I presume there can be a case when someone wants to run the rollback task manually. Introduce a config for that.

Create a mechanism that allows to roll back the phantom migrations.

Ideas on how to implement it (can be implemented both):

1. Add an option to the command line: rake db:rollback_branches --dry-run
2. Via env var. Apply dry run if ACTUAL_DB_SCHEMA_ROLLBACK_BRANCHES_DRY_RUN=true

The dry run should not roll back any migrations but rather output the phantom migrations list.

Prefer the same output format in the output as in the following Rails' task:

rails db:migrate:status

 Status   Migration ID    Migration Name
--------------------------------------------------
   up     20220923211114  Initial
   up     20221003092332  Add typical clients and samples links to vendor applications
   up     20221003094301  Add vendor id index to messages
   up     20221004031502  Add vendor company to users
   up     20221115233057  Add status to vendor applications
   up     20221129184509  Add type to quotes
   up     20221207181413  Create admin user approvals
   up     20221223154500  Add templates for sale to vendor application
   up     20230105173250  Create categories
   up     20230113173555  Add data to projects
   up     20230120192110  Create services
   up     20230126045226  Add terms accepted at to users
   up     20230131174014  Add active to services

This is a very cool gem, but since my app has multiple databases, I can't use it.

It's explicitly disallowed here:

Though it might be some work, in principle this feels doable to me. I'm happy to make an attempt at a PR.

Add info that it can be disabled and used manually. Add necessary functionality if needed.

I want to be able to roll back specific phantom migrations manually one by one from any branch that has the applied phantom migrations.

• Use guard clause to return if not dev mode
• Fix typo in Path

When saving the run migration in the temp folder, also remember the branch name (in a comment?).

Later, it can be used to show the status.

As the title, it destroy data when happen migration rollback.
Can the data be automatically exported before rollback, and then imported back when migrating again？

PS: English is not my native language; please excuse typing errors.

The tests should check:

• When migration is up it "remembers" the run migrations inside the tmp folder.
• If migration exists in tmp folder but if it's absent in migrations folder, it should be rolled back.
• If migration is not reversible it should report the message.

When the gem automatically rolls back, it's unclear why it was rolled back. Show some information in the output.

For example, this is a migration that was tried to be rolled back on running rails db:migrate:

== 20240208135613 AddNotNullConstraintToCategoryIdInProjects: migrating =======
-- change_column_null(:projects, :category_id, false)
   -> 0.0422s
== 20240208135613 AddNotNullConstraintToCategoryIdInProjects: migrated (0.0423s)

== 20240212121941 CreateQueSchema: reverting ==================================
rails aborted!
StandardError: An error has occurred, this and all later migrations canceled:

uninitialized constant CreateQueSchema::Que

    Que.migrate!(version: 0)
    ^^^
Did you mean?  Queue
~/tmp/migrated/20240212121941_create_que_schema.rb:11:in `down'

Caused by:
NameError: uninitialized constant CreateQueSchema::Que

    Que.migrate!(version: 0)
    ^^^
Did you mean?  Queue
~/tmp/migrated/20240212121941_create_que_schema.rb:11:in `down'
Tasks: TOP => db:_dump => db:rollback_branches
(See full trace by running task with --trace)

Note that it needs to be clarified why it was tried to roll it back. Add some info like: actual_db_schema rolls back this migration.

Also, it should not stop the migration on error raised, suppress it, show the error in the output, and move on.

I am a user of this gem. I have spotted what I might call a typo in the README.

The README explains that the location for the phantom migrations is tmp/migrations. It's actually tmp/migrated.

I made the above observation while testing out the library. I was looking for tmp/migrations but I didn't find it. Turns out that it was tmp/migrated.

we can also see that in the implementations here:

• actual_db_schema/lib/actual_db_schema.rb

  Line 31 in 3b5c4ad

  def self.migrated_folder

I suppose it would be worthwhile to update the README to avoid confusion for other potential users.

Put the rake tasks into their own namespace to avoid possible collision with other gems.

Update readme with an explanation of how to use Gemfile.local

warning: csv was loaded from the standard library, but will no longer be part of the default gems since Ruby 3.4.0. Add csv to your Gemfile or gemspec.