我建议用棉花糖。它允许您创建序列化器来表示支持关系和嵌套对象的模型实例。

以下是他们文档中的一个删节的例子。以ORM模型为例，作者:

class Author(db.Model):
    id = db.Column(db.Integer, primary_key=True)
    first = db.Column(db.String(80))
    last = db.Column(db.String(80))

该类的棉花糖模式是这样构造的:

class AuthorSchema(Schema):
    id = fields.Int(dump_only=True)
    first = fields.Str()
    last = fields.Str()
    formatted_name = fields.Method("format_name", dump_only=True)

    def format_name(self, author):
        return "{}, {}".format(author.last, author.first)

.．.并像这样使用:

author_schema = AuthorSchema()
author_schema.dump(Author.query.first())

.．.会产生这样的输出:

{
        "first": "Tim",
        "formatted_name": "Peters, Tim",
        "id": 1,
        "last": "Peters"
}

看看他们完整的Flask-SQLAlchemy示例。

一个名为marshmlow - SQLAlchemy的库专门集成了SQLAlchemy和marshmallow。在这个库中，上面描述的Author模型的模式如下所示:

class AuthorSchema(ModelSchema):
    class Meta:
        model = Author

该集成允许从SQLAlchemy Column类型推断字段类型。

marshmallow-sqlalchemy这里。

下面是一个解决方案，它允许您选择希望在输出中包含的关系。 注意:这是一个完整的重写，将dict/str作为一个参数，而不是一个列表。修复了一些东西..

def deep_dict(self, relations={}):
    """Output a dict of an SA object recursing as deep as you want.

    Takes one argument, relations which is a dictionary of relations we'd
    like to pull out. The relations dict items can be a single relation
    name or deeper relation names connected by sub dicts

    Example:
        Say we have a Person object with a family relationship
            person.deep_dict(relations={'family':None})
        Say the family object has homes as a relation then we can do
            person.deep_dict(relations={'family':{'homes':None}})
            OR
            person.deep_dict(relations={'family':'homes'})
        Say homes has a relation like rooms you can do
            person.deep_dict(relations={'family':{'homes':'rooms'}})
            and so on...
    """
    mydict =  dict((c, str(a)) for c, a in
                    self.__dict__.items() if c != '_sa_instance_state')
    if not relations:
        # just return ourselves
        return mydict

    # otherwise we need to go deeper
    if not isinstance(relations, dict) and not isinstance(relations, str):
        raise Exception("relations should be a dict, it is of type {}".format(type(relations)))

    # got here so check and handle if we were passed a dict
    if isinstance(relations, dict):
        # we were passed deeper info
        for left, right in relations.items():
            myrel = getattr(self, left)
            if isinstance(myrel, list):
                mydict[left] = [rel.deep_dict(relations=right) for rel in myrel]
            else:
                mydict[left] = myrel.deep_dict(relations=right)
    # if we get here check and handle if we were passed a string
    elif isinstance(relations, str):
        # passed a single item
        myrel = getattr(self, relations)
        left = relations
        if isinstance(myrel, list):
            mydict[left] = [rel.deep_dict(relations=None)
                                 for rel in myrel]
        else:
            mydict[left] = myrel.deep_dict(relations=None)

    return mydict

举个关于person/family/homes/rooms的例子…把它转换成json，你只需要

json.dumps(person.deep_dict(relations={'family':{'homes':'rooms'}}))

出于安全考虑，您不应该返回模型的所有字段。我更喜欢有选择性地选择他们。

Flask的json编码现在支持UUID, datetime和relationships(并为flask_sqlalchemy db添加了query和query_class。模型类)。编码器我更新如下:

app / json_encoder.py

    from sqlalchemy.ext.declarative import DeclarativeMeta
    from flask import json


    class AlchemyEncoder(json.JSONEncoder):
        def default(self, o):
            if isinstance(o.__class__, DeclarativeMeta):
                data = {}
                fields = o.__json__() if hasattr(o, '__json__') else dir(o)
                for field in [f for f in fields if not f.startswith('_') and f not in ['metadata', 'query', 'query_class']]:
                    value = o.__getattribute__(field)
                    try:
                        json.dumps(value)
                        data[field] = value
                    except TypeError:
                        data[field] = None
                return data
            return json.JSONEncoder.default(self, o)

app / __init__ . py

# json encoding
from app.json_encoder import AlchemyEncoder
app.json_encoder = AlchemyEncoder

有了这个，我可以选择添加一个__json__属性，返回我希望编码的字段列表:

app / models.py

class Queue(db.Model):
    id = db.Column(db.Integer, primary_key=True)
    song_id = db.Column(db.Integer, db.ForeignKey('song.id'), unique=True, nullable=False)
    song = db.relationship('Song', lazy='joined')
    type = db.Column(db.String(20), server_default=u'audio/mpeg')
    src = db.Column(db.String(255), nullable=False)
    created_at = db.Column(db.DateTime, server_default=db.func.now())
    updated_at = db.Column(db.DateTime, server_default=db.func.now(), onupdate=db.func.now())

    def __init__(self, song):
        self.song = song
        self.src = song.full_path

    def __json__(self):
        return ['song', 'src', 'type', 'created_at']

我添加@jsonapi到我的视图，返回结果列表，然后我的输出如下:

[

{

    "created_at": "Thu, 23 Jul 2015 11:36:53 GMT",
    "song": 

        {
            "full_path": "/static/music/Audioslave/Audioslave [2002]/1 Cochise.mp3",
            "id": 2,
            "path_name": "Audioslave/Audioslave [2002]/1 Cochise.mp3"
        },
    "src": "/static/music/Audioslave/Audioslave [2002]/1 Cochise.mp3",
    "type": "audio/mpeg"
}

]

也许你可以使用这样的类

from sqlalchemy.ext.declarative import declared_attr
from sqlalchemy import Table


class Custom:
    """Some custom logic here!"""

    __table__: Table  # def for mypy

    @declared_attr
    def __tablename__(cls):  # pylint: disable=no-self-argument
        return cls.__name__  # pylint: disable= no-member

    def to_dict(self) -> Dict[str, Any]:
        """Serializes only column data."""
        return {c.name: getattr(self, c.name) for c in self.__table__.columns}

Base = declarative_base(cls=Custom)

class MyOwnTable(Base):
    #COLUMNS!

所有对象都有to_dict方法

Python 3.7+将于2023年发布

您可以将数据类装饰器添加到您的模型中，并定义一个自定义JSON序列化器，然后是JSON。转储将工作(通过向cls提供自定义编码器)。在下面的例子中，db_row是DB类的一个实例:

json.dumps(db_row, cls=models.CustomJSONEncoder)

{"id": 25, "name": "A component", "author": "Bob", "modified": "2023-02-08T11:49:15.675837"}

可以很容易地修改定制JSON序列化器，使其与任何原生JSON不可序列化的类型兼容。

models.py

from datetime import datetime
import dataclasses
import json
from sqlalchemy import Column, Integer, String, DateTime
from database import Base


@dataclasses.dataclass # <<-- add this decorator 
class DB(Base):
    """Model used for SQLite database entries."""

    __tablename__ = "components"

    id: int = Column(Integer, primary_key=True, index=True)
    name: str = Column(String)
    author: str = Column(String)
    modified: datetime = Column(DateTime(timezone=True), default=datetime.utcnow)


class CustomJSONEncoder(json.JSONEncoder): # <<-- Add this custom encoder 
    """Custom JSON encoder for the DB class."""

    def default(self, o):
        if dataclasses.is_dataclass(o): # this serializes anything dataclass can handle  
            return dataclasses.asdict(o)
        if isinstance(o, datetime): # this adds support for datetime
            return o.isoformat()
        return super().default(o)

为了进一步扩展它，使它适用于你在数据库中可能使用的任何不可序列化的类型，在自定义编码器类中添加另一条if语句，返回一些可序列化的东西(例如str)。

(Sasha B的回答非常棒)

这特别地将datetime对象转换为字符串，在原始答案中将转换为None:

# Standard library imports
from datetime import datetime
import json

# 3rd party imports
from sqlalchemy.ext.declarative import DeclarativeMeta

class JsonEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj.__class__, DeclarativeMeta):
            dict = {}

            # Remove invalid fields and just get the column attributes
            columns = [x for x in dir(obj) if not x.startswith("_") and x != "metadata"]

            for column in columns:
                value = obj.__getattribute__(column)

                try:
                    json.dumps(value)
                    dict[column] = value
                except TypeError:
                    if isinstance(value, datetime):
                        dict[column] = value.__str__()
                    else:
                        dict[column] = None
            return dict

        return json.JSONEncoder.default(self, obj)