vutils-testing 1.0.2

Auxiliary library for writing tests

vutils-testing: Auxiliary Library for Writing Tests

This package provides a set of tools that help with writing tests. It helps with creating test data and types, mocking objects, patching, and verifying test results.

Installation

To install the package, type

$ pip install vutils-testing

How to Use

For more details, please follow the subsections below.

Type Factories

Sometimes tests require new types to be defined. To do this, vutils.testing.utils provides make_type function, which is a wrapper of type:

# Create type derived directly from object:
my_type = make_type("MyType")

# Create class derived directly from Exception:
my_error = make_type("MyError", Exception)

# Create class derived from A and B:
my_class = make_type("MyClass", (A, B))

# Create class derived from A with foo member:
my_another_class = make_type("MyAnotherClass", A, {"foo": 42})

# Create class derived from object with foo member:
my_test_class = make_type("MyTestClass", members={"foo": 42})

# Key-value arguments other than bases and members are passed to
# __init_subclass__:
my_fourth_class = make_type("MyFourthClass", bases=A, foo=42)

Mocking Objects and Patching

make_mock, make_callable, and PatcherFactory from vutils.testing.mock allow to create mock objects and patching things.

make_mock(*args, **kwargs) is a shortcut for unittest.mock.Mock

make_callable(x) creates also instance of unittest.mock.Mock, but it specifies its function-related behavior: if x is callable, it is used to do a side-effect, otherwise it is used as the return value.

# func_a() returns 3
func_a = make_callable(3)

container = []

# func_b() appends 42 to container
func_b = make_callable(lambda *x, **y: container.append(42))

# func_c() returns func_b
func_c = make_callable(lambda *x, **y: func_b)

PatcherFactory allows to use unittest.mock.patch multiple-times without need of nested with statements. When instantiated, setup method is called. setup method, implemented in the subclass, then may define set of patcher specifications via add_spec method:

class MyPatcher(PatcherFactory):

    @staticmethod
    def setup_foo(mock):
        mock.foo = "foo"

    @staticmethod
    def setup_baz(baz):
        baz["quux"] = 42

    def setup(self):
        self.baz = {}
        # When self.patch() is called:
        # - create a mock object, apply setup_foo on it, and patch foopkg.foo
        #   with it:
        self.add_spec("foopkg.foo", self.setup_foo)
        # - patch foopkg.bar with 42:
        self.add_spec("foopkg.bar", new=42)
        # - apply setup_baz on baz and patch foopkg.baz with it (create=True
        #   and other possible key-value arguments are passed to
        #   unittest.mock.patch):
        self.add_spec("foopkg.baz", self.setup_baz, new=self.baz, create=True)

patcher = MyPatcher()

with patcher.patch():
   # Patches are applied in order as specified by add_spec and reverted in
   # reverse order.
   do_something()

Covering mypy Specific Code

When a module contains code that is visible only to mypy, it is not executed during unit tests and hence reported as not covered. Function cover_typing from vutils.testing.utils module has the ability to execute such a code and therefore improve coverage reports:

# In foopkg/foo.py module:
if typing.TYPE_CHECKING:
    from foopkg import _A, _B, _C

# In test_coverage.py:
import pytest

from vutils.testing.utils import cover_typing

# Ensure the test run as last (this feature is available after installing
# pytest-order). cover_typing reloads the module which may have negative
# consequences on other tests
@pytest.mark.order("last")
def test_typing_code_is_covered():
    # When called, following happens:
    # - typing.TYPE_CHECKING is patched to True
    # - foopkg is patched with _A, _B, and _C symbols if they do not exist
    # - finally, foopkg.foo is reloaded
    cover_typing("foopkg.foo", ["_A", "_B", "_C"])

The story behind cover_typing is to keep source files clean from directives telling the pytest and linters what to do.

Sometimes a symbol can play two roles. Suppose that symbol _L is a type alias for list[object] when mypy is performing its checks and list otherwise:

# In foopkg/foo.py module:
if typing.TYPE_CHECKING:
    from foopkg import _L
else:
    _L = list


class ListType(_L):
    pass

To cover this case, ClassLikeSymbol from vutils.testing.utils comes to help. In test_coverage.py, just define _L like

class _L(metaclass=ClassLikeSymbol):
    pass

and then pass it to cover_typing:

cover_typing("foopkg.foo", [_L])

Deferred Instance Initialization

Patching may take no effect if the patched object appears in constructor and this constructor is called outside of patcher context. LazyInstance from vutils.testing.utils can defer initialization up to the time of method call:

class StderrWriter:
    def __init__(self):
        self.stream = sys.stderr

    def write(self, text):
        self.stream.write(text)

class StderrPatcher(PatcherFactory):
    def setup(self):
        self.stream = io.StringIO
        self.add_spec("sys.stderr", new=self.stream)

class MyTestCase(TestCase):
    def test_deferred_initialization(self):
        writerA = StderrWriter()
        writerB = LazyInstance(StderrWriter).create()
        patcher = StderrPatcher()

        # Patch sys.stderr:
        with patcher.patch():
            # Write Hello! to standard error output:
            writerA.write("Hello!\n")
            # Write Hi! to StringIO instance:
            writerB.write("Hi!\n")

Deferred assertRaises

Sometimes there are callable objects with a very similar prototypes and behavior so they can be run and checked with one universal function. However, if one of them raises an exception under specific circumstances, this must be also handled by the universal function, which adds to its complexity. For this reason, vutils.testing.utils introduces AssertRaises class which wraps exception raising assertions:

class FooError(Exception):
    detail = "foo"

def func_a(obj):
    obj.foo = 42

def func_b(obj):
    func_a(obj)
    raise FooError()

def Foo(TestCase):
    def run_and_check(self, func):
        obj = make_mock()
        func(obj)
        self.assertEqual(obj.foo, 42)

    def test_func(self):
        wfunc_b = AssertRaises(self, func_b, FooError)

        self.run_and_check(func_a)
        # Catch and store FooError:
        self.run_and_check(wfunc_b)
        # Check the caught exception:
        self.assertEqual(wfunc_b.get_exception().detail, "foo")

Enhanced TestCase

Module vutils.testing.testcase provides TestCase which is a subclass of unittest.TestCase extended about these methods:

• assert_called_with - assert that the mock object has been called once with the specified arguments and then reset it:

  class ExampleTestCase(TestCase):
      def test_assert_called_with(self):
          mock = make_mock(["foo"])
  
          mock.foo(1, 2, bar=3)
          self.assert_called_with(mock, 1, 2, bar=3)
  
          mock.foo(4)
          self.assert_called_with(mock, 4)
  

• assert_not_called - assert that the mock object has not been called:

  class ExampleTestCase(TestCase):
      def test_assert_not_called(self):
          mock = make_mock(["foo"])
  
          self.assert_not_called(mock.foo)