6.1.1 - 2026-03-17

Summary

This update fixes a bug that specifically impacted ORM sessions when flushing values to pass IDs between linked SQLAlchemy tables. We discovered that the internal get_lastrowid function has been off by one since the dialect’s inception (2015) before Exasol’s direct involvement, a bug that previously went unnoticed due to a gap in test coverage. If you were affected, this discrepancy caused incorrect ID referencing during data inserts, potentially breaking relational integrity in your sessions. We have now fixed the logic to ensure the returned IDs are accurate and implemented new tests that specifically validate row ID retrieval during ORM flushes.

The error would have only happened for ORM sessions like this example:

with Session(ENGINE) as session:
    new_user = User(first_name="Jax", last_name="Doe")

    session.add(new_user)
    session.flush()

    # Here, we use the ID from the post-flush SQLAlchemy object. Before this update,
    # this value was off by 1 (ID - 1). Now, this has been corrected (ID).
    new_email = EmailAddress(
        user_id=new_user.id,
        email_address="jax.doe@example.com",
    )

    session.add(new_email)
    session.commit()

While this method ensures accurate ID retrieval, it is not the most performant way to insert data in SQLAlchemy-Exasol. For high-volume inserts, consider using more efficient bulk processing methods.

Refactoring

• #724: Updated to exasol-toolbox 6.0.0

Bugfix

• #709: Fixed ORM session usage so that session IDs when flushed are not off by one

Dependency Updates

dev

• Updated dependency exasol-toolbox:5.1.0 to 6.0.0