dbt YAML columns missing from the warehouse

This rule flags models whose schema YAML declares a column the built relation does not have. Names are compared ignoring case. A relation for which the catalog lists no columns is skipped, as is a model not in the catalog.

A test or contract on a missing column protects nothing. The documentation also describes a shape the data does not have, so readers plan queries against columns that are not there.

Reconcile the declared columns with the ones that exist. Select the column in the model, or remove it from the schema YAML along with the tests that name it.

Before:

models:
  - name: fct_orders
    columns:
      - name: order_id
      - name: discount_code
        data_tests:
          - not_null

After:

models:
  - name: fct_orders
    columns:
      - name: order_id

A column added to YAML ahead of the model change that selects it is fine for as long as that change is in flight. Build it before the tests are relied on.

manifest.json and catalog.json, generated together by dbt docs generate. If the catalog is older than the manifest, a missing column could be an unbuilt change, so the rule declines.