linter: add require-enum-value-ordering rule

[cabello]

Adds a new lint rule require-enum-value-ordering that flags ALTER TYPE ... ADD VALUE statements missing BEFORE or AFTER.

Without explicit ordering, new enum values are silently appended to the end, which can cause unexpected ordering in status workflows, severity, priority, etc.

[netlify]

👷 Deploy request for squawkhq pending review.

Visit the deploys page to approve it

Name	Link
🔨 Latest commit	d79aac1

[cabello]

Hi 👋
We have our own migration system https://github.com/marcduez/pg-migrate and we want as much as possible for db migrations to be idempotent, safe and aligned, and we noticed that if two developers touch the same enum we might end up with a distinct set of schema dump depending on the order they run each migration.
Thank you for considering this change 🙇

[sbdchd]

Cool rule, thank you for putting this together!

I was curious how Django handles this, but turns out they don't support Postgres enum types unfortunately.

• Use native enums if the database supports them django-commons/django-enum#19
• https://code.djangoproject.com/ticket/24342

For the docs, I think it might be helpful to show the query pattern aka the select query you're making that depends on the ordering. Do you have an example?

[cabello]

Thank you for the reply, we don't use ORMs so we try to benefit to all features the database has datatypes, triggers, materialized views, etc.

It's the output of a pg_dump of the database, we commit that to git so CI can break if it changes after running migrations; it includes enum creation and internally Postgres must use some tinyint to represent that, it's not trivial to manipulate enums in Postgres.

If dev A dumps the schema, commits and the values were A B C then the CI runs it and due to the order of execution the values are A C B, then that dev has to find a way to sync with master otherwise they will always be reverting that change.