Usage

Droughty depends on a number factors to be useful. These factors can be split into two considerations, tooling and development approaches.

Tooling

Droughty is most useful when using one of the supported warehouses (goes without saying), dbt, looker and dbml. With these all being present, droughty can nicely fit within a development workflow and automate some of the boring stuff.

Development approaches

Looker approaches

Droughty outputs a base.layer which contains dimensions from the analytics layer. This is most useful when using either a series of refinements or extends that sit over the base layer. It’s best then paired with a layered modelling approach with a staging, integration and aggregate layer.

Why is this useful?

  • it allows you to have a floating base layer that’s coupled to the warehouse through machine generated code, saving time and perhaps reducing error

  • it ensures that the labelling and descriptions are identical within the warehouse and looker.

  • there’s scope to automate the generation of measures

Warehousing approaches

Droughty outputs lookml, dbml and dbt schemas based on an dataset/schema warehouse structure. It depends on a separate dataset/schema for the staging, integration and the warehouse/analytics layer. This is pretty easy to achieve with these dbt_project.yml configs:

models:
+persist_docs:
    relation: true
    columns: true

example_dw:
    staging:
        materialized: view
        schema: staging
    integration:
        materialized: view
        schema: integration
    warehouse:
        materialized: table
    data_quality:
        materialized: table
        schema: data_quality
    utils:
        materialized: table
        schema: logs

Droughty also depends on some naming conventions

  • primary key’s end in pk

  • foreign key’s end in fk

Summary

If you meet the majority of the requirements above, there’s a chance this will be useful… I personally use droughty with the following steps:

  • I run dbt-test to generate tests as I go throughout my dbt dev work

  • I run dbml before raising a PR so that there’s an update ERD of the warehouse for documentation and to include in the PR Summary

  • I run lookml as a pre-commit hook so that looker is always up-to-date with the warehouse

Any questions, raise them in GitHub. Thanks.