SQLAlchemy中与过滤器的自引用关系

Question

我正在尝试使用自定义JOIN逻辑创建自引用SQLAlchemy关系。

我们的数据库有一个事件表，其中包含“ eventID”，“ eventDate”和“ userID”列。 我正在实现的关系是“ previousEvents”。 我想使用'userID'将Events表加入自身，其中eventDate> eventDate用于具有相同userID的每个其他事件。

构造的SQL查询应如下所示：

SELECT original.eventID, previous.eventID, ... etc
FROM   Events as original
JOIN   Events as previous 
       ON  previous.userID = original.userID 
       AND previous.eventDate < original.eventDate

我的表定义如下所示：

from sqlalchemy import Column, ForeignKey, String, DateTime
from sqlalchemy.ext.declarative import declarative_base

class Event(Base):
    __tablename__ = 'event'

    # Attributes
    eventID = Column(String(length=256), primary_key=True)
    eventDate = Column(DateTime)
    userID = Column(String(length=256), ForeignKey('users.userID'))

class Users(Base):
    __tablename__ = 'users'

    # Attributes
    userID = Column(String(length=256), primary_key=True)
    userName = Column(String)

如何在事件中定义关系以获得理想的效果？

Answer 1

您需要一个替代的联接条件和自定义外来条件 ，因为没有实际的外键关系可以联接：

class Event(Base):                                                            
    __tablename__ = 'event'            

    # Attributes
    eventID = Column(String(length=256), primary_key=True)
    eventDate = Column(DateTime)
    userID = Column(String(length=256), ForeignKey('users.userID'))

    previousEvents = relationship(
        "Event", primaryjoin=and_(userID == foreign(userID),
                                  eventDate > foreign(eventDate)),
        viewonly=True)      

使用foreign()注释创建没有实际外键的连接条件。 还可以使用relationship()的foreign_keys参数，但是您必须非常小心地安排比较，以使列以正确的顺序排列，因为联接的两边都使用相同的列。

当您希望创建一个连接先前事件的查询时，您需要一个别名 ：

In [13]: prev_event = aliased(Event)

In [14]: print(session.query(Event).join(prev_event, Event.previousEvents))
SELECT event."eventID" AS "event_eventID", event."eventDate" AS "event_eventDate", event."userID" AS "event_userID" 
FROM event JOIN event AS event_1 ON event_1."userID" = event."userID" AND event_1."eventDate" < event."eventDate"

或者，如果您希望使用连接加载：

In [17]: print(session.query(Event).options(joinedload(Event.previousEvents)))
SELECT event."eventID" AS "event_eventID", event."eventDate" AS "event_eventDate", event."userID" AS "event_userID", event_1."eventID" AS "event_1_eventID", event_1."eventDate" AS "event_1_eventDate", event_1."userID" AS "event_1_userID" 
FROM event LEFT OUTER JOIN event AS event_1 ON event_1."userID" = event."userID" AND event_1."eventDate" < event."eventDate"