Using SQLAlchemy ORM for a non-primary key, unique, auto-incrementing id

a workaroud here for mysql:

After sqlalchemy create table,
　alter table manully (add AUTO_INCREMENT property) use DDL.
　( at table after_create event ).

then it will execute on demand like create_all()

test code (1.4.x):

from sqlalchemy.ext.declarative import declarative_base

# --- 1. define table
base = declarative_base()
class store(base):
    __tablename__ = 'store'
    id  = Column(Integer,    autoincrement=True, unique=True, primary_key=False) # AI here not work

    did = Column(String(64), unique=False, nullable=False)
    fid = Column(String(64), unique=False, nullable=False)

    __table_args__ = (
        PrimaryKeyConstraint('did', 'fid', name='idx_did_fid'),
    )


print(store)

# --- 2. fix autoincre on non-primary key // will execute as need, like create_all() dose
from sqlalchemy import event, DDL
event.listen(
    base.metadata, 'after_create',  DDL('ALTER TABLE `store` CHANGE `id` `id` INT(11) NULL DEFAULT NULL AUTO_INCREMENT')
)

# --- 3. create table as need
base.metadata.create_all(engine)

Postgresql 2022 answer:

the following code

import uuid
from sqlalchemy import Integer, Column, String, Sequence
from sqlalchemy.dialects.postgresql import UUID

class Test(Base):
    __tablename__ = 'Test'

    id_sec = Sequence(__tablename__ + "_id_seq")
    id = Column(Integer, id_sec, server_default=id_sec.next_value(), nullable=False)
    uuid = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)

    # your data
    .... 

generates the following table:

                                      Table "public.Test"
    Column     |       Type        | Collation | Nullable |                   Default                    
---------------+-------------------+-----------+----------+----------------------------------------------
 id            | integer           |           | not null | nextval('"Test_id_seq"'::regclass)
 uuid          | uuid              |           | not null | 

so you can insert rows without specifying the id

PostgreSQL 13.5
SQLAlchemy==1.4.29

SQLAlchemy 1.4 and above

In addition to sequences, you can use sqlalchemy.schema.Identity with some supported databases. OP’s example becomes:

from sqlalchemy import Identity
...

Base = declarative_base()
Session = sessionmaker(bind=db)
session = Session()

class Person(Base):
    __tablename__ = "people"
    id = Column(Integer, Identity(), index=True, unique=True)
    first_name = Column(String(30), primary_key=True)
    last_name = Column(String(30), primary_key=True)

if __name__ == "__main__":
    Base.metadata.create_all(db)
    session.add_all([
        Person(first_name="Winston", last_name="Moy"),
        Person(first_name="Bill", last_name="Gates"),
        Person(first_name="Steve", last_name="Jobs"),
        Person(first_name="Quinten", last_name="Coldwater")
    ])
    session.commit()

As of writing, the Identity construct is currently known to be supported by:

• PostgreSQL as of version 10.
• Oracle as of version 12. It also supports passing always=None to enable the default generated mode and the parameter on_null=True to specify “ON NULL” in conjunction with a “BY DEFAULT” identity column.
• Microsoft SQL Server. MSSQL uses a custom syntax that only supports the start and increment parameters, and ignores all other.

For more info on arguments and supported databases, see the following

• https://docs.sqlalchemy.org/en/14/core/defaults.html#sqlalchemy.schema.Identity
• https://docs.sqlalchemy.org/en/14/core/defaults.html#identity-ddl