python-pytest-cases

• lazy_value parameters are now cached by pytest node id only. So plugins can access the value without triggering an extra function call, but a new call is triggered for each pytest node, so as to prevent mutable object leakage across tests. Fixed #149 while ensuring no regression for #143.

• The ids argument of parametrize now accepts a (possibly infinite) generator of ids, e.g. (f"foo{i}" for i in itertools.count()), just as pytest does. This was not always the case, inparticular when parametrizing a @fixture. The ids arguments of fixture_union, param_fixture[s], etc. now also support this pattern. Fixed #148

See documentation page for details.

• Improved description for the glob argument in @parametrize_with_cases. Also made the implementation escape all regex special characters so that they can't be used. Finally a pattern should now match the entire case id (previously, a partial match would work if it was at the beginning of the string). One step towards #147

See documentation page for details.

• is_lazy is now part of public API, and _LazyValue now has a cache mechanism like _LazyTuple. Fixes #143

• @parametrize: custom ids are now correctly taken into account when a single lazy_valueis used for a tuple of parameters. This issue could be seen also with @parametrize_with_cases: idgen does not seem to be taken into account when cases are unpacked into a tuple. Fixes #144.

• Empty case ids are now replaced with '<empty_case_id>' to avoid ambiguous interpretation of test ids. Fixes #142.

See documentation page for details.

• new clone(self, remove_int_base=False) API on LazyValue and LazyTupleItem instances. With this new API, on old pytest < 5.3, other plugins such as pytest-harvest can easily clone the contents from lazy values without having them inherit from int - which was a dirty hack used by pytest-cases to trick pytest to generate acceptable test ids in these old pytest versions. Also improved the LazyValue, LazyTuple and LazyTupleItem object model with equality and repr. Fixes pytest-harvest#43

See documentation page for details.

• Marks set on a case class are now propagated to cases in the class. So you can use for example pytest-pilot more easily ! Fixes #139

See documentation page for details.

• Fixed "Created fixture names are not unique, please report" error when duplicate fixture reference is provided in a pytest.param. Fixes #138.

See documentation page for details.

• Fixed TypeError: _idval() got an unexpected keyword argument 'item' with pytest versions between 3.0.0 and 3.7.4. Fixed #136

See documentation page for details.

• @parametrize_with_cases now supports that argnames is a list or tuple, just as @pytest.mark.parametrize does. PR #132, by @saroad2.

See documentation page for details.

• Now enforcing usage of makefun 1.9.3 or above to avoid issue AttributeError: 'functools.partial' object has no attribute '__module__' mentioned in #128

See documentation page for details.

• Improved documentation to explain why @fixture should be used instead of @pytest.fixture. Fixed #125

• Fixed ValueError: fixture is being applied more than once to the same function when two functions parametrized with the same cases were sitting in the same file. Improved robustness when cases require fixtures, in particular when parametrized test/fixture sits in a class or when several of them sit in a class/module. Fixed #126

See documentation page for details.

• Added missing deprecation warning on @cases_generator. Fixes #124.

• Removed target and tags arguments of @cases_generator (deprecated api anyway) that were added by mistake in version 2.0.0 but never used.

See documentation page for details.

• Added support for pytest items without funcargs. Fixes interoperability with other pytest plugins such as pytest-black or pytest-flake8. Fixes #122

See documentation page for details.

Fixed issue with @parametrize_with_cases when two cases with the same id and both requiring a fixture were to be created. Fixed #117.

Fixture closure engine refactoring:

• When no fixture unions are present, the fixture closure is now identical to the default one in pytest, to avoid issues originating from other plugins fiddling with the closure. Fixes #116

• New SuperClosure class representing the "list" facade on top of the fixture tree (instead of FixtureClosureNode). In addition, this list facade now better handles editing the order of fixtures when possible. Fixes #111.

• Session and Module-scoped fixtures that are not used in all union alternatives are not any more torn town/setup across union alternatives. Fixes #120

See documentation page for details.

• Fixed TypeError with iterable argvalue in standard parametrize. Fixed #115.

See documentation page for details.

• Fixed wrong module string decomposition when passed to cases argument in @parametrize_with_cases. Fixes #113

• Autouse fixtures are now correctly used. Fixed #114

See documentation page for details.

Lazy values (so, test cases) now have a much nicer string representation ; in particular in pytest-harvest results tables. Fixes #112

See documentation page for details.

• New documentation page concerning theory of fixture unions. Fixes #109

• Using a fixture_ref in a new-style @parametrize (with **args or idgen) now outputs a correct id. Fixes #110

See documentation page for details.

I am very pleased to announce this new version of pytest-cases, providing a lot of major improvements. Creating powerful and complex test suites have never been so easy and intuitive !

Below is a complete list of changes, but the user guide has also been updated accordingly so feel free to have a look to get a complete example-based walkthrough.

A/ More powerful and flexible cases collection

New @parametrize_with_cases decorator to replace @cases_data (deprecated).

1. Aligned with pytest:

   • now argnames can contain several names, and the case functions are automatically unpacked into it. You don't need to perform a case.get() in the test anymore !

       @parametrize_with_cases("a,b")
       def test_foo(a, b):
           # use a and b directly !
           ...
     

   • cases are unpacked at test setup time, so the clock does not run while the case is created - in case you use pytest-harvest to collect the timings.

   • @parametrize_with_cases can be used on test functions as well as fixture functions (it was already the case in v1)

2. Easier to configure:

   • the decorator now has a single cases argument to indicate the cases, wherever they come from (no module argument anymore)

   • default (cases=AUTO) automatically looks for cases in the associated case module named test_xxx_cases.py. Users can easily switch to alternate pattern cases_xxx.py with cases=AUTO2. Fixes #91.

   • cases can sit inside a class, like what you're used to do with pytest. This additional style makes it much more convenient to organize cases and associated them with tests, when cases sit in the same file than the tests. Fixes #93.

   • an explicit sequence can be provided, it can mix all kind of sources: functions, classes, modules, and module names as strings (even relative ones!).

      @parametrize_with_cases("a", cases=(CasesClass, '.my_extra_cases'))
      def test_foo(a):
          ...
     

3. More powerful API for filtering:

   • a new prefix argument (default case_) can be used to define case functions for various type of parameters: welcome user_<id>, data_<id>, algo_<id>, model_<id> ! Fixes #108

   • a new glob argument receiving a glob-like string can be used to further filter cases based on their names. For example you can distinguish *_success from *_failure case ids, so as to dispatch them to the appropriate positive or negative test. Fixes #108

   • finally you can still use has_tag and/or provide a filter callable, but now the callable will receive the case function, and this case function has a f._pytestcase attribute containing the id, tags and marks - it is therefore much easier to implement custom filtering.

B/ Easier-to-define case functions

• Case functions can start with different prefixes to denote different kind of data: e.g. data_<id>, user_<id>, model_<id>, etc.

• Case functions can now be parametrized with @parametrize or @pytest.mark.parametrize, just as in pytest ! This includes the ability to put pytest marks on the whole case, or on some specific parameter values using pytest.param. @cases_generator is therefore now deprecated but its alternate style for ids and arguments definition was preserved in @parametrize, see below.

• Now case functions can require fixtures ! In that case they will be transformed into fixtures and injected as fixture_ref in the parametrization. Fixes #56.

• New single optional @case(id=None, tags=(), marks=()) decorator to replace @case_name and @case_tags (deprecated): a single simple way to customize all aspects of a case function. Also, @test_target completely disappears from the picture as it was just a tag like others - this could be misleading.

C/ Misc / pytest goodies

• New aliases for readability: @fixture for @fixture_plus, and@parametrize for @parametrize_plus (both aliases will coexist with the old names). Fixes #107.

• @parametrize was improved in order to support the alternate parametrization mode that was previously offered by @cases_generator, see api reference. That way, users will be able to choose the style of their choice. Fixes #57 and #106.

• @parametrize now raises an explicit error message when the user makes a mistake with the argnames. Fixes #105.

• More readable error messages in @parametrize when lazy_value does not return the same number of argvalues than expected from the argnames.

• Any error message associated to a lazy_value function call is not caught and hidden anymore but is emitted to the user, for easier debugging.

• Fixed issue with lazy_value when a single mark is passed in the constructor.

• lazy_value used as a tuple for several arguments now have a correct id generated even in old pytest version 2.

• New pytest goodie assert_exception that can be used as a context manager. Fixes #104.

See documentation page for details.

• lazy_value are now resolved at pytest setup stage, not pytest call stage. This is important for execution time recorded in the reports (see also pytest-harvest plugin). Fixes #102

• A function used as a lazy_value can now be marked with pytest marks. Fixes #99

• A lazy_value now has a nicer id when it is a partial. Fixes #97

• Removed annoying PytestUnknownMarkWarning warning message when a mark was used on a case. Fixes #100

See documentation page for details.

• New marker lazy_value for parametrize_plus. Fixes #92

See documentation page for details.