如何通过小袋鼠自动化浏览器回滚对数据库所做的更改

Question

I have long test that call a lot of functions that test the all application. I looking for a way that in case of failure in the test, I can rollback the specific changes that the automation did on the test.

For example:

test "add user and login", session do
  session
  |> add_user()
  # There can be more functions here...
end

def add_user(session, loops // 2) do
  try do
    session
    |> visit("example.com")
    |> fill_in(css("#user_name", with "John Doe")
    |> click(css("#add_user_button"))
    |> assert_has(css("#user_added_successfully_message")
  rescue
    msg -> if loops > 0, do: add_user(session, loops - 1), else: raise msg 
  end
end

In case of failure on the assert_has function (the user aded but the message don't show up), I want to rollback all the changes that happened on the database before the add_user function called again in the rescue.

Answer 1

In case you use Ecto to access the DB everywhere, you can use its sandbox mode .

Most likely you want to configure the sandbox pool in your config/test.exs (if not already there):

config :my_app, Repo, # my_app being the name of the application that holds the Repo
  pool: Ecto.Adapters.SQL.Sandbox

Then in your test_helper or tests do something like this:

setup do
  :ok = Ecto.Adapters.SQL.Sandbox.checkout(Repo)
  Ecto.Adapters.SQL.Sandbox.mode(Repo, {:shared, self()})
end

With this, all your tests run in separate transactions and are rolled back afterwards. Also the above makes sure that all the processes use the same connection and see the same transaction data (see :shared mode in the docs). This example is taken from the the docs where there is more information on that.

If you can not use Ectos sandbox mode for whatever reason, a good option could be to start a database transaction yourself and share the connection between your test and the code under test. That way you can manually roll back after each test.