lookml-tests

LookML Testing Standards

Testing is critical for maintaining trust in data. LookML tests allow us to verify that our semantic model behaves as expected and that the underlying data conforms to our assumptions.

1. File Organization

• Location: Define tests in tests/[explore_name].test.lkml.
• One Suite Per Explore: Each file should contain all the test definitions for a specific Explore.
• Naming Convention: [explore_name].test.lkml (e.g., orders.test.lkml).

2. Test Structure

Each test consists of an explore_source query and an assert statement.

test: [test_name] {
  explore_source: [explore_name] {
    column: [column_name] { field: [view_name].[field_name] }
    filters: {
      field: [view_name].[field_name]
      value: "[value]"
    }
  }

  assert: [assertion_name] {
    expression: ${[view_name].[field_name]} [operator] [value] ;;
  }
}

3. Types of Tests

A. Integrity Checks (Critical)

Verify that Primary Keys remain unique after joins. This is the best defense against "fanout" errors caused by incorrect one_to_many join definitions.

Example: Primary Key Uniqueness

test: orders_pk_is_unique {
  explore_source: orders {
    column: order_id {}
    column: count {}
    # Limit to recent data to save costs/time if table is large
    filters: {
      field: orders.created_date
      value: "last 7 days"
    }
  }

  assert: order_id_is_unique {
    expression: ${orders.count} = 1 ;;
  }
}

B. Accuracy Tests

Validate specific measure values against known constants or expectations.

Example: Revenue is Positive

test: revenue_is_positive {
  explore_source: orders {
    column: total_revenue {}
    filters: {
      field: orders.created_date
      value: "yesterday"
    }
  }

  assert: revenue_greater_than_zero {
    expression: ${orders.total_revenue} >= 0 ;;
  }
}

C. Business Logic Validation

Ensure calculations behave as expected. For example, checking that gross_margin is never greater than revenue or that lifetime_orders is never NULL for an active user.

Example: Logic Check

test: margin_less_than_revenue {
  explore_source: orders {
    column: total_revenue {}
    column: total_margin {}
  }
  
  assert: margin_is_valid {
    expression: ${orders.total_margin} <= ${orders.total_revenue} ;;
  }
}

4. Best Practices

• Descriptive Extensions: Use informative names for tests (orders_pk_is_unique) and assertions (order_id_is_unique).
• Performance: Use filters (e.g., last 7 days) to limit the scan size for large tables, unless verifying full history is required.
• Model Inclusion: Ensure test files are included in the model file (e.g., include: "/tests/*.test.lkml").