db-schema / core Goto Github PK

Postgres allows renaming indexes, and while it is impossible to use inside the schema diffing algorithm (as indexes are compared by name) it can still be used inside conditional migrations.

Adding an Operations::RenameIndex (and respective rename_index in the Migrator) and handling it with execute('ALTER INDEX name RENAME TO new_name') in the Reader should do it.

If the database schema contains any check constraints that mention an enum field, then running AlterEnumValues against that enum type fails.

A possible solution would be to drop check constraints before recreating the enum and create them back afterwards.

Cloud Firestore is a very popular and useful database. It would be nice if you include it

Hello!

I like your idea and I want to help.

Reordering and deleting values from enum types is not supported.

You can do it!

Sequel migration example:

Sequel.migration do
  no_transaction

  up do
    add_enum_value(:enum_locale, 'fr')
  end

  down do
    enum_values = schema(:users).to_h[:locale][:enum_values]

    alter_table :users do
      set_column_type :locale, String
    end

    self[:users].where(locale: 'fr').update(locale: 'en')

    drop_enum :enum_locale
    create_enum :enum_locale, enum_values - %w(fr)

    alter_table table do
      set_column_type(
        :locale,
        :enum_locale,
        using: 'locale::enum_locale'
      )
    end
  end
end

A temporary table created in Normalizer can depend on enums and/or extensions that may not exist yet (for example when applying the schema to an empty database).