db-naming (M7)

Naming is the schema's documentation: consistent, predictable identifiers make joins obvious and reduce the quoting/casing bugs that creep in across ORMs and engines. It is low-weight (Naming category, design axis) — real but never the headline. This module is design-axis and never caps.

What it checks

• Casing / pluralization consistency: a mix of snake_case and camelCase, or singular and plural table names (user, orders, OrderItem) in one schema — pick one and hold it.
• Reserved words & case-folding traps: identifiers that are SQL reserved words (user, order, group, select) or that rely on case ("User" quoted) — Postgres folds unquoted to lowercase, MySQL is filesystem/case-config dependent; mixed quoting causes "relation does not exist" bugs.
• FK naming: FK columns not following a predictable <referenced>_id pattern, so joins aren't self-evident.
• Boolean naming: booleans not prefixed is_/has_/can_, or negative names (not_active) that invert logic.
• Opaque / ambiguous names: data, info, value, flag, temp, col1, abbreviations without a glossary, or timestamps not suffixed _at.

Axis & severity

• Axis: design; magnitude almost always low, banded honestly.
• Reserved-word unquoted identifier that breaks across engines: severity 2–3, warn.
• Schema-wide casing/pluralization inconsistency: severity 2, warn, fixable: proposed (rename is high-blast-radius — never auto).
• Opaque column name / missing _at suffix: severity 1, warn.
• M7 never caps any score; it is the smallest-weight design category.