chensan
Why Phoenix Recycling?
I’m writing tests for my controller:
defp log_in_user(%{conn: conn, log_in_as: email}) do
role = role_fixture()
user = user_fixture(role, email: email)
conn = assign(conn, :current_user, user)
{:ok, conn: conn, user: user}
end
defp create_site(%{conn: conn, user: user}) do
site = site_fixture(user)
{:ok, site: site}
end
describe "delete site" do
setup [:log_in_user, :create_site]
@tag log_in_as: "sam"
test "deletes chosen site", %{conn: conn, site: site} do
conn = delete(conn, Routes.site_path(conn, :delete, site))
assert redirected_to(conn) == Routes.site_path(conn, :index)
assert_error_sent 404, fn ->
get(conn, Routes.site_path(conn, :show, site))
end
end
end
The weird thing here is conn.assigns.current_user becomes nil in assert_error_sent where it should have value.
I found some similar questions around the web:
- Troubleshooting a failed test 302 redirect instead of 200 - #3 by chensan
- https://stackoverflow.com/questions/50110449/phoenix-controller-test-case-loses-current-user
- https://stackoverflow.com/questions/46363292/losing-conn-assigns-in-the-middle-of-a-test
Seems the problem is caused by Phoenix.ConnTest – Phoenix v1.4.0, and there’s a fix:
saved_assigns = conn.assigns
conn =
conn
|> recycle()
|> Map.put(:assigns, saved_assigns)
I don’t quite understand why, it’s so unexpected in our tests. Anyone could help explain it?
Marked As Solved
LostKobrakai
recycle does essentially what happens if you start a new request in the browser. Any new request in the browser will start of without any assigns. Usually you’re kept logged in by putting something in the session for the user.
Also Liked
idi527
Maybe Plug.Test — Plug v1.20.2
setup %{conn: conn} do
{:ok, conn: Plug.Test.init_test_session(conn, user_id: 666)}
end
I wouldn’t change assigns manually, instead I’d pipe the requests through the actual plug pipelines with the session set via Plug.Test.init_test_session/2 to get the expected conn with all relevant assigns set.
LostKobrakai
LostKobrakai
That’s just not true. If you’re not using session based authentication you’d need to log in new for each request. The only data which survives between requests is the stuff you put into the session. Anything else is discarded at the end of a request. That’s true for the request you do with a browser in dev mode and exactly what recycle mimics for tests.
Last Post!
chensan
After reading the source code, I’d like to share what I learned so far.
When we run get(conn, Routes.site_path(conn, :show, site)), a new conn was created, and cookies and some request headers (~w(accept authorization)) from old conn are copied to new conn - that’s what Phoenix recycling does for us, it emulates the behavior of browser. But conn.assigns is not copied, results in a nil value of current_user.
The reason post(conn, Routes.session_path(conn, :create), session: %{email: user.email, password: "123456"} ) works is because Auth module selects another path:
def call(conn, _opts) do
user_id = get_session(conn, :user_id)
cond do
conn.assigns[:current_user] -> # <- the way programming phoenix book suggests
conn
user = user_id && Accounts.get_user(user_id) -> # <- post Routes.session_path just selects this way.
assign(conn, :current_user, user)
true ->
assign(conn, :current_user, nil)
end
end
Thanks all for your kind help ![]()
Popular in Questions
Other popular topics
Categories:
Sub Categories:
Forums
Popular Tags
- #ecto
- #liveview
- #troubleshooting
- #learning-elixir
- #deployment
- #library
- #erlang
- #testing
- #genserver
- #mix
- #absinthe
- #remote-other
- #otp
- #plug
- #how-to-question
- #macros
- #postgres
- #channels
- #elixirconf
- #exunit
- #discussion
- #code-sync
- #javascript
- #podcasts
- #onsite
- #dialyzer
- #docker
- #authentication
- #umbrella
- #full-time-contract
- #podcasts-by-brainlid
- #ecto-query
- #elixir-ls
- #phoenix_html
- #iex
- #blog-post
- #graphql
- #genstage
- #ai
- #websockets
- #supervisor
- #elixirconf-us
- #advent-of-code
- #distillery
- #processes
- #api
- #forms
- #metaprogramming
- #security
- #hex









