Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.
see 5 files with indirect coverage changes

I don't understand why this is preferable. What issues were you having seeding the database. The current solution is idempotent

When pytest is run it creates a bunch of data, thereby auto-incrementing the id in each table. Then, when behave is run it adds the different objects. Because of the auto-incrementing from pytest the hardcoded ids are out of date. This causes two issues with the data that is being seeded:

1. if sh run_bdd.sh is run more than once, it will try to add a second sensor with the same serial_no, which breaks the unique constraint.
2. the hardcoded thing_id and sensor_id in the deployment are no longer valid.

By keeping it flexible and tearing down the data after each test we won't have to worry about ids and the order in which tests are run (pytest or behave), and during development we can call sh run_bdd.sh any number of times and the unique constraints won't be broken.

I don't see the point of removing seed data. Removing data added by a test is acceptable, but removing seed data solely to re-add it does not make sense. The seed functions need to be idempotent (and for the most part, they are) without requiring the removal of data add the end of a test suite.