object. 1258 1259 .. index:: 1260 single: string; __format__() (object method) 1261 pair: string; conversion 1262 builtin: print 1263 1264 1265 .. method:: object.__format__(self, format_spec) 1266 1267 Called by the :func:`format` built-in function, 1268 and by extension, evaluation of :ref:`formatted string literals 1269 <f-strings>` and the :meth:`str.format` method, to produce a "formatted" 1270 string representation of an object. The ``format_spec`` argument is 1271 a string that contains a description of the formatting options desired. 1272 The interpretation of the ``format_spec`` argument is up to the type 1273 implementing :meth:`__format__`, however most classes will either 1274 delegate formatting to one of the built-in types, or use a similar 1275 formatting option syntax. 1276 1277 See :ref:`formatspec` for a description of the standard formatting syntax. 1278 1279 The return value must be a string object. 1280 1281 .. versionchanged:: 3.4 1282 The __format__ method of ``object`` itself raises a :exc:`TypeError` 1283 if passed any non-empty string. 1284 1285 1286 .. _richcmpfuncs: 1287 .. method:: object.__lt__(self, other) 1288 object.__le__(self, other) 1289 object.__eq__(self, other) 1290 object.__ne__(self, other) 1291 object.__gt__(self, other) 1292 object.__ge__(self, other) 1293 1294 .. index:: 1295 single: comparisons 1296 1297 These are the so-called "rich comparison" methods. The correspondence between 1298 operator symbols and method names is as follows: ``x<y`` calls ``x.__lt__(y)``, 1299 ``x<=y`` calls ``x.__le__(y)``, ``x==y`` calls ``x.__eq__(y)``, ``x!=y`` calls 1300 ``x.__ne__(y)``, ``x>y`` calls ``x.__gt__(y)``, and ``x>=y`` calls 1301 ``x.__ge__(y)``. 1302 1303 A rich comparison method may return the singleton ``NotImplemented`` if it does 1304 not implement the operation for a given pair of arguments. By convention, 1305 ``False`` and ``True`` are returned for a successful comparison. However, these 1306 methods can return any value, so if the comparison operator is used in a Boolean 1307 context (e.g., in the condition of an ``if`` statement), Python will call 1308 :func:`bool` on the value to determine if the result is true or false. 1309 1310 By default, :meth:`__ne__` delegates to :meth:`__eq__` and 1311 inverts the result unless it is ``NotImplemented``. There are no other 1312 implied relationships among the comparison operators, for example, 1313 the truth of ``(x<y or x==y)`` does not imply ``x<=y``. 1314 To automatically generate ordering operations from a single root operation, 1315 see :func:`functools.total_ordering`. 1316 1317 See the paragraph on :meth:`__hash__` for 1318 some important notes on creating :term:`hashable` objects which support 1319 custom comparison operations and are usable as dictionary keys. 1320 1321 There are no swapped-argument versions of these methods (to be used when the 1322 left argument does not support the operation but the right argument does); 1323 rather, :meth:`__lt__` and :meth:`__gt__` are each other's reflection, 1324 :meth:`__le__` and :meth:`__ge__` are each other's reflection, and 1325 :meth:`__eq__` and :meth:`__ne__` are their own reflection. 1326 If the operands are of different types, and right operand's type is 1327 a direct or indirect subclass of the left operand's type, 1328 the reflected method of the right operand has priority, otherwise 1329 the left operand's method has priority. Virtual subclassing is 1330 not considered. 1331 1332 .. method:: object.__hash__(self) 1333 1334 .. index:: 1335 object: dictionary 1336 builtin: hash 1337 1338 Called by built-in function :func:`hash` and for operations on members of 1339 hashed collections including :class:`set`, :class:`frozenset`, and 1340 :class:`dict`. :meth:`__hash__` should return an integer. The only required 1341 property is that objects which compare equal have the same hash value; it is 1342 advised to mix together the hash values of the components of the object that 1343 also play a part in comparison of objects by packing them into a tuple and 1344 hashing the tuple. Example:: 1345 1346 def __hash__(self): 1347 return hash((self.name, self.nick, self.color)) 1348 1349 .. note:: 1350 1351 :func:`hash` truncates the value returned from an object's custom 1352 :meth:`__hash__` method to the size of a :c:type:`Py_ssize_t`. This is 1353 typically 8 bytes on 64-bit builds and 4 bytes on 32-bit builds. If an 1354 object's :meth:`__hash__` must interoperate on builds of different bit 1355 sizes, be sure to check the width on all supported builds. An easy way 1356 to do this is with 1357 ``python -c "import sys; print(sys.hash_info.width)"``. 1358 1359 If a class does not define an :meth:`__eq__` method it should not define a 1360 :meth:`__hash__` operation either; if it defines :meth:`__eq__` but not 1361 :meth:`__hash__`, its instances will not be usable as items in hashable 1362 collections. If a class defines mutable objects and implements an 1363 :meth:`__eq__` method, it should not implement :meth:`__hash__`, since the 1364 implementation of hashable collections requires that a key's hash value is 1365 immutable (if the object's hash value changes, it will be in the wrong hash 1366 bucket). 1367 1368 User-defined classes have :meth:`__eq__` and :meth:`__hash__` methods 1369 by default; with them, all objects compare unequal (except with themselves) 1370 and ``x.__hash__()`` returns an appropriate value such that ``x == y`` 1371 implies both that ``x is y`` and ``hash(x) == hash(y)``. 1372 1373 A class that overrides :meth:`__eq__` and does not define :meth:`__hash__` 1374 will have its :meth:`__hash__` implicitly set to ``None``. When the 1375 :meth:`__hash__` method of a class is ``None``, instances of the class will 1376 raise an appropriate :exc:`TypeError` when a program attempts to retrieve 1377 their hash value, and will also be correctly identified as unhashable when 1378 checking ``isinstance(obj, collections.Hashable)``. 1379 1380 If a class that overrides :meth:`__eq__` needs to retain the implementation 1381 of :meth:`__hash__` from a parent class, the interpreter must be told this 1382 explicitly by setting ``__hash__ = <ParentClass>.__hash__``. 1383 1384 If a class that does not override :meth:`__eq__` wishes to suppress hash 1385 support, it should include ``__hash__ = None`` in the class definition. 1386 A class which defines its own :meth:`__hash__` that explicitly raises 1387 a :exc:`TypeError` would be incorrectly identified as hashable by 1388 an ``isinstance(obj, collections.Hashable)`` call. 1389 1390 1391 .. note:: 1392 1393 By default, the :meth:`__hash__` values of str, bytes and datetime 1394 objects are "salted" with an unpredictable random value. Although they 1395 remain constant within an individual Python process, they are not 1396 predictable between repeated invocations of Python. 1397 1398 This is intended to provide protection against a denial-of-service caused 1399 by carefully-chosen inputs that exploit the worst case performance of a 1400 dict insertion, O(n^2) complexity. See 1401 http://www.ocert.org/advisories/ocert-2011-003.html for details. 1402 1403 Changing hash values affects the iteration order of dicts, sets and 1404 other mappings. Python has never made guarantees about this ordering 1405 (and it typically varies between 32-bit and 64-bit builds). 1406 1407 See also :envvar:`PYTHONHASHSEED`. 1408 1409 .. versionchanged:: 3.3 1410 Hash randomization is enabled by default. 1411 1412 1413 .. method:: object.__bool__(self) 1414 1415 .. index:: single: __len__() (mapping object method) 1416 1417 Called to implement truth value testing and the built-in operation 1418 ``bool()``; should return ``False`` or ``True``. When this method is not 1419 defined, :meth:`__len__` is called, if it is defined, and the object is 1420 considered true if its result is nonzero. If a class defines neither 1421 :meth:`__len__` nor :meth:`__bool__`, all its instances are considered 1422 true. 1423 1424 1425 .. _attribute-access: 1426 1427 Customizing attribute access 1428 ---------------------------- 1429 1430 The following methods can be defined to customize the meaning of attribute 1431 access (use of, assignment to, or deletion of ``x.name``) for class instances. 1432 1433 .. XXX explain how descriptors interfere here! 1434 1435 1436 .. method:: object.__getattr__(self, name) 1437 1438 Called when an attribute lookup has not found the attribute in the usual places 1439 (i.e. it is not an instance attribute nor is it found in the class tree for 1440 ``self``). ``name`` is the attribute name. This method should return the 1441 (computed) attribute value or raise an :exc:`AttributeError` exception. 1442 1443 Note that if the attribute is found through the normal mechanism, 1444 :meth:`__getattr__` is not called. (This is an intentional asymmetry between 1445 :meth:`__getattr__` and :meth:`__setattr__`.) This is done both for efficiency 1446 reasons and because otherwise :meth:`__getattr__` would have no way to access 1447 other attributes of the instance. Note that at least for instance variables, 1448 you can fake total control by not inserting any values in the instance attribute 1449 dictionary (but instead inserting them in another object). See the 1450 :meth:`__getattribute__` method below for a way to actually get total control 1451 over attribute access. 1452 1453 1454 .. method:: object.__getattribute__(self, name) 1455 1456 Called unconditionally to implement attribute accesses for instances of the 1457 class. If the class also defines :meth:`__getattr__`, the latter will not be 1458 called unless :meth:`__getattribute__` either calls it explicitly or raises an 1459 :exc:`AttributeError`. This method should return the (computed) attribute value 1460 or raise an :exc:`AttributeError` exception. In order to avoid infinite 1461 recursion in this method, its implementation should always call the base class 1462 method with the same name to access any attributes it needs, for example, 1463 ``object.__getattribute__(self, name)``. 1464 1465 .. note:: 1466 1467 This method may still be bypassed when looking up special methods as the 1468 result of implicit invocation via language syntax or built-in functions. 1469 See :ref:`special-lookup`. 1470 1471 1472 .. method:: object.__setattr__(self, name, value) 1473 1474 Called when an attribute assignment is attempted. This is called instead of 1475 the normal mechanism (i.e. store the value in the instance dictionary). 1476 *name* is the attribute name, *value* is the value to be assigned to it. 1477 1478 If :meth:`__setattr__` wants to assign to an instance attribute, it should 1479 call the base class method with the same name, for example, 1480 ``object.__setattr__(self, name, value)``. 1481 1482 1483 .. method:: object.__delattr__(self, name) 1484 1485 Like :meth:`__setattr__` but for attribute deletion instead of assignment. This 1486 should only be implemented if ``del obj.name`` is meaningful for the object. 1487 1488 1489 .. method:: object.__dir__(self) 1490 1491 Called when :func:`dir` is called on the object. A sequence must be 1492 returned. :func:`dir` converts the returned sequence to a list and sorts it. 1493 1494 1495 .. _descriptors: 1496 1497 Implementing Descriptors 1498 ^^^^^^^^^^^^^^^^^^^^^^^^ 1499 1500 The following methods only apply when an instance of the class containing the 1501 method (a so-called *descriptor* class) appears in an *owner* class (the 1502 descriptor must be in either the owner's class dictionary or in the class 1503 dictionary for one of its parents). In the examples below, "the attribute" 1504 refers to the attribute whose name is the key of the property in the owner 1505 class' :attr:`~object.__dict__`. 1506 1507 1508 .. method:: object.__get__(self, instance, owner) 1509 1510 Called to get the attribute of the owner class (class attribute access) or of an 1511 instance of that class (instance attribute access). *owner* is always the owner 1512 class, while *instance* is the instance that the attribute was accessed through, 1513 or ``None`` when the attribute is accessed through the *owner*. This method 1514 should return the (computed) attribute value or raise an :exc:`AttributeError` 1515 exception. 1516 1517 1518 .. method:: object.__set__(self, instance, value) 1519 1520 Called to set the attribute on an instance *instance* of the owner class to a 1521 new value, *value*. 1522 1523 1524 .. method:: object.__delete__(self, instance) 1525 1526 Called to delete the attribute on an instance *instance* of the owner class. 1527 1528 1529 .. method:: object.__set_name__(self, owner, name) 1530 1531 Called at the time the owning class *owner* is created. The 1532 descriptor has been assigned to *name*. 1533 1534 .. versionadded:: 3.6 1535 1536 1537 The attribute :attr:`__objclass__` is interpreted by the :mod:`inspect` module 1538 as specifying the class where this object was defined (setting this 1539 appropriately can assist in runtime introspection of dynamic class attributes). 1540 For callables, it may indicate that an instance of the given type (or a 1541 subclass) is expected or required as the first positional argument (for example, 1542 CPython sets this attribute for unbound methods that are implemented in C). 1543 1544 1545 .. _descriptor-invocation: 1546 1547 Invoking Descriptors 1548 ^^^^^^^^^^^^^^^^^^^^ 1549 1550 In general, a descriptor is an object attribute with "binding behavior", one 1551 whose attribute access has been overridden by methods in the descriptor 1552 protocol: :meth:`__get__`, :meth:`__set__`, and :meth:`__delete__`. If any of 1553 those methods are defined for an object, it is said to be a descriptor. 1554 1555 The default behavior for attribute access is to get, set, or delete the 1556 attribute from an object's dictionary. For instance, ``a.x`` has a lookup chain 1557 starting with ``a.__dict__['x']``, then ``type(a).__dict__['x']``, and 1558 continuing through the base classes of ``type(a)`` excluding metaclasses. 1559 1560 However, if the looked-up value is an object defining one of the descriptor 1561 methods, then Python may override the default behavior and invoke the descriptor 1562 method instead. Where this occurs in the precedence chain depends on which 1563 descriptor methods were defined and how they were called. 1564 1565 The starting point for descriptor invocation is a binding, ``a.x``. How the 1566 arguments are assembled depends on ``a``: 1567 1568 Direct Call 1569 The simplest and least common call is when user code directly invokes a 1570 descriptor method: ``x.__get__(a)``. 1571 1572 Instance Binding 1573 If binding to an object instance, ``a.x`` is transformed into the call: 1574 ``type(a).__dict__['x'].__get__(a, type(a))``. 1575 1576 Class Binding 1577 If binding to a class, ``A.x`` is transformed into the call: 1578 ``A.__dict__['x'].__get__(None, A)``. 1579 1580 Super Binding 1581 If ``a`` is an instance of :class:`super`, then the binding ``super(B, 1582 obj).m()`` searches ``obj.__class__.__mro__`` for the base class ``A`` 1583 immediately preceding ``B`` and then invokes the descriptor with the call: 1584 ``A.__dict__['m'].__get__(obj, obj.__class__)``. 1585 1586 For instance bindings, the precedence of descriptor invocation depends on the 1587 which descriptor methods are defined. A descriptor can define any combination 1588 of :meth:`__get__`, :meth:`__set__` and :meth:`__delete__`. If it does not 1589 define :meth:`__get__`, then accessing the attribute will return the descriptor 1590 object itself unless there is a value in the object's instance dictionary. If 1591 the descriptor defines :meth:`__set__` and/or :meth:`__delete__`, it is a data 1592 descriptor; if it defines neither, it is a non-data descriptor. Normally, data 1593 descriptors define both :meth:`__get__` and :meth:`__set__`, while non-data 1594 descriptors have just the :meth:`__get__` method. Data descriptors with 1595 :meth:`__set__` and :meth:`__get__` defined always override a redefinition in an 1596 instance dictionary. In contrast, non-data descriptors can be overridden by 1597 instances. 1598 1599 Python methods (including :func:`staticmethod` and :func:`classmethod`) are 1600 implemented as non-data descriptors. Accordingly, instances can redefine and 1601 override methods. This allows individual instances to acquire behaviors that 1602 differ from other instances of the same class. 1603 1604 The :func:`property` function is implemented as a data descriptor. Accordingly, 1605 instances cannot override the behavior of a property. 1606 1607 1608 .. _slots: 1609 1610 __slots__ 1611 ^^^^^^^^^ 1612 1613 By default, instances of classes have a dictionary for attribute storage. This 1614 wastes space for objects having very few instance variables. The space 1615 consumption can become acute when creating large numbers of instances. 1616 1617 The default can be overridden by defining *__slots__* in a class definition. 1618 The *__slots__* declaration takes a sequence of instance variables and reserves 1619 just enough space in each instance to hold a value for each variable. Space is 1620 saved because *__dict__* is not created for each instance. 1621 1622 1623 .. data:: object.__slots__ 1624 1625 This class variable can be assigned a string, iterable, or sequence of 1626 strings with variable names used by instances. *__slots__* reserves space 1627 for the declared variables and prevents the automatic creation of *__dict__* 1628 and *__weakref__* for each instance. 1629 1630 1631 Notes on using *__slots__* 1632 """""""""""""""""""""""""" 1633 1634 * When inheriting from a class without *__slots__*, the *__dict__* attribute of 1635 that class will always be accessible, so a *__slots__* definition in the 1636 subclass is meaningless. 1637 1638 * Without a *__dict__* variable, instances cannot be assigned new variables not 1639 listed in the *__slots__* definition. Attempts to assign to an unlisted 1640 variable name raises :exc:`AttributeError`. If dynamic assignment of new 1641 variables is desired, then add ``'__dict__'`` to the sequence of strings in 1642 the *__slots__* declaration. 1643 1644 * Without a *__weakref__* variable for each instance, classes defining 1645 *__slots__* do not support weak references to its instances. If weak reference 1646 support is needed, then add ``'__weakref__'`` to the sequence of strings in the 1647 *__slots__* declaration. 1648 1649 * *__slots__* are implemented at the class level by creating descriptors 1650 (:ref:`descriptors`) for each variable name. As a result, class attributes 1651 cannot be used to set default values for instance variables defined by 1652 *__slots__*; otherwise, the class attribute would overwrite the descriptor 1653 assignment. 1654 1655 * The action of a *__slots__* declaration is limited to the class where it is 1656 defined. As a result, subclasses will have a *__dict__* unless they also define 1657 *__slots__* (which must only contain names of any *additional* slots). 1658 1659 * If a class defines a slot also defined in a base class, the instance variable 1660 defined by the base class slot is inaccessible (except by retrieving its 1661 descriptor directly from the base class). This renders the meaning of the 1662 program undefined. In the future, a check may be added to prevent this. 1663 1664 * Nonempty *__slots__* does not work for classes derived from "variable-length" 1665 built-in types such as :class:`int`, :class:`bytes` and :class:`tuple`. 1666 1667 * Any non-string iterable may be assigned to *__slots__*. Mappings may also be 1668 used; however, in the future, special meaning may be assigned to the values 1669 corresponding to each key. 1670 1671 * *__class__* assignment works only if both classes have the same *__slots__*. 1672 1673 1674 .. _class-customization: 1675 1676 Customizing class creation 1677 -------------------------- 1678 1679 Whenever a class inherits from another class, *__init_subclass__* is 1680 called on that class. This way, it is possible to write classes which 1681 change the behavior of subclasses. This is closely related to class 1682 decorators, but where class decorators only affect the specific class they're 1683 applied to, ``__init_subclass__`` solely applies to future subclasses of the 1684 class defining the method. 1685 1686 .. classmethod:: object.__init_subclass__(cls) 1687 1688 This method is called whenever the containing class is subclassed. 1689 *cls* is then the new subclass. If defined as a normal instance method, 1690 this method is implicitly converted to a class method. 1691 1692 Keyword arguments which are given to a new class are passed to 1693 the parent's class ``__init_subclass__``. For compatibility with 1694 other classes using ``__init_subclass__``, one should take out the 1695 needed keyword arguments and pass the others over to the base 1696 class, as in:: 1697 1698 class Philosopher: 1699 def __init_subclass__(cls, default_name, **kwargs): 1700 super().__init_subclass__(**kwargs) 1701 cls.default_name = default_name 1702 1703 class AustralianPhilosopher(Philosopher, default_name="Bruce"): 1704 pass 1705 1706 The default implementation ``object.__init_subclass__`` does 1707 nothing, but raises an error if it is called with any arguments. 1708 1709 .. note:: 1710 1711 The metaclass hint ``metaclass`` is consumed by the rest of the type 1712 machinery, and is never passed to ``__init_subclass__`` implementations. 1713 The actual metaclass (rather than the explicit hint) can be accessed as 1714 ``type(cls)``. 1715 1716 .. versionadded:: 3.6 1717 1718 1719 .. _metaclasses: 1720 1721 Metaclasses 1722 ^^^^^^^^^^^ 1723 1724 .. index:: 1725 single: metaclass 1726 builtin: type 1727 1728 By default, classes are constructed using :func:`type`. The class body is 1729 executed in a new namespace and the class name is bound locally to the 1730 result of ``type(name, bases, namespace)``. 1731 1732 The class creation process can be customized by passing the ``metaclass`` 1733 keyword argument in the class definition line, or by inheriting from an 1734 existing class that included such an argument. In the following example, 1735 both ``MyClass`` and ``MySubclass`` are instances of ``Meta``:: 1736 1737 class Meta(type): 1738 pass 1739 1740 class MyClass(metaclass=Meta): 1741 pass 1742 1743 class MySubclass(MyClass): 1744 pass 1745 1746 Any other keyword arguments that are specified in the class definition are 1747 passed through to all metaclass operations described below. 1748 1749 When a class definition is executed, the following steps occur: 1750 1751 * the appropriate metaclass is determined 1752 * the class namespace is prepared 1753 * the class body is executed 1754 * the class object is created 1755 1756 Determining the appropriate metaclass 1757 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 1758 .. index:: 1759 single: metaclass hint 1760 1761 The appropriate metaclass for a class definition is determined as follows: 1762 1763 * if no bases and no explicit metaclass are given, then :func:`type` is used 1764 * if an explicit metaclass is given and it is *not* an instance of 1765 :func:`type`, then it is used directly as the metaclass 1766 * if an instance of :func:`type` is given as the explicit metaclass, or 1767 bases are defined, then the most derived metaclass is used 1768 1769 The most derived metaclass is selected from the explicitly specified 1770 metaclass (if any) and the metaclasses (i.e. ``type(cls)``) of all specified 1771 base classes. The most derived metaclass is one which is a subtype of *all* 1772 of these candidate metaclasses. If none of the candidate metaclasses meets 1773 that criterion, then the class definition will fail with ``TypeError``. 1774 1775 1776 .. _prepare: 1777 1778 Preparing the class namespace 1779 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 1780 1781 .. index:: 1782 single: __prepare__ (metaclass method) 1783 1784 Once the appropriate metaclass has been identified, then the class namespace 1785 is prepared. If the metaclass has a ``__prepare__`` attribute, it is called 1786 as ``namespace = metaclass.__prepare__(name, bases, **kwds)`` (where the 1787 additional keyword arguments, if any, come from the class definition). 1788 1789 If the metaclass has no ``__prepare__`` attribute, then the class namespace 1790 is initialised as an empty ordered mapping. 1791 1792 .. seealso:: 1793 1794 :pep:`3115` - Metaclasses in Python 3000 1795 Introduced the ``__prepare__`` namespace hook 1796 1797 1798 Executing the class body 1799 ^^^^^^^^^^^^^^^^^^^^^^^^ 1800 1801 .. index:: 1802 single: class; body 1803 1804 The class body is executed (approximately) as 1805 ``exec(body, globals(), namespace)``. The key difference from a normal 1806 call to :func:`exec` is that lexical scoping allows the class body (including 1807 any methods) to reference names from the current and outer scopes when the 1808 class definition occurs inside a function. 1809 1810 However, even when the class definition occurs inside the function, methods 1811 defined inside the class still cannot see names defined at the class scope. 1812 Class variables must be accessed through the first parameter of instance or 1813 class methods, or through the implicit lexically scoped ``__class__`` reference 1814 described in the next section. 1815 1816 .. _class-object-creation: 1817 1818 Creating the class object 1819 ^^^^^^^^^^^^^^^^^^^^^^^^^ 1820 1821 .. index:: 1822 single: __class__ (method cell) 1823 single: __classcell__ (class namespace entry) 1824 1825 1826 Once the class namespace has been populated by executing the class body, 1827 the class object is created by calling 1828 ``metaclass(name, bases, namespace, **kwds)`` (the additional keywords 1829 passed here are the same as those passed to ``__prepare__``). 1830 1831 This class object is the one that will be referenced by the zero-argument 1832 form of :func:`super`. ``__class__`` is an implicit closure reference 1833 created by the compiler if any methods in a class body refer to either 1834 ``__class__`` or ``super``. This allows the zero argument form of 1835 :func:`super` to correctly identify the class being defined based on 1836 lexical scoping, while the class or instance that was used to make the 1837 current call is identified based on the first argument passed to the method. 1838 1839 .. impl-detail:: 1840 1841 In CPython 3.6 and later, the ``__class__`` cell is passed to the metaclass 1842 as a ``__classcell__`` entry in the class namespace. If present, this must 1843 be propagated up to the ``type.__new__`` call in order for the class to be 1844 initialised correctly. 1845 Failing to do so will result in a :exc:`DeprecationWarning` in Python 3.6, 1846 and a :exc:`RuntimeWarning` in the future. 1847 1848 When using the default metaclass :class:`type`, or any metaclass that ultimately 1849 calls ``type.__new__``, the following additional customisation steps are 1850 invoked after creating the class object: 1851 1852 * first, ``type.__new__`` collects all of the descriptors in the class 1853 namespace that define a :meth:`~object.__set_name__` method; 1854 * second, all of these ``__set_name__`` methods are called with the class 1855 being defined and the assigned name of that particular descriptor; and 1856 * finally, the :meth:`~object.__init_subclass__` hook is called on the 1857 immediate parent of the new class in its method resolution order. 1858 1859 After the class object is created, it is passed to the class decorators 1860 included in the class definition (if any) and the resulting object is bound 1861 in the local namespace as the defined class. 1862 1863 When a new class is created by ``type.__new__``, the object provided as the 1864 namespace parameter is copied to a new ordered mapping and the original 1865 object is discarded. The new copy is wrapped in a read-only proxy, which 1866 becomes the :attr:`~object.__dict__` attribute of the class object. 1867 1868 .. seealso:: 1869 1870 :pep:`3135` - New super 1871 Describes the implicit ``__class__`` closure reference 1872 1873 1874 Metaclass example 1875 ^^^^^^^^^^^^^^^^^ 1876 1877 The potential uses for metaclasses are boundless. Some ideas that have been 1878 explored include logging, interface checking, automatic delegation, automatic 1879 property creation, proxies, frameworks, and automatic resource 1880 locking/synchronization. 1881 1882 Here is an example of a metaclass that uses an :class:`collections.OrderedDict` 1883 to remember the order that class variables are defined:: 1884 1885 class OrderedClass(type): 1886 1887 @classmethod 1888 def __prepare__(metacls, name, bases, **kwds): 1889 return collections.OrderedDict() 1890 1891 def __new__(cls, name, bases, namespace, **kwds): 1892 result = type.__new__(cls, name, bases, dict(namespace)) 1893 result.members = tuple(namespace) 1894 return result 1895 1896 class A(metaclass=OrderedClass): 1897 def one(self): pass 1898 def two(self): pass 1899 def three(self): pass 1900 def four(self): pass 1901 1902 >>> A.members 1903 ('__module__', 'one', 'two', 'three', 'four') 1904 1905 When the class definition for *A* gets executed, the process begins with 1906 calling the metaclass's :meth:`__prepare__` method which returns an empty 1907 :class:`collections.OrderedDict`. That mapping records the methods and 1908 attributes of *A* as they are defined within the body of the class statement. 1909 Once those definitions are executed, the ordered dictionary is fully populated 1910 and the metaclass's :meth:`__new__` method gets invoked. That method builds 1911 the new type and it saves the ordered dictionary keys in an attribute 1912 called ``members``. 1913 1914 1915 Customizing instance and subclass checks 1916 ---------------------------------------- 1917 1918 The following methods are used to override the default behavior of the 1919 :func:`isinstance` and :func:`issubclass` built-in functions. 1920 1921 In particular, the metaclass :class:`abc.ABCMeta` implements these methods in 1922 order to allow the addition of Abstract Base Classes (ABCs) as "virtual base 1923 classes" to any class or type (including built-in types), including other 1924 ABCs. 1925 1926 .. method:: class.__instancecheck__(self, instance) 1927 1928 Return true if *instance* should be considered a (direct or indirect) 1929 instance of *class*. If defined, called to implement ``isinstance(instance, 1930 class)``. 1931 1932 1933 .. method:: class.__subclasscheck__(self, subclass) 1934 1935 Return true if *subclass* should be considered a (direct or indirect) 1936 subclass of *class*. If defined, called to implement ``issubclass(subclass, 1937 class)``. 1938 1939 1940 Note that these methods are looked up on the type (metaclass) of a class. They 1941 cannot be defined as class methods in the actual class. This is consistent with 1942 the lookup of special methods that are called on instances, only in this 1943 case the instance is itself a class. 1944 1945 .. seealso:: 1946 1947 :pep:`3119` - Introducing Abstract Base Classes 1948 Includes the specification for customizing :func:`isinstance` and 1949 :func:`issubclass` behavior through :meth:`~class.__instancecheck__` and 1950 :meth:`~class.__subclasscheck__`, with motivation for this functionality 1951 in the context of adding Abstract Base Classes (see the :mod:`abc` 1952 module) to the language. 1953 1954 1955 .. _callable-types: 1956 1957 Emulating callable objects 1958 -------------------------- 1959 1960 1961 .. method:: object.__call__(self[, args...]) 1962 1963 .. index:: pair: call; instance 1964 1965 Called when the instance is "called" as a function; if this method is defined, 1966 ``x(arg1, arg2, ...)`` is a shorthand for ``x.__call__(arg1, arg2, ...)``. 1967 1968 1969 .. _sequence-types: 1970 1971 Emulating container types 1972 ------------------------- 1973 1974 The following methods can be defined to implement container objects. Containers 1975 usually are sequences (such as lists or tuples) or mappings (like dictionaries), 1976 but can represent other containers as well. The first set of methods is used 1977 either to emulate a sequence or to emulate a mapping; the difference is that for 1978 a sequence, the allowable keys should be the integers *k* for which ``0 <= k < 1979 N`` where *N* is the length of the sequence, or slice objects, which define a 1980 range of items. It is also recommended that mappings provide the methods 1981 :meth:`keys`, :meth:`values`, :meth:`items`, :meth:`get`, :meth:`clear`, 1982 :meth:`setdefault`, :meth:`pop`, :meth:`popitem`, :meth:`!copy`, and 1983 :meth:`update` behaving similar to those for Python's standard dictionary 1984 objects. The :mod:`collections` module provides a 1985 :class:`~collections.abc.MutableMapping` 1986 abstract base class to help create those methods from a base set of 1987 :meth:`__getitem__`, :meth:`__setitem__`, :meth:`__delitem__`, and :meth:`keys`. 1988 Mutable sequences should provide methods :meth:`append`, :meth:`count`, 1989 :meth:`index`, :meth:`extend`, :meth:`insert`, :meth:`pop`, :meth:`remove`, 1990 :meth:`reverse` and :meth:`sort`, like Python standard list objects. Finally, 1991 sequence types should implement addition (meaning concatenation) and 1992 multiplication (meaning repetition) by defining the methods :meth:`__add__`, 1993 :meth:`__radd__`, :meth:`__iadd__`, :meth:`__mul__`, :meth:`__rmul__` and 1994 :meth:`__imul__` described below; they should not define other numerical 1995 operators. It is recommended that both mappings and sequences implement the 1996 :meth:`__contains__` method to allow efficient use of the ``in`` operator; for 1997 mappings, ``in`` should search the mapping's keys; for sequences, it should 1998 search through the values. It is further recommended that both mappings and 1999 sequences implement the :meth:`__iter__` method to allow efficient iteration 2000 through the container; for mappings, :meth:`__iter__` should be the same as 2001 :meth:`keys`; for sequences, it should iterate through the values. 2002 2003 .. method:: object.__len__(self) 2004 2005 .. index:: 2006 builtin: len 2007 single: __bool__() (object method) 2008 2009 Called to implement the built-in function :func:`len`. Should return the length 2010 of the object, an integer ``>=`` 0. Also, an object that doesn't define a 2011 :meth:`__bool__` method and whose :meth:`__len__` method returns zero is 2012 considered to be false in a Boolean context. 2013 2014 2015 .. method:: object.__length_hint__(self) 2016 2017 Called to implement :func:`operator.length_hint`. Should return an estimated 2018 length for the object (which may be greater or less than the actual length). 2019 The length must be an integer ``>=`` 0. This method is purely an 2020 optimization and is never required for correctness. 2021 2022 .. versionadded:: 3.4 2023 2024 .. note:: 2025 2026 Slicing is done exclusively with the following three methods. A call like :: 2027 2028 a[1:2] = b 2029 2030 is translated to :: 2031 2032 a[slice(1, 2, None)] = b 2033 2034 and so forth. Missing slice items are always filled in with ``None``. 2035 2036 2037 .. method:: object.__getitem__(self, key) 2038 2039 .. index:: object: slice 2040 2041 Called to implement evaluation of ``self[key]``. For sequence types, the 2042 accepted keys should be integers and slice objects. Note that the special 2043 interpretation of negative indexes (if the class wishes to emulate a sequence 2044 type) is up to the :meth:`__getitem__` method. If *key* is of an inappropriate 2045 type, :exc:`TypeError` may be raised; if of a value outside the set of indexes 2046 for the sequence (after any special interpretation of negative values), 2047 :exc:`IndexError` should be raised. For mapping types, if *key* is missing (not 2048 in the container), :exc:`KeyError` should be raised. 2049 2050 .. note:: 2051 2052 :keyword:`for` loops expect that an :exc:`IndexError` will be raised for illegal 2053 indexes to allow proper detection of the end of the sequence. 2054 2055 2056 .. method:: object.__missing__(self, key) 2057 2058 Called by :class:`dict`\ .\ :meth:`__getitem__` to implement ``self[key]`` for dict subclasses 2059 when key is not in the dictionary. 2060 2061 2062 .. method:: object.__setitem__(self, key, value) 2063 2064 Called to implement assignment to ``self[key]``. Same note as for 2065 :meth:`__getitem__`. This should only be implemented for mappings if the 2066 objects support changes to the values for keys, or if new keys can be added, or 2067 for sequences if elements can be replaced. The same exceptions should be raised 2068 for improper *key* values as for the :meth:`__getitem__` method. 2069 2070 2071 .. method:: object.__delitem__(self, key) 2072 2073 Called to implement deletion of ``self[key]``. Same note as for 2074 :meth:`__getitem__`. This should only be implemented for mappings if the 2075 objects support removal of keys, or for sequences if elements can be removed 2076 from the sequence. The same exceptions should be raised for improper *key* 2077 values as for the :meth:`__getitem__` method. 2078 2079 2080 .. method:: object.__iter__(self) 2081 2082 This method is called when an iterator is required for a container. This method 2083 should return a new iterator object that can iterate over all the objects in the 2084 container. For mappings, it should iterate over the keys of the container. 2085 2086 Iterator objects also need to implement this method; they are required to return 2087 themselves. For more information on iterator objects, see :ref:`typeiter`. 2088 2089 2090 .. method:: object.__reversed__(self) 2091 2092 Called (if present) by the :func:`reversed` built-in to implement 2093 reverse iteration. It should return a new iterator object that iterates 2094 over all the objects in the container in reverse order. 2095 2096 If the :meth:`__reversed__` method is not provided, the :func:`reversed` 2097 built-in will fall back to using the sequence protocol (:meth:`__len__` and 2098 :meth:`__getitem__`). Objects that support the sequence protocol should 2099 only provide :meth:`__reversed__` if they can provide an implementation 2100 that is more efficient than the one provided by :func:`reversed`. 2101 2102 2103 The membership test operators (:keyword:`in` and :keyword:`not in`) are normally 2104 implemented as an iteration through a sequence. However, container objects can 2105 supply the following special method with a more efficient implementation, which 2106 also does not require the object be a sequence. 2107 2108 .. method:: object.__contains__(self, item) 2109 2110 Called to implement membership test operators. Should return true if *item* 2111 is in *self*, false otherwise. For mapping objects, this should consider the 2112 keys of the mapping rather than the values or the key-item pairs. 2113 2114 For objects that don't define :meth:`__contains__`, the membership test first 2115 tries iteration via :meth:`__iter__`, then the old sequence iteration 2116 protocol via :meth:`__getitem__`, see :ref:`this section in the language 2117 reference <membership-test-details>`. 2118 2119 2120 .. _numeric-types: 2121 2122 Emulating numeric types 2123 ----------------------- 2124 2125 The following methods can be defined to emulate numeric objects. Methods 2126 corresponding to operations that are not supported by the particular kind of 2127 number implemented (e.g., bitwise operations for non-integral numbers) should be 2128 left undefined. 2129 2130 2131 .. method:: object.__add__(self, other) 2132 object.__sub__(self, other) 2133 object.__mul__(self, other) 2134 object.__matmul__(self, other) 2135 object.__truediv__(self, other) 2136 object.__floordiv__(self, other) 2137 object.__mod__(self, other) 2138 object.__divmod__(self, other) 2139 object.__pow__(self, other[, modulo]) 2140 object.__lshift__(self, other) 2141 object.__rshift__(self, other) 2142 object.__and__(self, other) 2143 object.__xor__(self, other) 2144 object.__or__(self, other) 2145 2146 .. index:: 2147 builtin: divmod 2148 builtin: pow 2149 builtin: pow 2150 2151 These methods are called to implement the binary arithmetic operations 2152 (``+``, ``-``, ``*``, ``@``, ``/``, ``//``, ``%``, :func:`divmod`, 2153 :func:`pow`, ``**``, ``<<``, ``>>``, ``&``, ``^``, ``|``). For instance, to 2154 evaluate the expression ``x + y``, where *x* is an instance of a class that 2155 has an :meth:`__add__` method, ``x.__add__(y)`` is called. The 2156 :meth:`__divmod__` method should be the equivalent to using 2157 :meth:`__floordiv__` and :meth:`__mod__`; it should not be related to 2158 :meth:`__truediv__`. Note that :meth:`__pow__` should be defined to accept 2159 an optional third argument if the ternary version of the built-in :func:`pow` 2160 function is to be supported. 2161 2162 If one of those methods does not support the operation with the supplied 2163 arguments, it should return ``NotImplemented``. 2164 2165 2166 .. method:: object.__radd__(self, other) 2167 object.__rsub__(self, other) 2168 object.__rmul__(self, other) 2169 object.__rmatmul__(self, other) 2170 object.__rtruediv__(self, other) 2171 object.__rfloordiv__(self, other) 2172 object.__rmod__(self, other) 2173 object.__rdivmod__(self, other) 2174 object.__rpow__(self, other) 2175 object.__rlshift__(self, other) 2176 object.__rrshift__(self, other) 2177 object.__rand__(self, other) 2178 object.__rxor__(self, other) 2179 object.__ror__(self, other) 2180 2181 .. index:: 2182 builtin: divmod 2183 builtin: pow 2184 2185 These methods are called to implement the binary arithmetic operations 2186 (``+``, ``-``, ``*``, ``@``, ``/``, ``//``, ``%``, :func:`divmod`, 2187 :func:`pow`, ``**``, ``<<``, ``>>``, ``&``, ``^``, ``|``) with reflected 2188 (swapped) operands. These functions are only called if the left operand does 2189 not support the corresponding operation [#]_ and the operands are of different 2190 types. [#]_ For instance, to evaluate the expression ``x - y``, where *y* is 2191 an instance of a class that has an :meth:`__rsub__` method, ``y.__rsub__(x)`` 2192 is called if ``x.__sub__(y)`` returns *NotImplemented*. 2193 2194 .. index:: builtin: pow 2195 2196 Note that ternary :func:`pow` will not try calling :meth:`__rpow__` (the 2197 coercion rules would become too complicated). 2198 2199 .. note:: 2200 2201 If the right operand's type is a subclass of the left operand's type and that 2202 subclass provides the reflected method for the operation, this method will be 2203 called before the left operand's non-reflected method. This behavior allows 2204 subclasses to override their ancestors' operations. 2205 2206 2207 .. method:: object.__iadd__(self, other) 2208 object.__isub__(self, other) 2209 object.__imul__(self, other) 2210 object.__imatmul__(self, other) 2211 object.__itruediv__(self, other) 2212 object.__ifloordiv__(self, other) 2213 object.__imod__(self, other) 2214 object.__ipow__(self, other[, modulo]) 2215 object.__ilshift__(self, other) 2216 object.__irshift__(self, other) 2217 object.__iand__(self, other) 2218 object.__ixor__(self, other) 2219 object.__ior__(self, other) 2220 2221 These methods are called to implement the augmented arithmetic assignments 2222 (``+=``, ``-=``, ``*=``, ``@=``, ``/=``, ``//=``, ``%=``, ``**=``, ``<<=``, 2223 ``>>=``, ``&=``, ``^=``, ``|=``). These methods should attempt to do the 2224 operation in-place (modifying *self*) and return the result (which could be, 2225 but does not have to be, *self*). If a specific method is not defined, the 2226 augmented assignment falls back to the normal methods. For instance, if *x* 2227 is an instance of a class with an :meth:`__iadd__` method, ``x += y`` is 2228 equivalent to ``x = x.__iadd__(y)`` . Otherwise, ``x.__add__(y)`` and 2229 ``y.__radd__(x)`` are considered, as with the evaluation of ``x + y``. In 2230 certain situations, augmented assignment can result in unexpected errors (see 2231 :ref:`faq-augmented-assignment-tuple-error`), but this behavior is in fact 2232 part of the data model. 2233 2234 2235 .. method:: object.__neg__(self) 2236 object.__pos__(self) 2237 object.__abs__(self) 2238 object.__invert__(self) 2239 2240 .. index:: builtin: abs 2241 2242 Called to implement the unary arithmetic operations (``-``, ``+``, :func:`abs` 2243 and ``~``). 2244 2245 2246 .. method:: object.__complex__(self) 2247 object.__int__(self) 2248 object.__float__(self) 2249 object.__round__(self, [,n]) 2250 2251 .. index:: 2252 builtin: complex 2253 builtin: int 2254 builtin: float 2255 builtin: round 2256 2257 Called to implement the built-in functions :func:`complex`, 2258 :func:`int`, :func:`float` and :func:`round`. Should return a value 2259 of the appropriate type. 2260 2261 2262 .. method:: object.__index__(self) 2263 2264 Called to implement :func:`operator.index`, and whenever Python needs to 2265 losslessly convert the numeric object to an integer object (such as in 2266 slicing, or in the built-in :func:`bin`, :func:`hex` and :func:`oct` 2267 functions). Presence of this method indicates that the numeric object is 2268 an integer type. Must return an integer. 2269 2270 .. note:: 2271 2272 In order to have a coherent integer type class, when :meth:`__index__` is 2273 defined :meth:`__int__` should also be defined, and both should return 2274 the same value. 2275 2276 2277 .. _context-managers: 2278 2279 With Statement Context Managers 2280 ------------------------------- 2281 2282 A :dfn:`context manager` is an object that defines the runtime context to be 2283 established when executing a :keyword:`with` statement. The context manager 2284 handles the entry into, and the exit from, the desired runtime context for the 2285 execution of the block of code. Context managers are normally invoked using the 2286 :keyword:`with` statement (described in section :ref:`with`), but can also be 2287 used by directly invoking their methods. 2288 2289 .. index:: 2290 statement: with 2291 single: context manager 2292 2293 Typical uses of context managers include saving and restoring various kinds of 2294 global state, locking and unlocking resources, closing opened files, etc. 2295 2296 For more information on context managers, see :ref:`typecontextmanager`. 2297 2298 2299 .. method:: object.__enter__(self) 2300 2301 Enter the runtime context related to this object. The :keyword:`with` statement 2302 will bind this method's return value to the target(s) specified in the 2303 :keyword:`as` clause of the statement, if any. 2304 2305 2306 .. method:: object.__exit__(self, exc_type, exc_value, traceback) 2307 2308 Exit the runtime context related to this object. The parameters describe the 2309 exception that caused the context to be exited. If the context was exited 2310 without an exception, all three arguments will be :const:`None`. 2311 2312 If an exception is supplied, and the method wishes to suppress the exception 2313 (i.e., prevent it from being propagated), it should return a true value. 2314 Otherwise, the exception will be processed normally upon exit from this method. 2315 2316 Note that :meth:`__exit__` methods should not reraise the passed-in exception; 2317 this is the caller's responsibility. 2318 2319 2320 .. seealso:: 2321 2322 :pep:`343` - The "with" statement 2323 The specification, background, and examples for the Python :keyword:`with` 2324 statement. 2325 2326 2327 .. _special-lookup: 2328 2329 Special method lookup 2330 --------------------- 2331 2332 For custom classes, implicit invocations of special methods are only guaranteed 2333 to work correctly if defined on an object's type, not in the object's instance 2334 dictionary. That behaviour is the reason why the following code raises an 2335 exception:: 2336 2337 >>> class C: 2338 ... pass 2339 ... 2340 >>> c = C() 2341 >>> c.__len__ = lambda: 5 2342 >>> len(c) 2343 Traceback (most recent call last): 2344 File "<stdin>", line 1, in <module> 2345 TypeError: object of type 'C' has no len() 2346 2347 The rationale behind this behaviour lies with a number of special methods such 2348 as :meth:`__hash__` and :meth:`__repr__` that are implemented by all objects, 2349 including type objects. If the implicit lookup of these methods used the 2350 conventional lookup process, they would fail when invoked on the type object 2351 itself:: 2352 2353 >>> 1 .__hash__() == hash(1) 2354 True 2355 >>> int.__hash__() == hash(int) 2356 Traceback (most recent call last): 2357 File "<stdin>", line 1, in <module> 2358 TypeError: descriptor '__hash__' of 'int' object needs an argument 2359 2360 Incorrectly attempting to invoke an unbound method of a class in this way is 2361 sometimes referred to as 'metaclass confusion', and is avoided by bypassing 2362 the instance when looking up special methods:: 2363 2364 >>> type(1).__hash__(1) == hash(1) 2365 True 2366 >>> type(int).__hash__(int) == hash(int) 2367 True 2368 2369 In addition to bypassing any instance attributes in the interest of 2370 correctness, implicit special method lookup generally also bypasses the 2371 :meth:`__getattribute__` method even of the object's metaclass:: 2372 2373 >>> class Meta(type): 2374 ... def __getattribute__(*args): 2375 ... print("Metaclass getattribute invoked") 2376 ... return type.__getattribute__(*args) 2377 ... 2378 >>> class C(object, metaclass=Meta): 2379 ... def __len__(self): 2380 ... return 10 2381 ... def __getattribute__(*args): 2382 ... print("Class getattribute invoked") 2383 ... return object.__getattribute__(*args) 2384 ... 2385 >>> c = C() 2386 >>> c.__len__() # Explicit lookup via instance 2387 Class getattribute invoked 2388 10 2389 >>> type(c).__len__(c) # Explicit lookup via type 2390 Metaclass getattribute invoked 2391 10 2392 >>> len(c) # Implicit lookup 2393 10 2394 2395 Bypassing the :meth:`__getattribute__` machinery in this fashion 2396 provides significant scope for speed optimisations within the 2397 interpreter, at the cost of some flexibility in the handling of 2398 special methods (the special method *must* be set on the class 2399 object itself in order to be consistently invoked by the interpreter). 2400 2401 2402 .. index:: 2403 single: coroutine 2404 2405 Coroutines 2406 ========== 2407 2408 2409 Awaitable Objects 2410 ----------------- 2411 2412 An :term:`awaitable` object generally implements an :meth:`__await__` method. 2413 :term:`Coroutine` objects returned from :keyword:`async def` functions 2414 are awaitable. 2415 2416 .. note:: 2417 2418 The :term:`generator iterator` objects returned from generators 2419 decorated with :func:`types.coroutine` or :func:`asyncio.coroutine` 2420 are also awaitable, but they do not implement :meth:`__await__`. 2421 2422 .. method:: object.__await__(self) 2423 2424 Must return an :term:`iterator`. Should be used to implement 2425 :term:`awaitable` objects. For instance, :class:`asyncio.Future` implements 2426 this method to be compatible with the :keyword:`await` expression. 2427 2428 .. versionadded:: 3.5 2429 2430 .. seealso:: :pep:`492` for additional information about awaitable objects. 2431 2432 2433 .. _coroutine-objects: 2434 2435 Coroutine Objects 2436 ----------------- 2437 2438 :term:`Coroutine` objects are :term:`awaitable` objects. 2439 A coroutine's execution can be controlled by calling :meth:`__await__` and 2440 iterating over the result. When the coroutine has finished executing and 2441 returns, the iterator raises :exc:`StopIteration`, and the exception's 2442 :attr:`~StopIteration.value` attribute holds the return value. If the 2443 coroutine raises an exception, it is propagated by the iterator. Coroutines 2444 should not directly raise unhandled :exc:`StopIteration` exceptions. 2445 2446 Coroutines also have the methods listed below, which are analogous to 2447 those of generators (see :ref:`generator-methods`). However, unlike 2448 generators, coroutines do not directly support iteration. 2449 2450 .. versionchanged:: 3.5.2 2451 It is a :exc:`RuntimeError` to await on a coroutine more than once. 2452 2453 2454 .. method:: coroutine.send(value) 2455 2456 Starts or resumes execution of the coroutine. If *value* is ``None``, 2457 this is equivalent to advancing the iterator returned by 2458 :meth:`__await__`. If *value* is not ``None``, this method delegates 2459 to the :meth:`~generator.send` method of the iterator that caused 2460 the coroutine to suspend. The result (return value, 2461 :exc:`StopIteration`, or other exception) is the same as when 2462 iterating over the :meth:`__await__` return value, described above. 2463 2464 .. method:: coroutine.throw(type[, value[, traceback]]) 2465 2466 Raises the specified exception in the coroutine. This method delegates 2467 to the :meth:`~generator.throw` method of the iterator that caused 2468 the coroutine to suspend, if it has such a method. Otherwise, 2469 the exception is raised at the suspension point. The result 2470 (return value, :exc:`StopIteration`, or other exception) is the same as 2471 when iterating over the :meth:`__await__` return value, described 2472 above. If the exception is not caught in the coroutine, it propagates 2473 back to the caller. 2474 2475 .. method:: coroutine.close() 2476 2477 Causes the coroutine to clean itself up and exit. If the coroutine 2478 is suspended, this method first delegates to the :meth:`~generator.close` 2479 method of the iterator that caused the coroutine to suspend, if it 2480 has such a method. Then it raises :exc:`GeneratorExit` at the 2481 suspension point, causing the coroutine to immediately clean itself up. 2482 Finally, the coroutine is marked as having finished executing, even if 2483 it was never started. 2484 2485 Coroutine objects are automatically closed using the above process when 2486 they are about to be destroyed. 2487 2488 .. _async-iterators: 2489 2490 Asynchronous Iterators 2491 ---------------------- 2492 2493 An *asynchronous iterable* is able to call asynchronous code in its 2494 ``__aiter__`` implementation, and an *asynchronous iterator* can call 2495 asynchronous code in its ``__anext__`` method. 2496 2497 Asynchronous iterators can be used in an :keyword:`async for` statement. 2498 2499 .. method:: object.__aiter__(self) 2500 2501 Must return an *asynchronous iterator* object. 2502 2503 .. method:: object.__anext__(self) 2504 2505 Must return an *awaitable* resulting in a next value of the iterator. Should 2506 raise a :exc:`StopAsyncIteration` error when the iteration is over. 2507 2508 An example of an asynchronous iterable object:: 2509 2510 class Reader: 2511 async def readline(self): 2512 ... 2513 2514 def __aiter__(self): 2515 return self 2516 2517 async def __anext__(self): 2518 val = await self.readline() 2519 if val == b'': 2520 raise StopAsyncIteration 2521 return val 2522 2523 .. versionadded:: 3.5 2524 2525 .. note:: 2526 2527 .. versionchanged:: 3.5.2 2528 Starting with CPython 3.5.2, ``__aiter__`` can directly return 2529 :term:`asynchronous iterators <asynchronous iterator>`. Returning 2530 an :term:`awaitable` object will result in a 2531 :exc:`PendingDeprecationWarning`. 2532 2533 The recommended way of writing backwards compatible code in 2534 CPython 3.5.x is to continue returning awaitables from 2535 ``__aiter__``. If you want to avoid the PendingDeprecationWarning 2536 and keep the code backwards compatible, the following decorator 2537 can be used:: 2538 2539 import functools 2540 import sys 2541 2542 if sys.version_info < (3, 5, 2): 2543 def aiter_compat(func): 2544 @functools.wraps(func) 2545 async def wrapper(self): 2546 return func(self) 2547 return wrapper 2548 else: 2549 def aiter_compat(func): 2550 return func 2551 2552 Example:: 2553 2554 class AsyncIterator: 2555 2556 @aiter_compat 2557 def __aiter__(self): 2558 return self 2559 2560 async def __anext__(self): 2561 ... 2562 2563 Starting with CPython 3.6, the :exc:`PendingDeprecationWarning` 2564 will be replaced with the :exc:`DeprecationWarning`. 2565 In CPython 3.7, returning an awaitable from ``__aiter__`` will 2566 result in a :exc:`RuntimeError`. 2567 2568 2569 Asynchronous Context Managers 2570 ----------------------------- 2571 2572 An *asynchronous context manager* is a *context manager* that is able to 2573 suspend execution in its ``__aenter__`` and ``__aexit__`` methods. 2574 2575 Asynchronous context managers can be used in an :keyword:`async with` statement. 2576 2577 .. method:: object.__aenter__(self) 2578 2579 This method is semantically similar to the :meth:`__enter__`, with only 2580 difference that it must return an *awaitable*. 2581 2582 .. method:: object.__aexit__(self, exc_type, exc_value, traceback) 2583 2584 This method is semantically similar to the :meth:`__exit__`, with only 2585 difference that it must return an *awaitable*. 2586 2587 An example of an asynchronous context manager class:: 2588 2589 class AsyncContextManager: 2590 async def __aenter__(self): 2591 await log('entering context') 2592 2593 async def __aexit__(self, exc_type, exc, tb): 2594 await log('exiting context') 2595 2596 .. versionadded:: 3.5 2597 2598 2599 .. rubric:: Footnotes 2600 2601 .. [#] It *is* possible in some cases to change an object's type, under certain 2602 controlled conditions. It generally isn't a good idea though, since it can 2603 lead to some very strange behaviour if it is handled incorrectly. 2604 2605 .. [#] The :meth:`__hash__`, :meth:`__iter__`, :meth:`__reversed__`, and 2606 :meth:`__contains__` methods have special handling for this; others 2607 will still raise a :exc:`TypeError`, but may do so by relying on 2608 the behavior that ``None`` is not callable. 2609 2610 .. [#] "Does not support" here means that the class has no such method, or 2611 the method returns ``NotImplemented``. Do not set the method to 2612 ``None`` if you want to force fallback to the right operand's reflected 2613 methodthat will instead have the opposite effect of explicitly 2614 *blocking* such fallback. 2615 2616 .. [#] For operands of the same type, it is assumed that if the non-reflected method 2617 (such as :meth:`__add__`) fails the operation is not supported, which is why the 2618 reflected method is not called. 2619
