首先让我来定义我的目标:
启用没有默认值的抽象类成员(不是属性,方法或实例成员)的定义(不是None或其他一些魔术值,但如果没有实现则抛出错误)。
在(1)的服务中,创建一个可重用的抽象机制,使得创建具有抽象成员的类变得微不足道,并且这样做的代码最简洁。
能够将抽象类成员附加到父类(作为子类,元类或任何其他方法),在定义具体子类之前,不需要定义抽象成员。
到目前为止我得到了什么:
metaclass_lib.py:
class AbstractRequiredClassMembersMetaclass(type):
def __new__(cls, name, bases, attrs):
missed_members = []
for class_member in cls.REQUIRED_CLASS_MEMBERS:
if class_member not in attrs:
missed_members.append(class_member)
if missed_members:
raise NotImplementedError('Class missing required members %s.' % missed_members)
return super(AbstractRequiredClassMembersMetaclass, cls).__new__(cls, name, bases, attrs)
class _MakeRequiredClassMemebersMetaclass(object):
existing_classes = {}
def __call__(self, name, required_members):
if name in _MakeRequiredClassMemebersMetaclass.exisiting_classes:
if required_members != _MakeRequiredClassMemebersMetaclass.exisiting_classes[name].REQUIRED_CLASS_MEMBERS:
raise RuntimeError('Class of name %s already defined with different required_members.' % name)
else:
NewClass = type(name, (AbstractRequiredClassMembersMetaclass,), {'REQUIRED_CLASS_MEMBERS' : required_members})
_MakeRequiredClassMemebersMetaclass.exisiting_classes[name] = NewClass
return _MakeRequiredClassMemebersMetaclass.exisiting_classes[name]
make_required_class_members_metaclass = _MakeRequiredClassMemebersMetaclass()
goods.py(用于说明的实施):
from metaclass_lib import make_required_class_members_metaclass
class AbstractGood(object):
__metaclass__ = make_required_class_members_metaclass('AbstractGoodMeta', ('unit_measure',))
def __init__(self):
self.unit_price = None
def total_cost(self, number_of_units):
return self.unit_price * self.number_of_units
class DeliGood(AbstractGood):
unit_measure = 'lbs'
def __init__(self):
self.sandwich_counter_product = False
class RefridgeratedGood(AbstractGood):
unit_measure = 'package'
def __init__(self):
self.freezer_safe = False
self.optimal_temp = '35'
这不起作用,因为在创建AbstractGood
type
对象期间,元类会窒息。问题是我想要所有具体商品来定义类成员,但我不想在任何抽象基础中定义类成员。我能想到的只是让元类只进行in attrs
检查关键字是否在name
(ex if 'Abstract' not in name
)中,但这看起来很脆弱而且很脆弱。
有更好的方法吗?
答案 0 :(得分:1)
也是一个黑客,但比检查名称更可靠。
您知道抽象基类没有基类
(或者在python 2中只有object
类型)。所以你可以检查一下:)
class AbstractBase(object):
class __metaclass__(type):
def __new__(mcs, name, bases, dict_):
if bases[0] is not object: # hack to not apply these on AbstractBase
assert 'unit_measure' in dict_
return type.__new__(mcs, name, bases, dict_)
class DeliGood(AbstractBase):
unit_measure = 'lbs'