Schema design standards and naming

Purpose

Schemas that are obvious to read and cheap to change.

snake_case; plural table names (panels), singular column names; primary key id uuid default gen_random_uuid().
Every table has created_at timestamptz default now(); mutable tables add updated_at (trigger-maintained) and updated_by.
Foreign keys always declared, named <table>_id, with explicit on delete behavior.
Units in column names (width_mm, weight_kg, angle_deg); store SI, convert at the edge.
Enums as Postgres enum types or text + check constraint; prefer check when values will change.
jsonb for document-shaped data (editor content, tool settings); never for data you'll filter or join on.
Indexes on every FK and every column used in where/order by; no premature composite indexes.
Soft delete (deleted_at) only where restore is a real requirement; otherwise hard delete with on delete cascade thought through.
Comment every table (comment on table … is '…') — it shows up in the dashboard and in generated types.

Storing arrays of ids instead of join tables; varchar(255) by habit (use text); nullable everything.

Owner: Matt · Last reviewed: 2026-09