SQLAlchemy: Getting a single object from joining multiple tables

Question

DB - SQLAlchemy setting

Consider a classic setting of two tables - user and api_key , represented by SQLAlchemy objects as:

class User(Base):
    __tablename__ = 'user'
    user_id = Column(String)
    user_name = Column(String)
    vioozer_api_key = Column(String, ForeignKey("api_key.api_key"))

class ApiKey(Base):
    __tablename__ = 'api_key'    
    api_key = Column(String(37), primary_key=True)

Other fields omitted for clarity.

My Query

Suppose that I want to get the user name and the api key for a specific user id:

user, api_key = database.db_session.query(User, ApiKey)\
    .join(ApiKey, User.vioozer_api_key==ApiKey.api_key)\
    .filter(User.user_id=='user_00000000000000000000000000000000').first()

I get two objects - user and api_key , from which I can fetch user.name and api_key.api_key .

The problem

I would like to wrap this call with a function, which would return a single objects whose fields would be the union of the user fields and the api_key fields - the same way a SQL join returns a table with the columns of both tables being joined. Is there a wayo to automatically get a object whose fields are the union of the fields of both tables?

Wha have I tried

I can define a mapper class for each Join operation, but I wonder if the mapping could be done automatically.

Answer 1

I'm surprised how little discussion there is on this... I need to join two tables and return a single object which has all columns from both tables. I'd expect that is a pretty common usecase - anyways here's the hack I made after hours of looking and failing to find a nice way

from sqlalchemy import inspect
def obj_to_dict(obj):
    return {c.key: getattr(obj, c.key) for c in inspect(obj).mapper.column_attrs}
def flatten_join(tup_list):
    return [{**obj_to_dict(a), **obj_to_dict(b)} for a,b in tup_list]

Here's an example of what it does...

########################## SETUP ##########################
from sqlalchemy import create_engine
from sqlalchemy.orm import sessionmaker
from sqlalchemy import Column, Integer, ForeignKey, Text
from sqlalchemy.ext.declarative import declarative_base

DeclarativeBase = declarative_base()


class Base(DeclarativeBase):
    __abstract__ = True

    def __repr__(self) -> str:
        items = []
        for key in self.__table__._columns._data.keys():
            val = self.__getattribute__(key)
            items.append(f'{key}={val}')
        key_vals = ' '.join(items)
        name = self.__class__.__name__
        return f"<{name}({key_vals})>"


class User(Base):
    __tablename__ = "user"

    id = Column(Integer, primary_key=True, index=True)
    username = Column(Text)


class Todo(Base):
    __tablename__ = "todo"

    id = Column(Integer, primary_key=True, index=True)
    title = Column(Text)
    user_id = Column(Integer, ForeignKey("user.id"))


engine = create_engine("sqlite:///:memory:")
SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)
db = SessionLocal()
Base.metadata.create_all(bind=engine)

########################## DEMO ##########################

db.add(User(**{"id": 1, "username": "jefmason"}))
db.add(Todo(**{"id": 1, "title": "Make coffee", "user_id": 1}))
db.add(Todo(**{"id": 2, "title": "Learn SQLAlchemy", "user_id": 1}))
db.commit()

result = db.query(Todo, User).join(User).all()

print(result)
# > [(<Todo(id=1 title=Make coffee user_id=1)>, <User(id=1 username=jefmason)>), (<Todo(id=2 title=Learn SQLAlchemy user_id=1)>, <User(id=1 username=jefmason)>)]

print(flatten_join(result))
# > [{'id': 1, 'title': 'Make coffee', 'user_id': 1, 'username': 'jefmason'}, {'id': 1, 'title': 'Learn SQLAlchemy', 'user_id': 1, 'username': 'jefmason'}]

Answer 2

Instead of querying objects, query for list of fields instead, in which case SQLAlchemy returns instances of KeyedTuple , which offers KeyedTuple._asdict() method you can use to return arbitrary dictionary:

def my_function(user_id):
    row =  database.db_session.query(User.name, ApiKey.api_key)\
        .join(ApiKey, User.vioozer_api_key==ApiKey.api_key)\
        .filter(User.user_id==user_id).first()
    return row._asdict()

my_data = my_function('user_00000000000000000000000000000000')

But for your particular query, you do not need even to join on ApiKey as the api_key field is present on the User table:

row = database.db_session.query(User.name, User.api_key)\
    .filter(User.user_id==user_id).first()

Answer 3

I would add an api_key relationship on User :

class User(Base):
    __tablename__ = 'user'
    user_id = Column(String)
    user_name = Column(String)
    vioozer_api_key = Column(String, ForeignKey("api_key.api_key"))
    api_key = Relationship('ApiKey', uselist=False)

Then you can do a query like this:

>>> user = database.db_session.query(User)\
      .filter(User.user_id=='user_00000000000000000000000000000000').first()
>>> user.api_key
<ApiKey>