I think it's really a pity CASCADE support is not added ATM. Hopefully it can be added later on, in a not too distant future. Obviously SQLite already knows what dependents exists, otherwise RENAME COLUMN wouldn't work as it does, so deleting those as well when the DROP COLUMN is qualified with an explicit CASCADE, should have been fairly easy I suspect. That's both standard behavior, but also very common behavior in my experience, to use CASCADE during drops. FWIW. > PS: If SQLite had a pragma to list dependencies between objects of the schema, then at least one would be able to programmatically do the CASCADE, by deleting those first in topological order, but w/o it, and w/o CASCADE, one has to be intimately aware of schema intricacies, which in large schemas, or schemas you inherited and didn't design yourself (or both!), is not always that easy, far from it.