SKILL.md

name	lookml-view
description	Use this skill to create or modify LookML Views. Covers basic view definitions, sql_table_name, file organization, and patterns.
license	Apache-2.0
metadata	publisher version google v1

Instructions

1. File Organization

• Standard Views: views/[datasource_name]/[view_name].view.lkml
• Extended Views: views/[datasource_name]/[view_name]_ext.view.lkml
• Refinements: views/[datasource_name]/[view_name]_rfn.view.lkml
• NDTs: views/[datasource_name]/[view_name]_ndt.view.lkml
• SDTs: views/[datasource_name]/[view_name]_sdt.view.lkml

2. Core Standards

1. sql_table_name: Required for standard views. Defaults are fragile; be explicit.
2. Primary Key:
   • Convention: place the primary key as the first dimension for readability.
   • Must have primary_key: yes — Why: essential for symmetric aggregates, which Looker uses to handle JOIN row duplication correctly (prevents fanout in multi-join queries).
3. Refinements: Use + before the view name (e.g., view: +users) to refine existing views without modifying the original file.
4. Extensions: Use extends to reuse logic from other views.
5. Derived Tables: Evaluate and nudge towards Native Derived Tables (NDTs) (explore_source) over SQL Derived Tables (sql) where possible. When creating entity rollups or summary facts (e.g., user_order_facts), assess whether an existing Explore can source the data. NDTs are recommended because they minimize code drift and automatically inherit underlying LookML descriptions, formatting, and Row-Level Security (RLS). If an NDT is too complex, or if specific SQL features are required, an SDT is fully acceptable.

3. Best Practices

• Naming: Use snake_case for view names and filenames.
• Descriptions: Add descriptions to the view itself if it helps explain its purpose (e.g., "Daily active users aggregated by region").
• Output Format: Ensure the file ends with a newline.

4. Related Skills

• lookml-refinements: Logic for include, refinements (+), and Hub & Spoke patterns.

Examples

Basic View

view: users {
  sql_table_name: `project.dataset.users` ;;

  dimension: user_id {
    primary_key: yes
    type: number
    sql: ${TABLE}.id ;;
    group_label: "IDs"
    description: "Unique internal user ID."
  }

  dimension: email {
    type: string
    sql: ${TABLE}.email ;;
    description: "User's email address."
  }
}

Extended View

include: "/views/base/users.view"

view: users_extended {
  extends: [users]

  dimension: lifetime_value {
    type: number
    sql: ${TABLE}.ltv ;;
    description: "Total lifetime value of the user."
  }
}

Refinement (Layering)

include: "/views/users.view"

view: +users {
  label: "All Users (Refined)"
  
  dimension: email {
    # Adding a description to an existing field
    description: "Primary contact email."
  }
}

Reference Skills

• Derived Tables: Standard (SDT) and Native (NDT) derived table patterns/materialization.