Print test fixture description in erlang eunit failure

Question

Is there a way to print the test description of an erlang test generator that uses fixtures? Using a generator makes it tricky to tell what test is actually failing and printing the description would help out.

Example:

-module(math_test).

-include_lib("eunit/include/eunit.hrl").
-define(test(Desc, F), {Desc, {setup, fun setup/0, fun cleanup/1, F}}).

setup() ->
  ok.

cleanup(_) ->
  ok.

math_test_ () ->
  [
    ?test("adds two numbers", fun add_two_numbers/0),
    ?test("subtract two numbers", fun subtract_two_numbers/0),
    ?test("undefined method called", fun undefined_error/0)
  ].

add_two_numbers () ->
  ?assertEqual(2, 1 + 3).

subtract_two_numbers () ->
  ?assertEqual(1, 2 - 2).

undefined_error () ->
  undefined_module:uh_oh().

And then running it

[root@a7c901c022bb src]# rebar3 eunit --module=math_test
===> Verifying dependencies...
===> Compiling math
===> Performing EUnit tests...
FFF
Failures:

  1) math_test:math_test_/0
     Failure/Error: ?assertEqual(2, 1 + 3)
       expected: 2
            got: 4
     %% /src/_build/test/lib/math/src/math_test.erl:20:in `math_test:-add_two_numbers/0-fun-0-/1`
     Output: 
     Output: 
  2) math_test:math_test_/0
     Failure/Error: ?assertEqual(1, 2 - 2)
       expected: 1
            got: 0
     %% /src/_build/test/lib/math/src/math_test.erl:23:in `math_test:-subtract_two_numbers/0-fun-0-/1`
     Output: 
     Output: 
  3) math_test:math_test_/0
     Failure/Error: {error,undef,[{undefined_module,uh_oh,[],[]}]}
     Output: 

The first 2 errors are ok, but not great -- you can at least see in the assertion where things actually went wrong.

However the 3rd error (calling undefined module/method) is where things go horribly wrong -- there's no real way to tell where it came from!

Is there a way to improve things, like printing the test description with the failure log?

Answer 1

One thing you can do is putting the test description on the test itself, not the entire setup tuple. That is, change this line:

-define(test(Desc, F), {Desc, {setup, fun setup/0, fun cleanup/1, F}}).

to:

-define(test(Desc, F), {setup, fun setup/0, fun cleanup/1, {Desc, F}}).

With that change, the test descriptions are printed:

Failures:

  1) math_test:math_test_/0: adds two numbers
     Failure/Error: ?assertEqual(2, 1 + 3)
       expected: 2
            got: 4
     %% /tmp/math_test/mylib/_build/test/lib/mylib/test/math_test.erl:20:in `math_test:-add_two_numbers/0-fun-0-/0`
     Output: 
     Output: 
  2) math_test:math_test_/0: subtract two numbers
     Failure/Error: ?assertEqual(1, 2 - 2)
       expected: 1
            got: 0
     %% /tmp/math_test/mylib/_build/test/lib/mylib/test/math_test.erl:23:in `math_test:-subtract_two_numbers/0-fun-0-/0`
     Output: 
     Output: 
  3) math_test:math_test_/0: undefined method called
     Failure/Error: {error,undef,[{undefined_module,uh_oh,[],[]}]}
     Output: 

---

Another thing to try is specifying the test functions with the ?_test macro instead of as plain fun terms:

math_test_ () ->
  [
    ?test("adds two numbers", ?_test(add_two_numbers())),
    ?test("subtract two numbers", ?_test(subtract_two_numbers())),
    ?test("undefined method called", ?_test(undefined_error()))
  ].

The ?_test macro remembers the line number it appeared on, and includes it in the output on test failure:

  1) math_test:math_test_/0:14: adds two numbers
  [...]
  2) math_test:math_test_/0:15: subtract two numbers
  [...]
  3) math_test:math_test_/0:16: undefined method called
  [...]

Now you can tell which line those tests were invoked from.

---

Yet another way to do it is to have the individual functions return eunit "test objects" instead of just running the tests. That would involve using ?_assertEqual instead of ?assertEqual, or wrapping the entire thing in ?_test:

math_test_ () ->
  [
    ?test("adds two numbers", add_two_numbers()),
    ?test("subtract two numbers", subtract_two_numbers()),
    ?test("undefined method called", undefined_error())
  ].

add_two_numbers () ->
  ?_assertEqual(2, 1 + 3).

subtract_two_numbers () ->
  ?_assertEqual(1, 2 - 2).

undefined_error () ->
  ?_test(undefined_module:uh_oh())

Then the output contains both the line numbers and the names of the individual test functions:

Failures:

  1) math_test:add_two_numbers/0:20: adds two numbers
     Failure/Error: ?assertEqual(2, 1 + 3)
       expected: 2
            got: 4
     %% /tmp/math_test/mylib/_build/test/lib/mylib/test/math_test.erl:20:in `math_test:-add_two_numbers/0-fun-0-/0`
     Output: 
     Output: 
  2) math_test:subtract_two_numbers/0:23: subtract two numbers
     Failure/Error: ?assertEqual(1, 2 - 2)
       expected: 1
            got: 0
     %% /tmp/math_test/mylib/_build/test/lib/mylib/test/math_test.erl:23:in `math_test:-subtract_two_numbers/0-fun-0-/0`
     Output: 
     Output: 
  3) math_test:undefined_error/0:26: undefined method called
     Failure/Error: {error,undef,[{undefined_module,uh_oh,[],[]}]}
     Output: 

Answer 2

The answers by @legoscia are good, but I also suspect that it's the error reporting implemented by rebar3 that's suboptimal for that kind of error. Running the tests directly from eunit with its default output, you get this:

2> eunit:test(math_test). 
math_test: math_test_...*failed*
in function math_test:'-add_two_numbers/0-fun-0-'/0 (math_test.erl, line 22)
**error:{assertEqual,[{module,math_test},
                      {line,22},
                      {expression,"1 + 3"},
                      {expected,2},
                      {value,4}]}
  output:<<"">>

math_test: math_test_...*failed*
in function math_test:'-subtract_two_numbers/0-fun-0-'/0 (math_test.erl, line 25)
**error:{assertEqual,[{module,math_test},
                      {line,25},
                      {expression,"2 - 2"},
                      {expected,1},
                      {value,0}]}
  output:<<"">>

math_test: math_test_...*failed*
in function undefined_module:uh_oh/0
  called as uh_oh()
**error:undef 
  output:<<"">>

=======================================================
  Failed: 3.  Skipped: 0.  Passed: 0.

With the 'verbose' option, it also prints the descriptions before each setup. Furthermore, moving the description to the test fun, and using the ?_test(...) macro to create test funs with more positional information than a plain fun, as @legoscia suggested, you get this output:

  math_test:18: math_test_ (undefined method called)...*failed*
in function undefined_module:uh_oh/0
  called as uh_oh()
**error:undef 
  output:<<"">>

You could report this to the rebar3 maintainers.