在SQLAlchemy中,我已经阅读了如何在声明声明时组合来自不同mixin的__table_args__
。
Combining Table/Mapper Arguments from Multiple Mixins
我遇到的问题是,该示例显示了如何在链的末尾(MRO中的最终类)完成此操作,但是如果我有这些Mixins并希望它发生在MyClientMixin
或Base
类,以避免为其他类型的mixin重复此代码?
class LaneCarrierCommitmentSummaryMixin(object):
""" Lane Carrier Commitment Summary.
A base class for LCCS to mixin with a client specific class.
"""
__tablename__ = 'lane_carrier_commitment_summary'
__table_args__ = ((UniqueConstraint(['hashmap_key', 'bow'],
name='uq_lane_carrier_commitment_summary_hashmap_key_bow')),)
class MyClientMixin(object):
""" MyClient Mixin class for providing the ETL schema. """
__table_args__ = {'schema': 'myclient_etl'}
class MyClientLaneCarrierCommitmentSummary(LaneCarrierCommitmentSummaryMixin, DateTrackedMixin, MyClientMixin, Base):
pass
我对这个概念感到苦苦挣扎。
答案 0 :(得分:8)
此基类将搜索要添加的__table_args__
的所有mixins,然后检查要添加的__local_table_args__
的当前类。这样,__local_table_args__
不会与声明的attr发生冲突。基类(cls.mro()
)以相反的顺序检查,以便链中较低的mixins被更高的mixin覆盖。
def _process_args(cls, attr, out_args, out_kwargs):
try:
args = getattr(cls, attr)
except AttributeError:
return
if isinstance(args, Mapping): # it's a dictionary
out_kwargs.update(args)
else: # it's a list
if isinstance(args[-1], Mapping): # it has a dictionary at the end
out_kwargs.update(args.pop())
out_args.extend(args)
class Base():
@declared_attr
def __mapper_args__(cls):
args = []
kwargs = {}
for mixin in reversed(cls.mro()):
_process_args(mixin, '__mapper_args__', args, kwargs)
_process_args(mixin, '__local_mapper_args__', args, kwargs)
return kwargs # mapper only takes dict
@declared_attr
def __table_args__(cls):
args = []
kwargs = {}
for mixin in reversed(cls.mro()):
_process_args(mixin, '__table_args__', args, kwargs)
_process_args(cls, '__local_table_args__', args, kwargs)
args.append(kwargs) # [item, item, ..., kwargs]
return tuple(args)
所有mixin都应该正常定义__table_args__
,但是继承自Base
的“真实”类应该定义__local_table_args__
。