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.