sqlalchemy 'relation does not exist' after flask migrate

[andre-qumulo]

The detailed write up is here: https://stackoverflow.com/questions/78481263/sqlalchemy-relation-does-not-exist-after-flask-migrate

but the short version is that I'm upgrading a Flask-SqlAlchemy app from SqlAlchemy 1.3 to SqlAlchemy 2.0. I have a unit test that walks through the migrations and am finding that when I do a db.session.execute(text('INSERT INTO ... on a table that is dropped and then added in the migration then I get an error from psychopg2 that the relation does not exist.

I am trying to understand if this is a Flask-migrate related issue or if there's somewhere else I should look.

Our flask migration code looks like:

    @contextlib.contextmanager
    def app_context_at_revision(self, revision_title: str) -> Iterator[None]:
        revision_hash = revision_title.split('_')[0]

        with self.app.app_context():
            flask_migrate.upgrade(directory=MIGRATION_DIR, revision=revision_hash)

        with self.app.app_context():
            yield

and we do a migrate.init_app(app, db) elsewhere in the code.

I can see the table create in the output (detailed in the stack overflow ticket) but it's like it's ignoring the migration entirely.

[miguelgrinberg]

Does the table exist in your database or not? It is not clear from your question if the problem is that the table does not get created, or that the table does exist but somehow the SQL query isn't seeing it.

[andre-qumulo]

I was able to get into the DB after running flask_migrate.upgrade that should have upgraded to several revisions. It is showing no relations.

[andre-qumulo]

My apologies for the confusion, I've been learning the system a bit better as I've been trying to debug this.

I've been comparing the behavior of the code to the previous one and am finding that the migration code is running in offline mode where as previously it was online.

I updated the test to run the upgrades as:

flask_migrate.upgrade(
    directory=MIGRATION_DIR, revision=self.initial_migration, sql=True
)

But when I put breakpoints in the env.py file then context.is_offline_mode() is True. I thought passing sql=True would make context.is_offline_mode() False. Any thoughts on why this is occurring?

[miguelgrinberg]

sql=true means that you want alembic to just output SQL code for the migrations, without actually running anything. This is what Alembic calls offline mode. The default mode with sql=false runs the migrations directly on the configured database.

[andre-qumulo]

Ah, thank you. I was getting confused. I replaced my env.py file with the one from this repo, removed sql=True, and any other references to running the migration as sql and now it seems to work...