As you can see here I tried to propose a fix to the issue you talked about.
Let me know what do you think.
The strange thing I noticed is how the maintainer of
ja_resource decided to test the library. He or she introduced a fake implementation of Ecto.Schema and Ecto.Repo, this makes the test to be potentially “false positive”, cause when we will try to use a newer version of Ecto, the tests will continue to use the fake implementation and obviously continue to pass, with no feedback about possible regressions.
Another problem I found is, since we have to use the fake implementation for our tests we are forced to maintain and update the code twice (our application code, and the fake implementation) .
In general, it is not recommended to fake what you cannot control. In this case I would propose a new issue there to invite to remove the fake implementation and to use the real Ecto. Just because this library is strictly coupled with Phoenix/Ecto.
What do you think?