Questions on Async code using Ecto

Hey Guys. I would like to ask a couple of questions about Ecto and Async tasks.

1 - Are there plans to improve the way we use Ecto + Async Tests + Async Code? Using SQL.Sandbox is in IMHO really confusing.

2 - Are there best practices to use async tests + async code ?

3 - Let’s imagine I have a GenServer Process that inserts 1000 lines each hour using Repo.insert instead of using Repo.insert_all. Would that be a problem to kill my process and my entire application? Does Ecto try to get a new connection from database pool for each insert using Repo.insert?

4 - Let’s imagine I have a async task in my code that does not matter that result to the user, but I want to test it. Should I send a message to the parent process to use assert_receive?

Thank you in advance.

  1. I do not know if it would be possible while keeping all niceties of SQL.Sandbox (especially fact that everything is ran within single transaction, which allows asynchronous testing with Postgres).
  2. There is documentation of Ecto.Adapters.SQL.Sandbox.
  3. Yes, Ecto checks out new connection from pool for each insert unless you will do it on your own via Repo.transaction or Repo.checkout
  4. It depends on the task. You can even test it synchronously if you want. There is no “one-fit-them-all” solution.

Thank you!