EddytorDocs & API
07 · Tables A→ZB · Plan the table

Naming conventions that age well

Schema, table, and column naming that stays readable as the catalog grows.

Names are cheap to set and expensive to change across many tables and downstream consumers. A little consistency up front keeps the catalog navigable.

Schemas & tables

  • Lowercase, singular-or-consistent nouns. Pick order vs orders and stick to it across the catalog.
  • Group by domain or pipeline stage in the schema: sales.orders, raw.events, curated.customers, reference.countries.
  • No environment in the name - separate dev/prod with a different storage connection, not orders_prod.
  • Avoid dates/versions in table names - versioning is built in (table history); don't bake orders_2026 into the name.

Columns

  • snake_case, descriptive, source-aligned. Column names are case-sensitive - match the source system and be consistent.
  • Suffix types of keys consistently: _id for identifiers, _at for timestamps, is_ for booleans (is_active).
  • Spell out rather than abbreviate cryptically (category not cat).

Why it matters

Names flow into the FQN, the sidebar, SQL queries, and every downstream tool reading the Delta table. Renaming a column or table later means updating consumers - so the convention you choose now is the one you live with.

Good to know

You can rename a table later, but its stable UUID - not its name - is its identity. Pick names humans read; let the UUID handle identity.

Next

On this page