dbt models that declare no columns

This rule flags dbt models whose schema YAML lists no columns at all. A model that declares even one column passes here, and the column rules take it from there.

A declared column list is what every column description, column test and contract hangs off. Without one there is nothing to check a schema change against, so a renamed or dropped column goes unnoticed.

Declare column schemas for the models that have none. Declare the model's columns in its schema YAML. The dbt-codegen package can generate the stub from the built model, so you only fill in descriptions and tests.

Before:

models:
  - name: dim_customers
    description: One row per customer.

After:

models:
  - name: dim_customers
    description: One row per customer.
    columns:
      - name: customer_id
      - name: first_order_at
      - name: lifetime_value

A short-lived or scratch model that nothing else reads. For everything else the stub costs a minute to generate.

manifest.json. If no model or source in the manifest carries a columns object at all, we cannot tell a missing schema from a manifest built without one. The rule then says it could not assess, rather than flagging every model.