SQLAlchemy One To Many Updates Failing - sqlalchemy

Question

Thank you in advance for your feedback on this. I've check quite a few sources including StackOverflow and still haven't found a resolution that works for this use case.
I'm developing a CRUD API using SQLAlchemy and the goal is to pass in a Pydantic object and update the database to reflect whatever is in the Pydantic object (including lists of attributes). This seems to only be an issue when the "many" table can't have a composite primary key because one of the fields is nullable.
Below is a very simple example where we have organizations and the organizations can have partner-preferred alternative names or generic alternative names. Essentially, you would be able to say that the name of the organization is "Microsoft" but the New York Stock Exchange refers to it as "MSFT". Additionally, in the past, we've seen the incorrect version, "Micro-Soft".
# Create the classes
class Organization(Base):
__tablename__='organization'
org_id=Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
org_name=Column(String(255))
alt_names = relationship("AltOrgName", lazy=False)
__table_args__={'schema': 'dbo'}
class AltOrgName(Base):
__tablename__='alt_org_name'
alt_org_name_id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
org_id=Column(UUID(as_uuid=True), ForeignKey(Organization.org_id))
alt_org_name = Column(String(255))
partner = Column(String(10)) # <-----Nullable, so it can't be part of the primary key
__table_args__=(UniqueConstraint(org_id, alt_org_name, partner), {'schema': 'dbo'})
From there you can create an organization with an alternative name:
# Create the organization
o1 = Organization(org_name="Microsoft",
alt_names = [AltOrgName(alt_org_name="MSFT", partner="New York Stock Exchange")]
)
db.add(o1)
db.commit()
Now I would like to fully overwrite the former version of the alternative names. In practice, this would be constructed from a Pydantic model, but this is functionally similar enough:
for an in o1.alt_names:
db.delete(an)
o1.alt_names = []
o1.alt_names.append(AltOrgName(alt_org_name="MSFT", partner="New York Stock Exchange"))
o1.alt_names.append(AltOrgName(alt_org_name="Micro-Soft", partner=None))
db.commit()
This leads to an integrity error:
IntegrityError: (pyodbc.IntegrityError) ('23000', "[23000] [Microsoft][ODBC Driver 17 for SQL Server][SQL Server]Violation of UNIQUE KEY constraint 'UQ__alt_org___89413C3E3CD4661D'. Cannot insert duplicate key in object 'dbo.alt_org_name'. The duplicate key value is (c8e907b3-fc71-40de-8cf1-0806e2ebbce6, MSFT, New York Stock Exchange). (2627) (SQLExecDirectW); [23000] [Microsoft][ODBC Driver 17 for SQL Server][SQL Server]The statement has been terminated. (3621)")
[SQL: INSERT INTO dbo.alt_org_name (alt_org_name_id, org_id, alt_org_name, partner) VALUES (?, ?, ?, ?)]
[parameters: ('d828928d-ba4a-49e9-989e-44fda8027de5', 'c8e907b3-fc71-40de-8cf1-0806e2ebbce6', 'MSFT', 'New York Stock Exchange')]
However, committing after the delete works.
for an in o1.alt_names:
db.delete(an)
db.commit() # <----------New line of code
o1.alt_names = []
o1.alt_names.append(AltOrgName(alt_org_name="MSFT", partner="New York Stock Exchange"))
o1.alt_names.append(AltOrgName(alt_org_name="Micro-Soft", partner=None))
db.commit()
Having two commits is an extremely undesirable code pattern since it creates the risk that the delete commit will work but the revision commit won't--which will ultimately wipe out all alternative names for the company. For instance, the code below would result in an integrity error and wipe out all alternative names for the organization without correctly replacing them:
for an in o1.alt_names:
db.delete(an)
db.commit()
o1.alt_names = []
o1.alt_names.append(AltOrgName(alt_org_name="MSFT", partner="New York Stock Exchange"))
o1.alt_names.append(AltOrgName(alt_org_name="Micro-Soft", partner=None))
o1.alt_names.append(AltOrgName(alt_org_name="Micro-Soft", partner=None)) # <-------------Duplicate
db.commit()
The issue definitely seems to be sqlalchemy's order of operations. If it performed the delete command first, there would be no issue. However, for some reason, it seems like the old alternative names are somehow still in the database at the time it attempts to insert the new ones. One thought is to somehow tell SQLAlchemy to perform the delete before attempting the inserts.
The other resolution is to make the primary key of the alt_org_name table (org_id, alt_org_name, partner) and make the null value for partner an empty string. However, I have a very similar situation where one of the nullable fields is a number so I would prefer a more robust solution if one exists.
Grateful for any ideas or suggestions.
EDIT
I got a response back from SQLAlchemy's team and adding db.flush() seems to fix it
for an in o1.alt_names:
db.delete(an)
db.flush() # <---- New line of code
o1.alt_names = []
o1.alt_names.append(AltOrgName(alt_org_name="MSFT", partner="New York Stock Exchange"))
o1.alt_names.append(AltOrgName(alt_org_name="Micro-Soft", partner=None))
db.commit()