dbt columns with no description

For every dbt model that declares its columns, this rule counts the columns with an empty description. It reports one finding per model, naming the first few columns.

Analysts guess at what an undocumented column means. Two of them guess differently, and two dashboards end up defining the same metric in two ways.

Document the declared columns. Describe the columns, starting with the ones people filter and aggregate on. Where a column means the same thing in several models, write the description once in a {% docs %} block and point each column at it with {{ doc() }}.

Before:

    columns:
      - name: net_revenue
      - name: ordered_at

After:

    columns:
      - name: net_revenue
        description: Order total after discounts and refunds, before tax, in GBP.
      - name: ordered_at
        description: "{{ doc('ordered_at') }}"

Keys and self-explanatory columns such as created_at can go without, if the model's own description says what one row is. Models that declare no columns are not judged here; the rule on undeclared columns covers them.

manifest.json alone.