Coverage increased (+0.002%) to 99.433% when pulling 8ec46da on RockBomber:master into 4e8b057 on ActivisionGameScience:master.

@RockBomber Thanks for the PR. I'm not sure fail() should be allowed in soft assertions. I guess I was thinking that if you call fail() in a test, then you really want the test to fail. Can you share a use case or a testing workflow where it makes sense to fail soft?

Hello, @saturnboy !
Example of my case:

from assertpy import assert_that, fail, soft_assertions

def case_1(test_data_1):
    if test_data_1.field_1 == '1':
        assert_that(test_data_1.field_2).is_equal_to('1')
    elif test_data_1.field_1 == '2':
        assert_that(test_data_1.field_2).is_equal_to('2')
    else:
        fail('Unexpected value of field_1: %s' % test_data_1.field_1)

def case_2(test_data_2):
    if test_data_2.field_1 == '3':
        assert_that(test_data_2.field_2).is_equal_to('3')
    elif test_data_2.field_1 == '4':
        assert_that(test_data_2.field_2).is_equal_to('4')
    else:
        fail('Unexpected value of field_1: %s' % test_data_2.field_1)

def test_my_cases(test_data_1, test_data_2):
    with soft_assertions():
        case_1(test_data_1)
        case_2(test_data_2)

Of course, I can add first additional assertion like
assert_that(test_data_1.field_1).is_in('1','2')
But if I will need add another acceptable value of field_1, I must remember to add it to both places.

@RockBomber It looks like you are testing dependent values (like an if-and-only-if in a math proof), for example: iff a=1, then b=2. The design of fail() in unit testing is to force a test failure, still not convinced it should warn when inside the soft assertions context.

I'd encourage you think of one branch as the default, like this:

def case_1(test_data_1):
    if test_data_1.field_1 == '1':
        assert_that(test_data_1.field_2).is_equal_to('1')
    else: # default case
        assert_that(test_data_1.field_1).is_equal_to('2')
        assert_that(test_data_1.field_2).is_equal_to('2')

Ok, My PR violates design of original fail().
I can suggest new function sfail() instead:

def sfail(msg=''):
    global _soft_ctx
    if _soft_ctx:
        global _soft_err
        out = 'Fail: %s!' % msg if msg else 'Fail!'
        _soft_err.append(out)
        return
    fail(msg)

@RockBomber found a bug in the contextmanager, so thanks for that. Added some tests (see ca90355) in test_soft.py, so you can see the exact intended behavior of fail() in assertpy. Closing this PR.

Please open a new PR for soft_fail() if you still want that functionality. Please include some tests.