I often use ExUnit.CaptureLog to capture information about a log message. But it doesn’t give you any information about the logged metadata. I’ve been adding structured logging recently and I’d like to assert that my code is correctly emitting the correct structured logs, but ExUnit.CaptureLog doesn’t keep or return the metadata. How have others solved this? Maybe we could create a PR to ex_unit that would add a function that returns the metadata along with the captured message?
At first glance, seems like the tricky part would be getting the results back to what’s calling capture_log
- the current code uses a StringIO
after formatting the event:
Hi, Jason !
In one project I format logs in json and include metadata there in the message, so it’s pretty straight forward… (Jason.decode
the captured log and assert on internals).
However, I guess that’s not your case, so I think we could add :logger
handler and try to assert on the metadata within log_event
.
Since handlers run in the same process as the one that emits the log we could send the roughly something like this:log_event
to self()
and receive in test
defmodule LoggerTestHelper do
def log(log_event, config) do
send(config.test_pid, {:logged_event, log_event})
end
end
defmodule Test do
use ExUnit.Case, async: true
require Logger
setup do
:logger.add_handler(:test_handler, LoggerTestHelper, %{test_pid: self()})
on_exit(fn -> :logger.remove_handler(:test_handler) end)
end
test "metadata" do
Logger.info("hello", foo: :bar)
assert_receive {:logged_event, %{meta: meta} = _log_event}
assert meta.foo == :bar
end
end
UPD: I just realized, that event might be logged by a process other than the test itself, and sending log_event
to self()
is quite limiting… Hence, I updated the example passing test_pid
to logger handler config and having it to send log_event
to test_pid
explicitly.
Hi Rudolf!
Oooh, I didn’t expect adding a handler within the test to end up that straightforward and clean! I’ve marked your answer as the solution because it does indeed solve the problem, thank you!
I’ve also extended your example helper slightly to improve the ergonomics, mainly so that it’s easy to get all the logged messages that match a given string, and to return a formatted version of the string (although I’m sure that there’s other cases that would be needed besides {:string, iolist}
if this were to hit production:
defmodule LoggerTestHelper do
def log(log_event, config) do
message =
case log_event.msg do
{:string, iolist} -> to_string(iolist)
end
send(config.test_pid, {:logged_event, message, log_event})
end
def logs_containing_string(search_string) do
{:messages, all_messages} = :erlang.process_info(self(), :messages)
for {:logged_event, message, log_event} <- all_messages,
String.contains?(message, search_string) do
{message, log_event}
end
end
end
And here’s my example test:
test "logs with metadata", %{conn: conn} do
{_result, _log} =
ExUnit.CaptureLog.with_log(fn ->
conn = get(conn, ~p"/posts")
assert html_response(conn, 200) =~ "Listing Posts"
end)
assert [{message, log}] =
LoggerTestHelper.logs_containing_string("Sent 200 for GET /posts")
assert message =~ "2 queries executed in"
assert message =~ "2 logs captured"
assert log.meta.method == "GET"
assert log.meta.request_path == "/posts"
assert log.meta.duration
assert log.meta.total_queries == 2
for query <- log.meta.ecto_queries do
assert query =~ "FROM \"posts\""
end
end
If you’re curious about the content of the test I’ve been playing with the ideas of Wide Logs (or Canonical Logs) that are captured quite well in this Stripe blog post: