如何在 sqlalchemy 经典映射上使用 cattrs 构造数据类？

Question

我正在尝试在我使用 sqlalchemy 的应用程序中使用 cattr 来构造一些 classes。出于某种原因，在我向数据库提交 class 后，无法构造回任何嵌套数据 class。我收到错误：

in emit_backref_from_collection_append_event
    child_state, child_dict = instance_state(child), instance_dict(child)
AttributeError: 'dict' object has no attribute '_sa_instance_state'

这是一个代码示例。

from dataclasses import dataclass
from sqlalchemy.orm import backref, relationship, registry, sessionmaker
from sqlalchemy import Table, Column, ForeignKey, create_engine
from sqlalchemy.sql.sqltypes import Integer, String
import cattr

from src.vlep.domain.model import Patient

mapper_registry = registry()

parents = Table(
    'parents', mapper_registry.metadata,
    Column('id', Integer, primary_key=True, autoincrement=True),
    Column('name', String(255)),
)

children = Table(
    'children', mapper_registry.metadata,
    Column('id', Integer, primary_key=True, autoincrement=True),
    Column('name', String(255)),
    Column('parent_id', ForeignKey('parents.id', ondelete='SET NULL'))
)


def start_mappers():
    children_mapper = mapper_registry.map_imperatively(
        Child,
        children,
    )
    parent_mapper = mapper_registry.map_imperatively(
        Parent,
        parents,
        properties={
            'children': relationship(
                children_mapper,
                backref=backref('parents'))
        }
    )


@dataclass
class Child:

    name: str


@dataclass
class Parent:

    name: str
    children: list


if __name__ == "__main__":

    engine = create_engine("sqlite:///:memory:")
    mapper_registry.metadata.create_all(engine)
    start_mappers()
    session = sessionmaker(bind=engine)()

    c1 = Child('Uguinho')
    c2 = Child('Luizinho')
    p = Parent('Fulano', [c1, c2])

    session.add(p)
    session.commit()

    try:
        # This works
        d = cattr.unstructure(p, Child)
        p2 = cattr.structure(d, Child)
        d2 = cattr.unstructure(p2, Child)
        print(d2)
    finally:
        # This does not
        d = cattr.unstructure(p, Parent)
        p2 = cattr.structure(d, Parent)
        d2 = cattr.unstructure(p2, Parent)
        print(d2)

根据错误消息，我猜测 sqlalchemy 正在尝试检查 parent_mapper 上的属性 'children' 是否是 Child 的列表并且它正在获取dict 而不是（它处于非结构化状态）。我什至不知道 sqlalchemy 是如何进行这项检查的。再次猜测，我想它总是会做一些一致性检查。无论如何，我现在知道如何解决这个问题了。

Answer 1

所以我仔细研究了 SQLAlchemy 文档，发现并改编了 an example 让您的代码工作：

from dataclasses import dataclass
from typing import List

import cattr
from sqlalchemy.orm import (
    relationship,
    sessionmaker,
    registry,
)
from sqlalchemy import Table, Column, ForeignKey, create_engine, MetaData
from sqlalchemy.sql.sqltypes import Integer, String

# commented this out for the example
# from src.vlep.domain.model import Patient

mapper_registry = registry()


@dataclass
class Child:

    name: str


@dataclass
class Parent:

    name: str
    children: List[Child]


metadata_obj = MetaData()

parent = Table(
    "parent",
    metadata_obj,
    Column("id", Integer, primary_key=True, autoincrement=True),
    Column("name", String(255)),
)

child = Table(
    "child",
    metadata_obj,
    Column("id", Integer, primary_key=True, autoincrement=True),
    Column("name", String(255)),
    Column("parent_id", ForeignKey("parent.id", ondelete="SET NULL")),
)


def start_mappers():
    mapper_registry.map_imperatively(
        Parent,
        parent,
        properties={"children": relationship(Child, backref="parent")},
    )
    mapper_registry.map_imperatively(Child, child)


if __name__ == "__main__":

    engine = create_engine("sqlite:///:memory:")
    metadata_obj.create_all(engine)
    start_mappers()
    session = sessionmaker(bind=engine)()

    c1 = Child("Uguinho")
    c2 = Child("Luizinho")
    p = Parent("Fulano", [c1, c2])

    session.add(p)
    session.commit()

    try:
        # This works
        d = cattr.unstructure(p, Child)
        p2 = cattr.structure(d, Child)
        d2 = cattr.unstructure(p2, Child)
        print(d2)
    finally:
        # This now works
        d = cattr.unstructure(p, Parent)
        p2 = cattr.structure(d, Parent)
        d2 = cattr.unstructure(p2, Parent)
        print(d2)

输出：

{'name': 'Fulano'}
{'name': 'Fulano', 'children': [{'name': 'Uguinho'}, {'name': 'Luizinho'}]}

我认为可能主要问题之一是您将“backref”声明为 backref=backref('parents')，而那里的函数包装器是不必要的？不太确定...

无论如何，数据类现在用于正确映射（据我所知）。我希望这个例子对你有用！

如何在 sqlalchemy 经典映射上使用 cattrs 构造数据类？

How to structure dataclass with cattrs on sqlalchemy classical mapping?

python

sqlalchemy