cpython/Lib/enum.py

2054 lines
77 KiB
Python
Raw Normal View History

import sys
import builtins as bltns
from types import MappingProxyType, DynamicClassAttribute
__all__ = [
'EnumType', 'EnumMeta',
'Enum', 'IntEnum', 'StrEnum', 'Flag', 'IntFlag', 'ReprEnum',
'auto', 'unique', 'property', 'verify', 'member', 'nonmember',
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
'FlagBoundary', 'STRICT', 'CONFORM', 'EJECT', 'KEEP',
'global_flag_repr', 'global_enum_repr', 'global_str', 'global_enum',
'EnumCheck', 'CONTINUOUS', 'NAMED_FLAGS', 'UNIQUE',
'pickle_by_global_name', 'pickle_by_enum_name',
]
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
# Dummy value for Enum and Flag as there are explicit checks for them
# before they have been created.
# This is also why there are checks in EnumType like `if Enum is not None`
Enum = Flag = EJECT = _stdlib_enums = ReprEnum = None
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
class nonmember(object):
"""
2022-09-21 07:31:05 +08:00
Protects item from becoming an Enum member during class creation.
"""
def __init__(self, value):
self.value = value
class member(object):
"""
2022-09-21 07:31:05 +08:00
Forces item to become an Enum member during class creation.
"""
def __init__(self, value):
self.value = value
def _is_descriptor(obj):
"""
Returns True if obj is a descriptor, False otherwise.
"""
return (
hasattr(obj, '__get__') or
hasattr(obj, '__set__') or
hasattr(obj, '__delete__')
)
def _is_dunder(name):
"""
Returns True if a __dunder__ name, False otherwise.
"""
return (
len(name) > 4 and
name[:2] == name[-2:] == '__' and
name[2] != '_' and
name[-3] != '_'
)
def _is_sunder(name):
"""
Returns True if a _sunder_ name, False otherwise.
"""
return (
len(name) > 2 and
name[0] == name[-1] == '_' and
name[1] != '_' and
name[-2] != '_'
)
def _is_internal_class(cls_name, obj):
# do not use `re` as `re` imports `enum`
if not isinstance(obj, type):
return False
qualname = getattr(obj, '__qualname__', '')
s_pattern = cls_name + '.' + getattr(obj, '__name__', '')
e_pattern = '.' + s_pattern
return qualname == s_pattern or qualname.endswith(e_pattern)
def _is_private(cls_name, name):
# do not use `re` as `re` imports `enum`
pattern = '_%s__' % (cls_name, )
pat_len = len(pattern)
if (
len(name) > pat_len
and name.startswith(pattern)
and (name[-1] != '_' or name[-2] != '_')
):
return True
else:
return False
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def _is_single_bit(num):
"""
True if only one bit set in num (should be an int)
"""
if num == 0:
return False
num &= num - 1
return num == 0
def _make_class_unpicklable(obj):
"""
Make the given obj un-picklable.
obj should be either a dictionary, or an Enum
"""
def _break_on_call_reduce(self, proto):
raise TypeError('%r cannot be pickled' % self)
if isinstance(obj, dict):
obj['__reduce_ex__'] = _break_on_call_reduce
obj['__module__'] = '<unknown>'
else:
setattr(obj, '__reduce_ex__', _break_on_call_reduce)
setattr(obj, '__module__', '<unknown>')
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def _iter_bits_lsb(num):
# num must be a positive integer
original = num
if isinstance(num, Enum):
num = num.value
if num < 0:
raise ValueError('%r is not a positive integer' % original)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
while num:
b = num & (~num + 1)
yield b
num ^= b
def show_flag_values(value):
return list(_iter_bits_lsb(value))
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def bin(num, max_bits=None):
"""
Like built-in bin(), except negative values are represented in
twos-compliment, and the leading bit always indicates sign
(0=positive, 1=negative).
>>> bin(10)
'0b0 1010'
>>> bin(~10) # ~10 is -11
'0b1 0101'
"""
ceiling = 2 ** (num).bit_length()
if num >= 0:
s = bltns.bin(num + ceiling).replace('1', '0', 1)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
else:
s = bltns.bin(~num ^ (ceiling - 1) + ceiling)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
sign = s[:3]
digits = s[3:]
if max_bits is not None:
if len(digits) < max_bits:
digits = (sign[-1] * max_bits + digits)[-max_bits:]
return "%s %s" % (sign, digits)
def _dedent(text):
"""
Like textwrap.dedent. Rewritten because we cannot import textwrap.
"""
lines = text.split('\n')
for i, ch in enumerate(lines[0]):
if ch != ' ':
break
for j, l in enumerate(lines):
lines[j] = l[i:]
return '\n'.join(lines)
class _auto_null:
def __repr__(self):
return '_auto_null'
_auto_null = _auto_null()
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
class auto:
"""
Instances are replaced with an appropriate value in Enum class suites.
"""
def __init__(self, value=_auto_null):
self.value = value
def __repr__(self):
return "auto(%r)" % self.value
class property(DynamicClassAttribute):
"""
This is a descriptor, used to define attributes that act differently
when accessed through an enum member and through an enum class.
Instance access is the same as property(), but access to an attribute
through the enum class will instead look in the class' _member_map_ for
a corresponding enum member.
"""
member = None
_attr_type = None
_cls_type = None
def __get__(self, instance, ownerclass=None):
if instance is None:
if self.member is not None:
return self.member
else:
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
raise AttributeError(
'%r has no attribute %r' % (ownerclass, self.name)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
)
if self.fget is not None:
# use previous enum.property
return self.fget(instance)
elif self._attr_type == 'attr':
# look up previous attibute
return getattr(self._cls_type, self.name)
elif self._attr_type == 'desc':
# use previous descriptor
return getattr(instance._value_, self.name)
# look for a member by this name.
try:
return ownerclass._member_map_[self.name]
except KeyError:
raise AttributeError(
'%r has no attribute %r' % (ownerclass, self.name)
) from None
def __set__(self, instance, value):
if self.fset is not None:
return self.fset(instance, value)
raise AttributeError(
"<enum %r> cannot set attribute %r" % (self.clsname, self.name)
)
def __delete__(self, instance):
if self.fdel is not None:
return self.fdel(instance)
raise AttributeError(
"<enum %r> cannot delete attribute %r" % (self.clsname, self.name)
)
def __set_name__(self, ownerclass, name):
self.name = name
self.clsname = ownerclass.__name__
class _proto_member:
"""
intermediate step for enum members between class execution and final creation
"""
def __init__(self, value):
self.value = value
def __set_name__(self, enum_class, member_name):
"""
convert each quasi-member into an instance of the new enum class
"""
# first step: remove ourself from enum_class
delattr(enum_class, member_name)
# second step: create member based on enum_class
value = self.value
if not isinstance(value, tuple):
args = (value, )
else:
args = value
if enum_class._member_type_ is tuple: # special case for tuple enums
args = (args, ) # wrap it one more time
if not enum_class._use_args_:
enum_member = enum_class._new_member_(enum_class)
else:
enum_member = enum_class._new_member_(enum_class, *args)
if not hasattr(enum_member, '_value_'):
if enum_class._member_type_ is object:
enum_member._value_ = value
else:
try:
enum_member._value_ = enum_class._member_type_(*args)
except Exception as exc:
new_exc = TypeError(
'_value_ not set in __new__, unable to create it'
)
new_exc.__cause__ = exc
raise new_exc
value = enum_member._value_
enum_member._name_ = member_name
enum_member.__objclass__ = enum_class
enum_member.__init__(*args)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
enum_member._sort_order_ = len(enum_class._member_names_)
if Flag is not None and issubclass(enum_class, Flag):
enum_class._flag_mask_ |= value
if _is_single_bit(value):
enum_class._singles_mask_ |= value
enum_class._all_bits_ = 2 ** ((enum_class._flag_mask_).bit_length()) - 1
# If another member with the same value was already defined, the
# new member becomes an alias to the existing one.
try:
try:
# try to do a fast lookup to avoid the quadratic loop
enum_member = enum_class._value2member_map_[value]
except TypeError:
for name, canonical_member in enum_class._member_map_.items():
if canonical_member._value_ == value:
enum_member = canonical_member
break
else:
raise KeyError
except KeyError:
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
# this could still be an alias if the value is multi-bit and the
# class is a flag class
if (
Flag is None
or not issubclass(enum_class, Flag)
):
# no other instances found, record this member in _member_names_
enum_class._member_names_.append(member_name)
elif (
Flag is not None
and issubclass(enum_class, Flag)
and _is_single_bit(value)
):
# no other instances found, record this member in _member_names_
enum_class._member_names_.append(member_name)
# if necessary, get redirect in place and then add it to _member_map_
found_descriptor = None
descriptor_type = None
class_type = None
for base in enum_class.__mro__[1:]:
attr = base.__dict__.get(member_name)
if attr is not None:
if isinstance(attr, (property, DynamicClassAttribute)):
found_descriptor = attr
class_type = base
descriptor_type = 'enum'
break
elif _is_descriptor(attr):
found_descriptor = attr
descriptor_type = descriptor_type or 'desc'
class_type = class_type or base
continue
else:
descriptor_type = 'attr'
class_type = base
if found_descriptor:
redirect = property()
redirect.member = enum_member
redirect.__set_name__(enum_class, member_name)
if descriptor_type in ('enum','desc'):
# earlier descriptor found; copy fget, fset, fdel to this one.
redirect.fget = getattr(found_descriptor, 'fget', None)
redirect._get = getattr(found_descriptor, '__get__', None)
redirect.fset = getattr(found_descriptor, 'fset', None)
redirect._set = getattr(found_descriptor, '__set__', None)
redirect.fdel = getattr(found_descriptor, 'fdel', None)
redirect._del = getattr(found_descriptor, '__delete__', None)
redirect._attr_type = descriptor_type
redirect._cls_type = class_type
setattr(enum_class, member_name, redirect)
else:
setattr(enum_class, member_name, enum_member)
# now add to _member_map_ (even aliases)
enum_class._member_map_[member_name] = enum_member
try:
# This may fail if value is not hashable. We can't add the value
# to the map, and by-value lookups for this value will be
# linear.
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
enum_class._value2member_map_.setdefault(value, enum_member)
except TypeError:
# keep track of the value in a list so containment checks are quick
enum_class._unhashable_values_.append(value)
class _EnumDict(dict):
"""
Track enum member order and ensure member names are not reused.
EnumType will use the names found in self._member_names as the
enumeration member names.
"""
def __init__(self):
super().__init__()
self._member_names = {} # use a dict to keep insertion order
self._last_values = []
self._ignore = []
self._auto_called = False
def __setitem__(self, key, value):
"""
Changes anything not dundered or not a descriptor.
If an enum member name is used twice, an error is raised; duplicate
values are not checked for.
Single underscore (sunder) names are reserved.
"""
if _is_private(self._cls_name, key):
# do nothing, name will be a normal attribute
pass
elif _is_sunder(key):
if key not in (
2021-04-15 21:49:54 +08:00
'_order_',
'_generate_next_value_', '_numeric_repr_', '_missing_', '_ignore_',
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
'_iter_member_', '_iter_member_by_value_', '_iter_member_by_def_',
):
raise ValueError(
'_sunder_ names, such as %r, are reserved for future Enum use'
% (key, )
)
if key == '_generate_next_value_':
# check if members already defined as auto()
if self._auto_called:
raise TypeError("_generate_next_value_ must be defined before members")
_gnv = value.__func__ if isinstance(value, staticmethod) else value
setattr(self, '_generate_next_value', _gnv)
elif key == '_ignore_':
if isinstance(value, str):
value = value.replace(',',' ').split()
else:
value = list(value)
self._ignore = value
already = set(value) & set(self._member_names)
if already:
raise ValueError(
'_ignore_ cannot specify already set names: %r'
% (already, )
)
elif _is_dunder(key):
if key == '__order__':
key = '_order_'
elif key in self._member_names:
# descriptor overwriting an enum?
raise TypeError('%r already defined as %r' % (key, self[key]))
elif key in self._ignore:
pass
elif isinstance(value, nonmember):
# unwrap value here; it won't be processed by the below `else`
value = value.value
elif _is_descriptor(value):
pass
elif _is_internal_class(self._cls_name, value):
# do nothing, name will be a normal attribute
pass
else:
if key in self:
# enum overwriting a descriptor?
raise TypeError('%r already defined as %r' % (key, self[key]))
elif isinstance(value, member):
# unwrap value here -- it will become a member
value = value.value
non_auto_store = True
single = False
if isinstance(value, auto):
single = True
value = (value, )
if type(value) is tuple and any(isinstance(v, auto) for v in value):
# insist on an actual tuple, no subclasses, in keeping with only supporting
# top-level auto() usage (not contained in any other data structure)
auto_valued = []
for v in value:
if isinstance(v, auto):
non_auto_store = False
if v.value == _auto_null:
v.value = self._generate_next_value(
key, 1, len(self._member_names), self._last_values[:],
)
self._auto_called = True
v = v.value
self._last_values.append(v)
auto_valued.append(v)
if single:
value = auto_valued[0]
else:
value = tuple(auto_valued)
self._member_names[key] = None
if non_auto_store:
self._last_values.append(value)
super().__setitem__(key, value)
def update(self, members, **more_members):
try:
for name in members.keys():
self[name] = members[name]
except AttributeError:
for name, value in members:
self[name] = value
for name, value in more_members.items():
self[name] = value
class EnumType(type):
"""
Metaclass for Enum
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
@classmethod
def __prepare__(metacls, cls, bases, **kwds):
# check that previous enum members do not exist
metacls._check_for_existing_members_(cls, bases)
# create the namespace dict
enum_dict = _EnumDict()
enum_dict._cls_name = cls
# inherit previous flags and _generate_next_value_ function
member_type, first_enum = metacls._get_mixins_(cls, bases)
if first_enum is not None:
enum_dict['_generate_next_value_'] = getattr(
first_enum, '_generate_next_value_', None,
)
return enum_dict
2016-08-20 15:00:52 +08:00
def __new__(metacls, cls, bases, classdict, *, boundary=None, _simple=False, **kwds):
# an Enum class is final once enumeration items have been defined; it
# cannot be mixed with other types (int, float, etc.) if it has an
# inherited __new__ unless a new __new__ is defined (or the resulting
# class will fail).
#
if _simple:
return super().__new__(metacls, cls, bases, classdict, **kwds)
#
# remove any keys listed in _ignore_
classdict.setdefault('_ignore_', []).append('_ignore_')
ignore = classdict['_ignore_']
for key in ignore:
classdict.pop(key, None)
#
# grab member names
member_names = classdict._member_names
#
# check for illegal enum names (any others?)
invalid_names = set(member_names) & {'mro', ''}
if invalid_names:
raise ValueError('invalid enum member name(s) %s' % (
','.join(repr(n) for n in invalid_names)
))
#
# adjust the sunders
_order_ = classdict.pop('_order_', None)
_gnv = classdict.get('_generate_next_value_')
if _gnv is not None and type(_gnv) is not staticmethod:
_gnv = staticmethod(_gnv)
# convert to normal dict
classdict = dict(classdict.items())
if _gnv is not None:
classdict['_generate_next_value_'] = _gnv
#
# data type of member and the controlling Enum class
member_type, first_enum = metacls._get_mixins_(cls, bases)
__new__, save_new, use_args = metacls._find_new_(
classdict, member_type, first_enum,
)
classdict['_new_member_'] = __new__
classdict['_use_args_'] = use_args
#
# convert future enum members into temporary _proto_members
for name in member_names:
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
value = classdict[name]
classdict[name] = _proto_member(value)
#
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
# house-keeping structures
classdict['_member_names_'] = []
classdict['_member_map_'] = {}
classdict['_value2member_map_'] = {}
classdict['_unhashable_values_'] = []
classdict['_member_type_'] = member_type
# now set the __repr__ for the value
classdict['_value_repr_'] = metacls._find_data_repr_(cls, bases)
#
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
# Flag structures (will be removed if final class is not a Flag
classdict['_boundary_'] = (
boundary
or getattr(first_enum, '_boundary_', None)
)
classdict['_flag_mask_'] = 0
classdict['_singles_mask_'] = 0
classdict['_all_bits_'] = 0
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
classdict['_inverted_'] = None
try:
exc = None
enum_class = super().__new__(metacls, cls, bases, classdict, **kwds)
except RuntimeError as e:
# any exceptions raised by member.__new__ will get converted to a
# RuntimeError, so get that original exception back and raise it instead
exc = e.__cause__ or e
if exc is not None:
raise exc
#
# update classdict with any changes made by __init_subclass__
classdict.update(enum_class.__dict__)
#
# double check that repr and friends are not the mixin's or various
# things break (such as pickle)
# however, if the method is defined in the Enum itself, don't replace
# it
#
# Also, special handling for ReprEnum
if ReprEnum is not None and ReprEnum in bases:
if member_type is object:
raise TypeError(
'ReprEnum subclasses must be mixed with a data type (i.e.'
' int, str, float, etc.)'
)
if '__format__' not in classdict:
enum_class.__format__ = member_type.__format__
classdict['__format__'] = enum_class.__format__
if '__str__' not in classdict:
method = member_type.__str__
if method is object.__str__:
# if member_type does not define __str__, object.__str__ will use
# its __repr__ instead, so we'll also use its __repr__
method = member_type.__repr__
enum_class.__str__ = method
classdict['__str__'] = enum_class.__str__
for name in ('__repr__', '__str__', '__format__', '__reduce_ex__'):
if name not in classdict:
# check for mixin overrides before replacing
enum_method = getattr(first_enum, name)
found_method = getattr(enum_class, name)
object_method = getattr(object, name)
data_type_method = getattr(member_type, name)
if found_method in (data_type_method, object_method):
setattr(enum_class, name, enum_method)
#
# for Flag, add __or__, __and__, __xor__, and __invert__
if Flag is not None and issubclass(enum_class, Flag):
for name in (
'__or__', '__and__', '__xor__',
'__ror__', '__rand__', '__rxor__',
'__invert__'
):
if name not in classdict:
enum_method = getattr(Flag, name)
setattr(enum_class, name, enum_method)
classdict[name] = enum_method
#
# replace any other __new__ with our own (as long as Enum is not None,
# anyway) -- again, this is to support pickle
if Enum is not None:
# if the user defined their own __new__, save it before it gets
# clobbered in case they subclass later
if save_new:
enum_class.__new_member__ = __new__
enum_class.__new__ = Enum.__new__
#
# py3 support for definition order (helps keep py2/py3 code in sync)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
#
# _order_ checking is spread out into three/four steps
# - if enum_class is a Flag:
# - remove any non-single-bit flags from _order_
# - remove any aliases from _order_
# - check that _order_ and _member_names_ match
#
# step 1: ensure we have a list
if _order_ is not None:
if isinstance(_order_, str):
_order_ = _order_.replace(',', ' ').split()
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
#
# remove Flag structures if final class is not a Flag
if (
Flag is None and cls != 'Flag'
or Flag is not None and not issubclass(enum_class, Flag)
):
delattr(enum_class, '_boundary_')
delattr(enum_class, '_flag_mask_')
delattr(enum_class, '_singles_mask_')
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
delattr(enum_class, '_all_bits_')
delattr(enum_class, '_inverted_')
elif Flag is not None and issubclass(enum_class, Flag):
# set correct __iter__
member_list = [m._value_ for m in enum_class]
if member_list != sorted(member_list):
enum_class._iter_member_ = enum_class._iter_member_by_def_
if _order_:
# _order_ step 2: remove any items from _order_ that are not single-bit
_order_ = [
o
for o in _order_
if o not in enum_class._member_map_ or _is_single_bit(enum_class[o]._value_)
]
#
if _order_:
# _order_ step 3: remove aliases from _order_
_order_ = [
o
for o in _order_
if (
o not in enum_class._member_map_
or
(o in enum_class._member_map_ and o in enum_class._member_names_)
)]
# _order_ step 4: verify that _order_ and _member_names_ match
if _order_ != enum_class._member_names_:
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
raise TypeError(
'member order does not match _order_:\n %r\n %r'
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
% (enum_class._member_names_, _order_)
)
#
return enum_class
def __bool__(cls):
"""
classes/types should always be True.
"""
return True
def __call__(cls, value, names=None, *values, module=None, qualname=None, type=None, start=1, boundary=None):
"""
Either returns an existing member, or creates a new enum class.
This method is used both when an enum class is given a value to match
to an enumeration member (i.e. Color(3)) and for the functional API
(i.e. Color = Enum('Color', names='RED GREEN BLUE')).
The value lookup branch is chosen if the enum is final.
When used for the functional API:
`value` will be the name of the new class.
`names` should be either a string of white-space/comma delimited names
(values will start at `start`), or an iterator/mapping of name, value pairs.
`module` should be set to the module this class is being created in;
if it is not set, an attempt to find that module will be made, but if
it fails the class will not be picklable.
`qualname` should be set to the actual location this class can be found
at in its module; by default it is set to the global scope. If this is
not correct, unpickling will fail in some circumstances.
`type`, if set, will be mixed in as the first base class.
"""
if cls._member_map_:
# simple value lookup if members exist
if names:
value = (value, names) + values
return cls.__new__(cls, value)
# otherwise, functional API: we're creating a new Enum type
if names is None and type is None:
# no body? no data-type? possibly wrong usage
raise TypeError(
f"{cls} has no members; specify `names=()` if you meant to create a new, empty, enum"
)
return cls._create_(
class_name=value,
names=names,
module=module,
qualname=qualname,
type=type,
start=start,
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
boundary=boundary,
)
def __contains__(cls, value):
"""Return True if `value` is in `cls`.
`value` is in `cls` if:
1) `value` is a member of `cls`, or
2) `value` is the value of one of the `cls`'s members.
"""
if isinstance(value, cls):
return True
return value in cls._value2member_map_ or value in cls._unhashable_values_
def __delattr__(cls, attr):
# nicer error message when someone tries to delete an attribute
# (see issue19025).
if attr in cls._member_map_:
raise AttributeError("%r cannot delete member %r." % (cls.__name__, attr))
super().__delattr__(attr)
def __dir__(cls):
interesting = set([
'__class__', '__contains__', '__doc__', '__getitem__',
'__iter__', '__len__', '__members__', '__module__',
'__name__', '__qualname__',
]
+ cls._member_names_
)
if cls._new_member_ is not object.__new__:
interesting.add('__new__')
if cls.__init_subclass__ is not object.__init_subclass__:
interesting.add('__init_subclass__')
if cls._member_type_ is object:
return sorted(interesting)
else:
# return whatever mixed-in data type has
return sorted(set(dir(cls._member_type_)) | interesting)
def __getitem__(cls, name):
"""
Return the member matching `name`.
"""
return cls._member_map_[name]
def __iter__(cls):
"""
Return members in definition order.
"""
return (cls._member_map_[name] for name in cls._member_names_)
def __len__(cls):
"""
Return the number of members (no aliases)
"""
return len(cls._member_names_)
@bltns.property
def __members__(cls):
"""
Returns a mapping of member name->value.
This mapping lists all enum members, including aliases. Note that this
is a read-only view of the internal mapping.
"""
return MappingProxyType(cls._member_map_)
def __repr__(cls):
if Flag is not None and issubclass(cls, Flag):
return "<flag %r>" % cls.__name__
else:
return "<enum %r>" % cls.__name__
def __reversed__(cls):
"""
Return members in reverse definition order.
"""
return (cls._member_map_[name] for name in reversed(cls._member_names_))
def __setattr__(cls, name, value):
"""
Block attempts to reassign Enum members.
A simple assignment to the class namespace only changes one of the
several possible ways to get an Enum member from the Enum class,
resulting in an inconsistent Enumeration.
"""
member_map = cls.__dict__.get('_member_map_', {})
if name in member_map:
raise AttributeError('cannot reassign member %r' % (name, ))
super().__setattr__(name, value)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def _create_(cls, class_name, names, *, module=None, qualname=None, type=None, start=1, boundary=None):
"""
Convenience method to create a new Enum class.
`names` can be:
* A string containing member names, separated either with spaces or
commas. Values are incremented by 1 from `start`.
* An iterable of member names. Values are incremented by 1 from `start`.
* An iterable of (member name, value) pairs.
* A mapping of member name -> value pairs.
"""
metacls = cls.__class__
bases = (cls, ) if type is None else (type, cls)
_, first_enum = cls._get_mixins_(class_name, bases)
classdict = metacls.__prepare__(class_name, bases)
# special processing needed for names?
if isinstance(names, str):
names = names.replace(',', ' ').split()
if isinstance(names, (tuple, list)) and names and isinstance(names[0], str):
original_names, names = names, []
last_values = []
for count, name in enumerate(original_names):
value = first_enum._generate_next_value_(name, start, count, last_values[:])
last_values.append(value)
names.append((name, value))
if names is None:
names = ()
# Here, names is either an iterable of (name, value) or a mapping.
for item in names:
if isinstance(item, str):
member_name, member_value = item, names[item]
else:
member_name, member_value = item
classdict[member_name] = member_value
if module is None:
try:
module = sys._getframemodulename(2)
except AttributeError:
# Fall back on _getframe if _getframemodulename is missing
try:
module = sys._getframe(2).f_globals['__name__']
except (AttributeError, ValueError, KeyError):
pass
if module is None:
_make_class_unpicklable(classdict)
else:
classdict['__module__'] = module
if qualname is not None:
classdict['__qualname__'] = qualname
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
return metacls.__new__(metacls, class_name, bases, classdict, boundary=boundary)
def _convert_(cls, name, module, filter, source=None, *, boundary=None, as_global=False):
"""
Create a new Enum subclass that replaces a collection of global constants
"""
# convert all constants from source (or module) that pass filter() to
# a new Enum called name, and export the enum and its members back to
# module;
# also, replace the __reduce_ex__ method so unpickling works in
# previous Python versions
module_globals = sys.modules[module].__dict__
if source:
source = source.__dict__
else:
source = module_globals
# _value2member_map_ is populated in the same order every time
# for a consistent reverse mapping of number to name when there
# are multiple names for the same number.
members = [
(name, value)
for name, value in source.items()
if filter(name)]
try:
# sort by value
members.sort(key=lambda t: (t[1], t[0]))
except TypeError:
# unless some values aren't comparable, in which case sort by name
members.sort(key=lambda t: t[0])
body = {t[0]: t[1] for t in members}
body['__module__'] = module
tmp_cls = type(name, (object, ), body)
cls = _simple_enum(etype=cls, boundary=boundary or KEEP)(tmp_cls)
if as_global:
global_enum(cls)
else:
sys.modules[cls.__module__].__dict__.update(cls.__members__)
module_globals[name] = cls
return cls
@classmethod
def _check_for_existing_members_(mcls, class_name, bases):
for chain in bases:
for base in chain.__mro__:
if isinstance(base, EnumType) and base._member_names_:
raise TypeError(
"<enum %r> cannot extend %r"
% (class_name, base)
)
@classmethod
def _get_mixins_(mcls, class_name, bases):
"""
Returns the type for creating enum members, and the first inherited
enum class.
bases: the tuple of bases that was given to __new__
"""
if not bases:
return object, Enum
# ensure final parent class is an Enum derivative, find any concrete
# data type, and check that Enum has no members
first_enum = bases[-1]
if not isinstance(first_enum, EnumType):
raise TypeError("new enumerations should be created as "
"`EnumName([mixin_type, ...] [data_type,] enum_type)`")
member_type = mcls._find_data_type_(class_name, bases) or object
return member_type, first_enum
@classmethod
def _find_data_repr_(mcls, class_name, bases):
for chain in bases:
for base in chain.__mro__:
if base is object:
continue
elif isinstance(base, EnumType):
# if we hit an Enum, use it's _value_repr_
return base._value_repr_
elif '__repr__' in base.__dict__:
# this is our data repr
# double-check if a dataclass with a default __repr__
if (
'__dataclass_fields__' in base.__dict__
and '__dataclass_params__' in base.__dict__
and base.__dict__['__dataclass_params__'].repr
):
return _dataclass_repr
else:
return base.__dict__['__repr__']
return None
@classmethod
def _find_data_type_(mcls, class_name, bases):
# a datatype has a __new__ method, or a __dataclass_fields__ attribute
data_types = set()
base_chain = set()
for chain in bases:
candidate = None
for base in chain.__mro__:
base_chain.add(base)
if base is object:
continue
elif isinstance(base, EnumType):
if base._member_type_ is not object:
data_types.add(base._member_type_)
break
elif '__new__' in base.__dict__ or '__dataclass_fields__' in base.__dict__:
data_types.add(candidate or base)
break
else:
candidate = candidate or base
if len(data_types) > 1:
raise TypeError('too many data types for %r: %r' % (class_name, data_types))
elif data_types:
return data_types.pop()
else:
return None
@classmethod
def _find_new_(mcls, classdict, member_type, first_enum):
"""
Returns the __new__ to be used for creating the enum members.
classdict: the class dictionary given to __new__
member_type: the data type whose __new__ will be used by default
first_enum: enumeration to check for an overriding __new__
"""
# now find the correct __new__, checking to see of one was defined
# by the user; also check earlier enum classes in case a __new__ was
# saved as __new_member__
__new__ = classdict.get('__new__', None)
# should __new__ be saved as __new_member__ later?
save_new = first_enum is not None and __new__ is not None
if __new__ is None:
# check all possibles for __new_member__ before falling back to
# __new__
for method in ('__new_member__', '__new__'):
for possible in (member_type, first_enum):
target = getattr(possible, method, None)
if target not in {
None,
None.__new__,
object.__new__,
Enum.__new__,
}:
__new__ = target
break
if __new__ is not None:
break
else:
__new__ = object.__new__
# if a non-object.__new__ is used then whatever value/tuple was
# assigned to the enum member name will be passed to __new__ and to the
# new enum member's __init__
if first_enum is None or __new__ in (Enum.__new__, object.__new__):
use_args = False
else:
use_args = True
return __new__, save_new, use_args
EnumMeta = EnumType
class Enum(metaclass=EnumType):
"""
Create a collection of name/value pairs.
Example enumeration:
>>> class Color(Enum):
... RED = 1
... BLUE = 2
... GREEN = 3
Access them by:
- attribute access:
>>> Color.RED
<Color.RED: 1>
- value lookup:
>>> Color(1)
<Color.RED: 1>
- name lookup:
>>> Color['RED']
<Color.RED: 1>
Enumerations can be iterated over, and know how many members they have:
>>> len(Color)
3
>>> list(Color)
[<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]
Methods can be added to enumerations, and members can have their own
attributes -- see the documentation for details.
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
@classmethod
def __signature__(cls):
if cls._member_names_:
return '(*values)'
else:
return '(new_class_name, /, names, *, module=None, qualname=None, type=None, start=1, boundary=None)'
def __new__(cls, value):
# all enum instances are actually created during class construction
# without calling this method; this method is called by the metaclass'
# __call__ (i.e. Color(3) ), and by pickle
if type(value) is cls:
# For lookups like Color(Color.RED)
return value
# by-value search for a matching enum member
# see if it's in the reverse mapping (for hashable values)
try:
return cls._value2member_map_[value]
except KeyError:
# Not found, no need to do long O(n) search
pass
except TypeError:
# not there, now do long search -- O(n) behavior
for member in cls._member_map_.values():
if member._value_ == value:
return member
# still not found -- verify that members exist, in-case somebody got here mistakenly
# (such as via super when trying to override __new__)
if not cls._member_map_:
raise TypeError("%r has no members defined" % cls)
#
# still not found -- try _missing_ hook
try:
exc = None
result = cls._missing_(value)
except Exception as e:
exc = e
result = None
try:
if isinstance(result, cls):
return result
elif (
Flag is not None and issubclass(cls, Flag)
and cls._boundary_ is EJECT and isinstance(result, int)
):
return result
else:
ve_exc = ValueError("%r is not a valid %s" % (value, cls.__qualname__))
if result is None and exc is None:
raise ve_exc
elif exc is None:
exc = TypeError(
'error in %s._missing_: returned %r instead of None or a valid member'
% (cls.__name__, result)
)
if not isinstance(exc, ValueError):
exc.__context__ = ve_exc
raise exc
finally:
# ensure all variables that could hold an exception are destroyed
exc = None
ve_exc = None
def __init__(self, *args, **kwds):
pass
@staticmethod
2022-06-23 22:48:25 +08:00
def _generate_next_value_(name, start, count, last_values):
"""
Generate the next value when not given.
name: the name of the member
start: the initial start value or None
count: the number of existing members
2022-06-23 22:48:25 +08:00
last_values: the list of values assigned
"""
2022-06-23 22:48:25 +08:00
if not last_values:
return start
try:
last_value = sorted(last_values).pop()
except TypeError:
raise TypeError('unable to sort non-numeric values') from None
try:
return last_value + 1
except TypeError:
raise TypeError('unable to increment %r' % (last_value, )) from None
@classmethod
def _missing_(cls, value):
return None
def __repr__(self):
v_repr = self.__class__._value_repr_ or repr
return "<%s.%s: %s>" % (self.__class__.__name__, self._name_, v_repr(self._value_))
def __str__(self):
return "%s.%s" % (self.__class__.__name__, self._name_, )
def __dir__(self):
"""
Returns all members and all public methods
"""
if self.__class__._member_type_ is object:
interesting = set(['__class__', '__doc__', '__eq__', '__hash__', '__module__', 'name', 'value'])
else:
interesting = set(object.__dir__(self))
for name in getattr(self, '__dict__', []):
if name[0] != '_':
interesting.add(name)
for cls in self.__class__.mro():
for name, obj in cls.__dict__.items():
if name[0] == '_':
continue
if isinstance(obj, property):
# that's an enum.property
if obj.fget is not None or name not in self._member_map_:
interesting.add(name)
else:
# in case it was added by `dir(self)`
interesting.discard(name)
else:
interesting.add(name)
names = sorted(
set(['__class__', '__doc__', '__eq__', '__hash__', '__module__'])
| interesting
)
return names
def __format__(self, format_spec):
return str.__format__(str(self), format_spec)
def __hash__(self):
return hash(self._name_)
def __reduce_ex__(self, proto):
return self.__class__, (self._value_, )
def __deepcopy__(self,memo):
return self
def __copy__(self):
return self
# enum.property is used to provide access to the `name` and
# `value` attributes of enum members while keeping some measure of
# protection from modification, while still allowing for an enumeration
# to have members named `name` and `value`. This works because each
# instance of enum.property saves its companion member, which it returns
# on class lookup; on instance lookup it either executes a provided function
# or raises an AttributeError.
@property
def name(self):
"""The name of the Enum member."""
return self._name_
@property
def value(self):
"""The value of the Enum member."""
return self._value_
class ReprEnum(Enum):
"""
Only changes the repr(), leaving str() and format() to the mixed-in type.
"""
class IntEnum(int, ReprEnum):
"""
Enum where members are also (and must be) ints
"""
class StrEnum(str, ReprEnum):
"""
Enum where members are also (and must be) strings
"""
def __new__(cls, *values):
"values must already be of type `str`"
if len(values) > 3:
raise TypeError('too many arguments for str(): %r' % (values, ))
if len(values) == 1:
# it must be a string
if not isinstance(values[0], str):
raise TypeError('%r is not a string' % (values[0], ))
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if len(values) >= 2:
# check that encoding argument is a string
if not isinstance(values[1], str):
raise TypeError('encoding must be a string, not %r' % (values[1], ))
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if len(values) == 3:
# check that errors argument is a string
if not isinstance(values[2], str):
raise TypeError('errors must be a string, not %r' % (values[2]))
value = str(*values)
member = str.__new__(cls, value)
member._value_ = value
return member
@staticmethod
def _generate_next_value_(name, start, count, last_values):
"""
Return the lower-cased version of the member name.
"""
return name.lower()
def pickle_by_global_name(self, proto):
# should not be used with Flag-type enums
return self.name
_reduce_ex_by_global_name = pickle_by_global_name
def pickle_by_enum_name(self, proto):
# should not be used with Flag-type enums
return getattr, (self.__class__, self._name_)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
class FlagBoundary(StrEnum):
"""
control how out of range values are handled
"strict" -> error is raised [default for Flag]
"conform" -> extra bits are discarded
"eject" -> lose flag status
"keep" -> keep flag status and all bits [default for IntFlag]
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
"""
STRICT = auto()
CONFORM = auto()
EJECT = auto()
KEEP = auto()
STRICT, CONFORM, EJECT, KEEP = FlagBoundary
class Flag(Enum, boundary=STRICT):
"""
Support for flags
"""
_numeric_repr_ = repr
@staticmethod
def _generate_next_value_(name, start, count, last_values):
"""
Generate the next value when not given.
name: the name of the member
start: the initial start value or None
count: the number of existing members
2022-06-23 22:48:25 +08:00
last_values: the last value assigned or None
"""
if not count:
return start if start is not None else 1
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
last_value = max(last_values)
try:
high_bit = _high_bit(last_value)
except Exception:
raise TypeError('invalid flag value %r' % last_value) from None
return 2 ** (high_bit+1)
@classmethod
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def _iter_member_by_value_(cls, value):
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
Extract all members from the value in definition (i.e. increasing value) order.
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
for val in _iter_bits_lsb(value & cls._flag_mask_):
yield cls._value2member_map_.get(val)
_iter_member_ = _iter_member_by_value_
@classmethod
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
def _iter_member_by_def_(cls, value):
"""
Extract all members from the value in definition order.
"""
yield from sorted(
cls._iter_member_by_value_(value),
key=lambda m: m._sort_order_,
)
@classmethod
def _missing_(cls, value):
"""
2021-04-15 21:49:54 +08:00
Create a composite member containing all canonical members present in `value`.
If non-member values are present, result depends on `_boundary_` setting.
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if not isinstance(value, int):
raise ValueError(
"%r is not a valid %s" % (value, cls.__qualname__)
)
# check boundaries
# - value must be in range (e.g. -16 <-> +15, i.e. ~15 <-> 15)
# - value must not include any skipped flags (e.g. if bit 2 is not
# defined, then 0d10 is invalid)
flag_mask = cls._flag_mask_
singles_mask = cls._singles_mask_
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
all_bits = cls._all_bits_
neg_value = None
if (
not ~all_bits <= value <= all_bits
or value & (all_bits ^ flag_mask)
):
if cls._boundary_ is STRICT:
max_bits = max(value.bit_length(), flag_mask.bit_length())
raise ValueError(
"%r invalid value %r\n given %s\n allowed %s" % (
cls, value, bin(value, max_bits), bin(flag_mask, max_bits),
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
))
elif cls._boundary_ is CONFORM:
value = value & flag_mask
elif cls._boundary_ is EJECT:
return value
elif cls._boundary_ is KEEP:
if value < 0:
value = (
max(all_bits+1, 2**(value.bit_length()))
+ value
)
else:
raise ValueError(
'%r unknown flag boundary %r' % (cls, cls._boundary_, )
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
)
if value < 0:
neg_value = value
value = all_bits + 1 + value
# get members and unknown
unknown = value & ~flag_mask
aliases = value & ~singles_mask
member_value = value & singles_mask
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if unknown and cls._boundary_ is not KEEP:
raise ValueError(
'%s(%r) --> unknown values %r [%s]'
% (cls.__name__, value, unknown, bin(unknown))
)
# normal Flag?
if cls._member_type_ is object:
# construct a singleton enum pseudo-member
pseudo_member = object.__new__(cls)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
else:
pseudo_member = cls._member_type_.__new__(cls, value)
if not hasattr(pseudo_member, '_value_'):
pseudo_member._value_ = value
if member_value or aliases:
members = []
combined_value = 0
for m in cls._iter_member_(member_value):
members.append(m)
combined_value |= m._value_
if aliases:
value = member_value | aliases
for n, pm in cls._member_map_.items():
if pm not in members and pm._value_ and pm._value_ & value == pm._value_:
members.append(pm)
combined_value |= pm._value_
unknown = value ^ combined_value
pseudo_member._name_ = '|'.join([m._name_ for m in members])
if not combined_value:
pseudo_member._name_ = None
elif unknown and cls._boundary_ is STRICT:
raise ValueError('%r: no members with value %r' % (cls, unknown))
elif unknown:
pseudo_member._name_ += '|%s' % cls._numeric_repr_(unknown)
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
else:
pseudo_member._name_ = None
# use setdefault in case another thread already created a composite
# with this value
# note: zero is a special case -- always add it
pseudo_member = cls._value2member_map_.setdefault(value, pseudo_member)
if neg_value is not None:
cls._value2member_map_[neg_value] = pseudo_member
return pseudo_member
def __contains__(self, other):
"""
Returns True if self has at least the same flags set as other.
"""
if not isinstance(other, self.__class__):
raise TypeError(
"unsupported operand type(s) for 'in': %r and %r" % (
type(other).__qualname__, self.__class__.__qualname__))
return other._value_ & self._value_ == other._value_
def __iter__(self):
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
Returns flags in definition order.
"""
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
yield from self._iter_member_(self._value_)
def __len__(self):
return self._value_.bit_count()
def __repr__(self):
cls_name = self.__class__.__name__
v_repr = self.__class__._value_repr_ or repr
if self._name_ is None:
return "<%s: %s>" % (cls_name, v_repr(self._value_))
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
else:
return "<%s.%s: %s>" % (cls_name, self._name_, v_repr(self._value_))
def __str__(self):
cls_name = self.__class__.__name__
if self._name_ is None:
return '%s(%r)' % (cls_name, self._value_)
else:
return "%s.%s" % (cls_name, self._name_)
def __bool__(self):
return bool(self._value_)
def __or__(self, other):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if isinstance(other, self.__class__):
other = other._value_
elif self._member_type_ is not object and isinstance(other, self._member_type_):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
other = other
else:
return NotImplemented
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
value = self._value_
return self.__class__(value | other)
def __and__(self, other):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if isinstance(other, self.__class__):
other = other._value_
elif self._member_type_ is not object and isinstance(other, self._member_type_):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
other = other
else:
return NotImplemented
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
value = self._value_
return self.__class__(value & other)
def __xor__(self, other):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
if isinstance(other, self.__class__):
other = other._value_
elif self._member_type_ is not object and isinstance(other, self._member_type_):
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
other = other
else:
return NotImplemented
bpo-38250: [Enum] single-bit flags are canonical (GH-24215) Flag members are now divided by one-bit verses multi-bit, with multi-bit being treated as aliases. Iterating over a flag only returns the contained single-bit flags. Iterating, repr(), and str() show members in definition order. When constructing combined-member flags, any extra integer values are either discarded (CONFORM), turned into ints (EJECT) or treated as errors (STRICT). Flag classes can specify which of those three behaviors is desired: >>> class Test(Flag, boundary=CONFORM): ... ONE = 1 ... TWO = 2 ... >>> Test(5) <Test.ONE: 1> Besides the three above behaviors, there is also KEEP, which should not be used unless necessary -- for example, _convert_ specifies KEEP as there are flag sets in the stdlib that are incomplete and/or inconsistent (e.g. ssl.Options). KEEP will, as the name suggests, keep all bits; however, iterating over a flag with extra bits will only return the canonical flags contained, not the extra bits. Iteration is now in member definition order. If member definition order matches increasing value order, then a more efficient method of flag decomposition is used; otherwise, sort() is called on the results of that method to get definition order. ``re`` module: repr() has been modified to support as closely as possible its previous output; the big difference is that inverted flags cannot be output as before because the inversion operation now always returns the comparable positive result; i.e. re.A|re.I|re.M|re.S is ~(re.L|re.U|re.S|re.T|re.DEBUG) in both of the above terms, the ``value`` is 282. re's tests have been updated to reflect the modifications to repr().
2021-01-26 06:26:19 +08:00
value = self._value_
return self.__class__(value ^ other)
def __invert__(self):
if self._inverted_ is None:
if self._boundary_ in (EJECT, KEEP):
self._inverted_ = self.__class__(~self._value_)
else:
self._inverted_ = self.__class__(self._singles_mask_ & ~self._value_)
return self._inverted_
__rand__ = __and__
__ror__ = __or__
__rxor__ = __xor__
class IntFlag(int, ReprEnum, Flag, boundary=KEEP):
"""
Support for integer-based Flags
"""
def _high_bit(value):
"""
returns index of highest bit, or -1 if value is zero or negative
"""
return value.bit_length() - 1
def unique(enumeration):
"""
Class decorator for enumerations ensuring unique member values.
"""
duplicates = []
for name, member in enumeration.__members__.items():
if name != member.name:
duplicates.append((name, member.name))
if duplicates:
alias_details = ', '.join(
["%s -> %s" % (alias, name) for (alias, name) in duplicates])
raise ValueError('duplicate values found in %r: %s' %
(enumeration, alias_details))
return enumeration
def _dataclass_repr(self):
dcf = self.__dataclass_fields__
return ', '.join(
'%s=%r' % (k, getattr(self, k))
for k in dcf.keys()
if dcf[k].repr
)
def global_enum_repr(self):
"""
use module.enum_name instead of class.enum_name
the module is the last module in case of a multi-module name
"""
module = self.__class__.__module__.split('.')[-1]
return '%s.%s' % (module, self._name_)
def global_flag_repr(self):
"""
use module.flag_name instead of class.flag_name
the module is the last module in case of a multi-module name
"""
module = self.__class__.__module__.split('.')[-1]
cls_name = self.__class__.__name__
if self._name_ is None:
return "%s.%s(%r)" % (module, cls_name, self._value_)
if _is_single_bit(self):
return '%s.%s' % (module, self._name_)
if self._boundary_ is not FlagBoundary.KEEP:
return '|'.join(['%s.%s' % (module, name) for name in self.name.split('|')])
else:
name = []
for n in self._name_.split('|'):
if n[0].isdigit():
name.append(n)
else:
name.append('%s.%s' % (module, n))
return '|'.join(name)
def global_str(self):
"""
use enum_name instead of class.enum_name
"""
if self._name_ is None:
cls_name = self.__class__.__name__
return "%s(%r)" % (cls_name, self._value_)
else:
return self._name_
def global_enum(cls, update_str=False):
"""
decorator that makes the repr() of an enum member reference its module
instead of its class; also exports all members to the enum's module's
global namespace
"""
if issubclass(cls, Flag):
cls.__repr__ = global_flag_repr
else:
cls.__repr__ = global_enum_repr
if not issubclass(cls, ReprEnum) or update_str:
cls.__str__ = global_str
sys.modules[cls.__module__].__dict__.update(cls.__members__)
return cls
def _simple_enum(etype=Enum, *, boundary=None, use_args=None):
"""
Class decorator that converts a normal class into an :class:`Enum`. No
safety checks are done, and some advanced behavior (such as
:func:`__init_subclass__`) is not available. Enum creation can be faster
using :func:`_simple_enum`.
>>> from enum import Enum, _simple_enum
>>> @_simple_enum(Enum)
... class Color:
... RED = auto()
... GREEN = auto()
... BLUE = auto()
>>> Color
<enum 'Color'>
"""
def convert_class(cls):
nonlocal use_args
cls_name = cls.__name__
if use_args is None:
use_args = etype._use_args_
__new__ = cls.__dict__.get('__new__')
if __new__ is not None:
new_member = __new__.__func__
else:
new_member = etype._member_type_.__new__
attrs = {}
body = {}
if __new__ is not None:
body['__new_member__'] = new_member
body['_new_member_'] = new_member
body['_use_args_'] = use_args
body['_generate_next_value_'] = gnv = etype._generate_next_value_
body['_member_names_'] = member_names = []
body['_member_map_'] = member_map = {}
body['_value2member_map_'] = value2member_map = {}
body['_unhashable_values_'] = []
body['_member_type_'] = member_type = etype._member_type_
body['_value_repr_'] = etype._value_repr_
if issubclass(etype, Flag):
body['_boundary_'] = boundary or etype._boundary_
body['_flag_mask_'] = None
body['_all_bits_'] = None
body['_singles_mask_'] = None
body['_inverted_'] = None
body['__or__'] = Flag.__or__
body['__xor__'] = Flag.__xor__
body['__and__'] = Flag.__and__
body['__ror__'] = Flag.__ror__
body['__rxor__'] = Flag.__rxor__
body['__rand__'] = Flag.__rand__
body['__invert__'] = Flag.__invert__
for name, obj in cls.__dict__.items():
if name in ('__dict__', '__weakref__'):
continue
if _is_dunder(name) or _is_private(cls_name, name) or _is_sunder(name) or _is_descriptor(obj):
body[name] = obj
else:
attrs[name] = obj
if cls.__dict__.get('__doc__') is None:
body['__doc__'] = 'An enumeration.'
#
# double check that repr and friends are not the mixin's or various
# things break (such as pickle)
# however, if the method is defined in the Enum itself, don't replace
# it
enum_class = type(cls_name, (etype, ), body, boundary=boundary, _simple=True)
for name in ('__repr__', '__str__', '__format__', '__reduce_ex__'):
if name not in body:
# check for mixin overrides before replacing
enum_method = getattr(etype, name)
found_method = getattr(enum_class, name)
object_method = getattr(object, name)
data_type_method = getattr(member_type, name)
if found_method in (data_type_method, object_method):
setattr(enum_class, name, enum_method)
gnv_last_values = []
if issubclass(enum_class, Flag):
# Flag / IntFlag
single_bits = multi_bits = 0
for name, value in attrs.items():
if isinstance(value, auto) and auto.value is _auto_null:
value = gnv(name, 1, len(member_names), gnv_last_values)
if value in value2member_map:
# an alias to an existing member
member = value2member_map[value]
redirect = property()
redirect.member = member
redirect.__set_name__(enum_class, name)
setattr(enum_class, name, redirect)
member_map[name] = member
else:
# create the member
if use_args:
if not isinstance(value, tuple):
value = (value, )
member = new_member(enum_class, *value)
value = value[0]
else:
member = new_member(enum_class)
if __new__ is None:
member._value_ = value
member._name_ = name
member.__objclass__ = enum_class
member.__init__(value)
redirect = property()
redirect.member = member
redirect.__set_name__(enum_class, name)
setattr(enum_class, name, redirect)
member_map[name] = member
member._sort_order_ = len(member_names)
value2member_map[value] = member
if _is_single_bit(value):
# not a multi-bit alias, record in _member_names_ and _flag_mask_
member_names.append(name)
single_bits |= value
else:
multi_bits |= value
gnv_last_values.append(value)
enum_class._flag_mask_ = single_bits | multi_bits
enum_class._singles_mask_ = single_bits
enum_class._all_bits_ = 2 ** ((single_bits|multi_bits).bit_length()) - 1
# set correct __iter__
member_list = [m._value_ for m in enum_class]
if member_list != sorted(member_list):
enum_class._iter_member_ = enum_class._iter_member_by_def_
else:
# Enum / IntEnum / StrEnum
for name, value in attrs.items():
if isinstance(value, auto):
if value.value is _auto_null:
value.value = gnv(name, 1, len(member_names), gnv_last_values)
value = value.value
if value in value2member_map:
# an alias to an existing member
member = value2member_map[value]
redirect = property()
redirect.member = member
redirect.__set_name__(enum_class, name)
setattr(enum_class, name, redirect)
member_map[name] = member
else:
# create the member
if use_args:
if not isinstance(value, tuple):
value = (value, )
member = new_member(enum_class, *value)
value = value[0]
else:
member = new_member(enum_class)
if __new__ is None:
member._value_ = value
member._name_ = name
member.__objclass__ = enum_class
member.__init__(value)
member._sort_order_ = len(member_names)
redirect = property()
redirect.member = member
redirect.__set_name__(enum_class, name)
setattr(enum_class, name, redirect)
member_map[name] = member
value2member_map[value] = member
member_names.append(name)
gnv_last_values.append(value)
if '__new__' in body:
enum_class.__new_member__ = enum_class.__new__
enum_class.__new__ = Enum.__new__
return enum_class
return convert_class
@_simple_enum(StrEnum)
class EnumCheck:
"""
various conditions to check an enumeration for
"""
CONTINUOUS = "no skipped integer values"
NAMED_FLAGS = "multi-flag aliases may not contain unnamed flags"
UNIQUE = "one name per value"
CONTINUOUS, NAMED_FLAGS, UNIQUE = EnumCheck
class verify:
"""
Check an enumeration for various constraints. (see EnumCheck)
"""
def __init__(self, *checks):
self.checks = checks
def __call__(self, enumeration):
checks = self.checks
cls_name = enumeration.__name__
if Flag is not None and issubclass(enumeration, Flag):
enum_type = 'flag'
elif issubclass(enumeration, Enum):
enum_type = 'enum'
else:
raise TypeError("the 'verify' decorator only works with Enum and Flag")
for check in checks:
if check is UNIQUE:
# check for duplicate names
duplicates = []
for name, member in enumeration.__members__.items():
if name != member.name:
duplicates.append((name, member.name))
if duplicates:
alias_details = ', '.join(
["%s -> %s" % (alias, name) for (alias, name) in duplicates])
raise ValueError('aliases found in %r: %s' %
(enumeration, alias_details))
elif check is CONTINUOUS:
values = set(e.value for e in enumeration)
if len(values) < 2:
continue
low, high = min(values), max(values)
missing = []
if enum_type == 'flag':
# check for powers of two
for i in range(_high_bit(low)+1, _high_bit(high)):
if 2**i not in values:
missing.append(2**i)
elif enum_type == 'enum':
# check for powers of one
for i in range(low+1, high):
if i not in values:
missing.append(i)
else:
raise Exception('verify: unknown type %r' % enum_type)
if missing:
raise ValueError(('invalid %s %r: missing values %s' % (
enum_type, cls_name, ', '.join((str(m) for m in missing)))
)[:256])
# limit max length to protect against DOS attacks
elif check is NAMED_FLAGS:
# examine each alias and check for unnamed flags
member_names = enumeration._member_names_
member_values = [m.value for m in enumeration]
missing_names = []
missing_value = 0
for name, alias in enumeration._member_map_.items():
if name in member_names:
# not an alias
continue
if alias.value < 0:
# negative numbers are not checked
continue
values = list(_iter_bits_lsb(alias.value))
missed = [v for v in values if v not in member_values]
if missed:
missing_names.append(name)
for val in missed:
missing_value |= val
if missing_names:
if len(missing_names) == 1:
alias = 'alias %s is missing' % missing_names[0]
else:
alias = 'aliases %s and %s are missing' % (
', '.join(missing_names[:-1]), missing_names[-1]
)
if _is_single_bit(missing_value):
value = 'value 0x%x' % missing_value
else:
value = 'combined values of 0x%x' % missing_value
raise ValueError(
'invalid Flag %r: %s %s [use enum.show_flag_values(value) for details]'
% (cls_name, alias, value)
)
return enumeration
def _test_simple_enum(checked_enum, simple_enum):
"""
A function that can be used to test an enum created with :func:`_simple_enum`
against the version created by subclassing :class:`Enum`::
>>> from enum import Enum, _simple_enum, _test_simple_enum
>>> @_simple_enum(Enum)
... class Color:
... RED = auto()
... GREEN = auto()
... BLUE = auto()
>>> class CheckedColor(Enum):
... RED = auto()
... GREEN = auto()
... BLUE = auto()
>>> _test_simple_enum(CheckedColor, Color)
If differences are found, a :exc:`TypeError` is raised.
"""
failed = []
if checked_enum.__dict__ != simple_enum.__dict__:
checked_dict = checked_enum.__dict__
checked_keys = list(checked_dict.keys())
simple_dict = simple_enum.__dict__
simple_keys = list(simple_dict.keys())
member_names = set(
list(checked_enum._member_map_.keys())
+ list(simple_enum._member_map_.keys())
)
for key in set(checked_keys + simple_keys):
if key in ('__module__', '_member_map_', '_value2member_map_', '__doc__'):
# keys known to be different, or very long
continue
elif key in member_names:
# members are checked below
continue
elif key not in simple_keys:
failed.append("missing key: %r" % (key, ))
elif key not in checked_keys:
failed.append("extra key: %r" % (key, ))
else:
checked_value = checked_dict[key]
simple_value = simple_dict[key]
if callable(checked_value) or isinstance(checked_value, bltns.property):
continue
if key == '__doc__':
# remove all spaces/tabs
compressed_checked_value = checked_value.replace(' ','').replace('\t','')
compressed_simple_value = simple_value.replace(' ','').replace('\t','')
if compressed_checked_value != compressed_simple_value:
failed.append("%r:\n %s\n %s" % (
key,
"checked -> %r" % (checked_value, ),
"simple -> %r" % (simple_value, ),
))
elif checked_value != simple_value:
failed.append("%r:\n %s\n %s" % (
key,
"checked -> %r" % (checked_value, ),
"simple -> %r" % (simple_value, ),
))
failed.sort()
for name in member_names:
failed_member = []
if name not in simple_keys:
failed.append('missing member from simple enum: %r' % name)
elif name not in checked_keys:
failed.append('extra member in simple enum: %r' % name)
else:
checked_member_dict = checked_enum[name].__dict__
checked_member_keys = list(checked_member_dict.keys())
simple_member_dict = simple_enum[name].__dict__
simple_member_keys = list(simple_member_dict.keys())
for key in set(checked_member_keys + simple_member_keys):
if key in ('__module__', '__objclass__', '_inverted_'):
# keys known to be different or absent
continue
elif key not in simple_member_keys:
failed_member.append("missing key %r not in the simple enum member %r" % (key, name))
elif key not in checked_member_keys:
failed_member.append("extra key %r in simple enum member %r" % (key, name))
else:
checked_value = checked_member_dict[key]
simple_value = simple_member_dict[key]
if checked_value != simple_value:
failed_member.append("%r:\n %s\n %s" % (
key,
"checked member -> %r" % (checked_value, ),
"simple member -> %r" % (simple_value, ),
))
if failed_member:
failed.append('%r member mismatch:\n %s' % (
name, '\n '.join(failed_member),
))
for method in (
'__str__', '__repr__', '__reduce_ex__', '__format__',
'__getnewargs_ex__', '__getnewargs__', '__reduce_ex__', '__reduce__'
):
if method in simple_keys and method in checked_keys:
# cannot compare functions, and it exists in both, so we're good
continue
elif method not in simple_keys and method not in checked_keys:
# method is inherited -- check it out
checked_method = getattr(checked_enum, method, None)
simple_method = getattr(simple_enum, method, None)
if hasattr(checked_method, '__func__'):
checked_method = checked_method.__func__
simple_method = simple_method.__func__
if checked_method != simple_method:
failed.append("%r: %-30s %s" % (
method,
"checked -> %r" % (checked_method, ),
"simple -> %r" % (simple_method, ),
))
else:
# if the method existed in only one of the enums, it will have been caught
# in the first checks above
pass
if failed:
raise TypeError('enum mismatch:\n %s' % '\n '.join(failed))
def _old_convert_(etype, name, module, filter, source=None, *, boundary=None):
"""
Create a new Enum subclass that replaces a collection of global constants
"""
# convert all constants from source (or module) that pass filter() to
# a new Enum called name, and export the enum and its members back to
# module;
# also, replace the __reduce_ex__ method so unpickling works in
# previous Python versions
module_globals = sys.modules[module].__dict__
if source:
source = source.__dict__
else:
source = module_globals
# _value2member_map_ is populated in the same order every time
# for a consistent reverse mapping of number to name when there
# are multiple names for the same number.
members = [
(name, value)
for name, value in source.items()
if filter(name)]
try:
# sort by value
members.sort(key=lambda t: (t[1], t[0]))
except TypeError:
# unless some values aren't comparable, in which case sort by name
members.sort(key=lambda t: t[0])
cls = etype(name, members, module=module, boundary=boundary or KEEP)
return cls
_stdlib_enums = IntEnum, StrEnum, IntFlag