python-peps/peps/pep-0435.rst

PEP: 435
Title: Adding an Enum type to the Python standard library
Version: $Revision$
Last-Modified: $Date$
Author: Barry Warsaw <barry@python.org>,
        Eli Bendersky <eliben@gmail.com>,
        Ethan Furman <ethan@stoneleaf.us>
Status: Final
Type: Standards Track
Content-Type: text/x-rst
Created: 23-Feb-2013
Python-Version: 3.4
Post-History: 23-Feb-2013, 02-May-2013
Replaces: 354
Resolution: https://mail.python.org/pipermail/python-dev/2013-May/126112.html


Abstract
========

This PEP proposes adding an enumeration type to the Python standard library.

An enumeration is a set of symbolic names bound to unique, constant values.
Within an enumeration, the values can be compared by identity, and the
enumeration itself can be iterated over.


Status of discussions
=====================

The idea of adding an enum type to Python is not new - :pep:`354` is a
previous attempt that was rejected in 2005.  Recently a new set of discussions
was initiated [3]_ on the ``python-ideas`` mailing list.  Many new ideas were
proposed in several threads; after a lengthy discussion Guido proposed adding
``flufl.enum`` to the standard library [4]_.  During the PyCon 2013 language
summit the issue was discussed further.  It became clear that many developers
want to see an enum that subclasses ``int``, which can allow us to replace
many integer constants in the standard library by enums with friendly string
representations, without ceding backwards compatibility.  An additional
discussion among several interested core developers led to the proposal of
having ``IntEnum`` as a special case of ``Enum``.

The key dividing issue between ``Enum`` and ``IntEnum`` is whether comparing
to integers is semantically meaningful.  For most uses of enumerations, it's
a **feature** to reject comparison to integers; enums that compare to integers
lead, through transitivity, to comparisons between enums of unrelated types,
which isn't desirable in most cases.  For some uses, however, greater
interoperability with integers is desired. For instance, this is the case for
replacing existing standard library constants (such as ``socket.AF_INET``)
with enumerations.

Further discussion in late April 2013 led to the conclusion that enumeration
members should belong to the type of their enum: ``type(Color.red) == Color``.
Guido has pronounced a decision on this issue [5]_, as well as related issues
of not allowing to subclass enums [6]_, unless they define no enumeration
members [7]_.

The PEP was accepted by Guido on May 10th, 2013 [1]_.


Motivation
==========

*[Based partly on the Motivation stated in* :pep:`354`\ *]*

The properties of an enumeration are useful for defining an immutable, related
set of constant values that may or may not have a semantic meaning.  Classic
examples are days of the week (Sunday through Saturday) and school assessment
grades ('A' through 'D', and 'F').  Other examples include error status values
and states within a defined process.

It is possible to simply define a sequence of values of some other basic type,
such as ``int`` or ``str``, to represent discrete arbitrary values.  However,
an enumeration ensures that such values are distinct from any others including,
importantly, values within other enumerations, and that operations without
meaning ("Wednesday times two") are not defined for these values.  It also
provides a convenient printable representation of enum values without requiring
tedious repetition while defining them (i.e. no ``GREEN = 'green'``).


Module and type name
====================

We propose to add a module named ``enum`` to the standard library.  The main
type exposed by this module is ``Enum``.  Hence, to import the ``Enum`` type
user code will run::

    >>> from enum import Enum


Proposed semantics for the new enumeration type
===============================================

Creating an Enum
----------------

Enumerations are created using the class syntax, which makes them easy to read
and write.  An alternative creation method is described in `Functional API`_.
To define an enumeration, subclass ``Enum`` as follows::

    >>> from enum import Enum
    >>> class Color(Enum):
    ...     red = 1
    ...     green = 2
    ...     blue = 3

**A note on nomenclature**: we call ``Color`` an *enumeration* (or *enum*)
and ``Color.red``, ``Color.green`` are *enumeration members* (or
*enum members*).  Enumeration members also have *values* (the value of
``Color.red`` is ``1``, etc.)

Enumeration members have human readable string representations::

    >>> print(Color.red)
    Color.red

...while their ``repr`` has more information::

    >>> print(repr(Color.red))
    <Color.red: 1>

The *type* of an enumeration member is the enumeration it belongs to::

    >>> type(Color.red)
    <Enum 'Color'>
    >>> isinstance(Color.green, Color)
    True
    >>>

Enums also have a property that contains just their item name::

    >>> print(Color.red.name)
    red

Enumerations support iteration, in definition order::

    >>> class Shake(Enum):
    ...   vanilla = 7
    ...   chocolate = 4
    ...   cookies = 9
    ...   mint = 3
    ...
    >>> for shake in Shake:
    ...   print(shake)
    ...
    Shake.vanilla
    Shake.chocolate
    Shake.cookies
    Shake.mint

Enumeration members are hashable, so they can be used in dictionaries and sets::

    >>> apples = {}
    >>> apples[Color.red] = 'red delicious'
    >>> apples[Color.green] = 'granny smith'
    >>> apples
    {<Color.red: 1>: 'red delicious', <Color.green: 2>: 'granny smith'}


Programmatic access to enumeration members
------------------------------------------

Sometimes it's useful to access members in enumerations programmatically (i.e.
situations where ``Color.red`` won't do because the exact color is not known
at program-writing time).  ``Enum`` allows such access::

    >>> Color(1)
    <Color.red: 1>
    >>> Color(3)
    <Color.blue: 3>

If you want to access enum members by *name*, use item access::

    >>> Color['red']
    <Color.red: 1>
    >>> Color['green']
    <Color.green: 2>


Duplicating enum members and values
-----------------------------------

Having two enum members with the same name is invalid::

    >>> class Shape(Enum):
    ...   square = 2
    ...   square = 3
    ...
    Traceback (most recent call last):
    ...
    TypeError: Attempted to reuse key: square

However, two enum members are allowed to have the same value.  Given two members
A and B with the same value (and A defined first), B is an alias to A.  By-value
lookup of the value of A and B will return A.  By-name lookup of B will also
return A::

    >>> class Shape(Enum):
    ...   square = 2
    ...   diamond = 1
    ...   circle = 3
    ...   alias_for_square = 2
    ...
    >>> Shape.square
    <Shape.square: 2>
    >>> Shape.alias_for_square
    <Shape.square: 2>
    >>> Shape(2)
    <Shape.square: 2>

Iterating over the members of an enum does not provide the aliases::

    >>> list(Shape)
    [<Shape.square: 2>, <Shape.diamond: 1>, <Shape.circle: 3>]

The special attribute ``__members__`` is an ordered dictionary mapping names
to members.  It includes all names defined in the enumeration, including the
aliases::

    >>> for name, member in Shape.__members__.items():
    ...   name, member
    ...
    ('square', <Shape.square: 2>)
    ('diamond', <Shape.diamond: 1>)
    ('circle', <Shape.circle: 3>)
    ('alias_for_square', <Shape.square: 2>)

The ``__members__`` attribute can be used for detailed programmatic access to
the enumeration members.  For example, finding all the aliases::

    >>> [name for name, member in Shape.__members__.items() if member.name != name]
    ['alias_for_square']

Comparisons
-----------

Enumeration members are compared by identity::

    >>> Color.red is Color.red
    True
    >>> Color.red is Color.blue
    False
    >>> Color.red is not Color.blue
    True

Ordered comparisons between enumeration values are *not* supported.  Enums are
not integers (but see `IntEnum`_ below)::

    >>> Color.red < Color.blue
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    TypeError: unorderable types: Color() < Color()

Equality comparisons are defined though::

    >>> Color.blue == Color.red
    False
    >>> Color.blue != Color.red
    True
    >>> Color.blue == Color.blue
    True

Comparisons against non-enumeration values will always compare not equal
(again, ``IntEnum`` was explicitly designed to behave differently, see
below)::

    >>> Color.blue == 2
    False


Allowed members and attributes of enumerations
----------------------------------------------

The examples above use integers for enumeration values.  Using integers is
short and handy (and provided by default by the `Functional API`_), but not
strictly enforced.  In the vast majority of use-cases, one doesn't care what
the actual value of an enumeration is.  But if the value *is* important,
enumerations can have arbitrary values.

Enumerations are Python classes, and can have methods and special methods as
usual.  If we have this enumeration::

    class Mood(Enum):
      funky = 1
      happy = 3

      def describe(self):
        # self is the member here
        return self.name, self.value

      def __str__(self):
        return 'my custom str! {0}'.format(self.value)

      @classmethod
      def favorite_mood(cls):
        # cls here is the enumeration
        return cls.happy

Then::

    >>> Mood.favorite_mood()
    <Mood.happy: 3>
    >>> Mood.happy.describe()
    ('happy', 3)
    >>> str(Mood.funky)
    'my custom str! 1'

The rules for what is allowed are as follows: all attributes defined within an
enumeration will become members of this enumeration, with the exception of
*__dunder__* names and descriptors [9]_; methods are descriptors too.


Restricted subclassing of enumerations
--------------------------------------

Subclassing an enumeration is allowed only if the enumeration does not define
any members.  So this is forbidden::

    >>> class MoreColor(Color):
    ...   pink = 17
    ...
    TypeError: Cannot extend enumerations

But this is allowed::

    >>> class Foo(Enum):
    ...   def some_behavior(self):
    ...     pass
    ...
    >>> class Bar(Foo):
    ...   happy = 1
    ...   sad = 2
    ...

The rationale for this decision was given by Guido in [6]_.  Allowing to
subclass enums that define members would lead to a violation of some
important invariants of types and instances.  On the other hand, it
makes sense to allow sharing some common behavior between a group of
enumerations, and subclassing empty enumerations is also used to implement
``IntEnum``.


IntEnum
-------

A variation of ``Enum`` is proposed which is also a subclass of ``int``.
Members of an ``IntEnum`` can be compared to integers; by extension,
integer enumerations of different types can also be compared to each other::

    >>> from enum import IntEnum
    >>> class Shape(IntEnum):
    ...   circle = 1
    ...   square = 2
    ...
    >>> class Request(IntEnum):
    ...   post = 1
    ...   get = 2
    ...
    >>> Shape == 1
    False
    >>> Shape.circle == 1
    True
    >>> Shape.circle == Request.post
    True

However they still can't be compared to ``Enum``::

    >>> class Shape(IntEnum):
    ...   circle = 1
    ...   square = 2
    ...
    >>> class Color(Enum):
    ...   red = 1
    ...   green = 2
    ...
    >>> Shape.circle == Color.red
    False

``IntEnum`` values behave like integers in other ways you'd expect::

    >>> int(Shape.circle)
    1
    >>> ['a', 'b', 'c'][Shape.circle]
    'b'
    >>> [i for i in range(Shape.square)]
    [0, 1]

For the vast majority of code, ``Enum`` is strongly recommended,
since ``IntEnum`` breaks some semantic promises of an enumeration (by
being comparable to integers, and thus by transitivity to other
unrelated enumerations). It should be used only in special cases where
there's no other choice; for example, when integer constants are
replaced with enumerations and backwards compatibility is required
with code that still expects integers.


Other derived enumerations
--------------------------

``IntEnum`` will be part of the ``enum`` module.  However, it would be very
simple to implement independently::

    class IntEnum(int, Enum):
        pass

This demonstrates how similar derived enumerations can be defined, for example
a ``StrEnum`` that mixes in ``str`` instead of ``int``.

Some rules:

1. When subclassing Enum, mix-in types must appear before Enum itself in the
   sequence of bases, as in the ``IntEnum`` example above.
2. While Enum can have members of any type, once you mix in an additional
   type, all the members must have values of that type, e.g. ``int`` above.
   This restriction does not apply to mix-ins which only add methods
   and don't specify another data type such as ``int`` or ``str``.


Pickling
--------

Enumerations can be pickled and unpickled::

    >>> from enum.tests.fruit import Fruit
    >>> from pickle import dumps, loads
    >>> Fruit.tomato is loads(dumps(Fruit.tomato))
    True

The usual restrictions for pickling apply: picklable enums must be defined in
the top level of a module, since unpickling requires them to be importable
from that module.


Functional API
--------------

The ``Enum`` class is callable, providing the following functional API::

    >>> Animal = Enum('Animal', 'ant bee cat dog')
    >>> Animal
    <Enum 'Animal'>
    >>> Animal.ant
    <Animal.ant: 1>
    >>> Animal.ant.value
    1
    >>> list(Animal)
    [<Animal.ant: 1>, <Animal.bee: 2>, <Animal.cat: 3>, <Animal.dog: 4>]

The semantics of this API resemble ``namedtuple``. The first argument
of the call to ``Enum`` is the name of the enumeration.  Pickling enums
created with the functional API will work on CPython and PyPy, but for
IronPython and Jython you may need to specify the module name explicitly
as follows::

    >>> Animals = Enum('Animals', 'ant bee cat dog', module=__name__)

The second argument is the *source* of enumeration member names.  It can be a
whitespace-separated string of names, a sequence of names, a sequence of
2-tuples with key/value pairs, or a mapping (e.g. dictionary) of names to
values.  The last two options enable assigning arbitrary values to
enumerations; the others auto-assign increasing integers starting with 1.  A
new class derived from ``Enum`` is returned.  In other words, the above
assignment to ``Animal`` is equivalent to::

    >>> class Animals(Enum):
    ...   ant = 1
    ...   bee = 2
    ...   cat = 3
    ...   dog = 4

The reason for defaulting to ``1`` as the starting number and not ``0`` is
that ``0`` is ``False`` in a boolean sense, but enum members all evaluate
to ``True``.


Proposed variations
===================

Some variations were proposed during the discussions in the mailing list.
Here's some of the more popular ones.


flufl.enum
----------

``flufl.enum`` was the reference implementation upon which this PEP was
originally based.  Eventually, it was decided against the inclusion of
``flufl.enum`` because its design separated enumeration members from
enumerations, so the former are not instances of the latter.  Its design
also explicitly permits subclassing enumerations for extending them with
more members (due to the member/enum separation, the type invariants are not
violated in ``flufl.enum`` with such a scheme).


Not having to specify values for enums
--------------------------------------

Michael Foord proposed (and Tim Delaney provided a proof-of-concept
implementation) to use metaclass magic that makes this possible::

    class Color(Enum):
        red, green, blue

The values get actually assigned only when first looked up.

Pros: cleaner syntax that requires less typing for a very common task (just
listing enumeration names without caring about the values).

Cons: involves much magic in the implementation, which makes even the
definition of such enums baffling when first seen.  Besides, explicit is
better than implicit.

Using special names or forms to auto-assign enum values
-------------------------------------------------------

A different approach to avoid specifying enum values is to use a special name
or form to auto assign them.  For example::

    class Color(Enum):
        red = None          # auto-assigned to 0
        green = None        # auto-assigned to 1
        blue = None         # auto-assigned to 2

More flexibly::

    class Color(Enum):
        red = 7
        green = None        # auto-assigned to 8
        blue = 19
        purple = None       # auto-assigned to 20

Some variations on this theme:

#. A special name ``auto`` imported from the enum package.
#. Georg Brandl proposed ellipsis (``...``) instead of ``None`` to achieve the
   same effect.

Pros: no need to manually enter values. Makes it easier to change the enum and
extend it, especially for large enumerations.

Cons: actually longer to type in many simple cases.  The argument of explicit
vs. implicit applies here as well.


Use-cases in the standard library
=================================

The Python standard library has many places where the usage of enums would be
beneficial to replace other idioms currently used to represent them.  Such
usages can be divided to two categories: user-code facing constants, and
internal constants.

User-code facing constants like ``os.SEEK_*``, ``socket`` module constants,
decimal rounding modes and HTML error codes could require backwards
compatibility since user code may expect integers.  ``IntEnum`` as described
above provides the required semantics; being a subclass of ``int``, it does not
affect user code that expects integers, while on the other hand allowing
printable representations for enumeration values::

    >>> import socket
    >>> family = socket.AF_INET
    >>> family == 2
    True
    >>> print(family)
    SocketFamily.AF_INET

Internal constants are not seen by user code but are employed internally by
stdlib modules.  These can be implemented with ``Enum``.  Some examples
uncovered by a very partial skim through the stdlib: ``binhex``, ``imaplib``,
``http/client``, ``urllib/robotparser``, ``idlelib``, ``concurrent.futures``,
``turtledemo``.

In addition, looking at the code of the Twisted library, there are many use
cases for replacing internal state constants with enums.  The same can be said
about a lot of networking code (especially implementation of protocols) and
can be seen in test protocols written with the Tulip library as well.


Acknowledgments
===============

This PEP was initially proposing including the ``flufl.enum`` package [8]_
by Barry Warsaw into the stdlib, and is inspired in large parts by it.
Ben Finney is the author of the earlier enumeration :pep:`354`.


References
==========

.. [1] https://mail.python.org/pipermail/python-dev/2013-May/126112.html
.. [3] https://mail.python.org/pipermail/python-ideas/2013-January/019003.html
.. [4] https://mail.python.org/pipermail/python-ideas/2013-February/019373.html
.. [5] To make enums behave similarly to Python classes like bool, and
       behave in a more intuitive way.  It would be surprising if the type of
       ``Color.red`` would not be ``Color``. (Discussion in
       https://mail.python.org/pipermail/python-dev/2013-April/125687.html)
.. [6] Subclassing enums and adding new members creates an unresolvable
       situation; on one hand ``MoreColor.red`` and ``Color.red`` should
       not be the same object, and on the other ``isinstance`` checks become
       confusing if they are not.  The discussion also links to Stack Overflow
       discussions that make additional arguments.
       (https://mail.python.org/pipermail/python-dev/2013-April/125716.html)
.. [7] It may be useful to have a class defining some behavior (methods, with
       no actual enumeration members) mixed into an enum, and this would not
       create the problem discussed in [6]_.  (Discussion in
       https://mail.python.org/pipermail/python-dev/2013-May/125859.html)
.. [8] http://pythonhosted.org/flufl.enum/
.. [9] http://docs.python.org/3/howto/descriptor.html


Copyright
=========

This document has been placed in the public domain.


..
   Local Variables:
   mode: indented-text
   indent-tabs-mode: nil
   sentence-end-double-space: t
   fill-column: 70
   coding: utf-8
   End: