2.0.0

Highlights

First-class async/await and property/descriptor support. Chaining. InOrder.
Enhanced captor, new patch_attr and patch_dict.

• Deprecate verifyNoMoreInteractions in favor of ensureNoUnverifiedInteractions.

• Deprecate verifyNoUnwantedInteractions in favor of verifyExpectedInteractions.

• Context managers now check usage and any expectations (set via expect) on exit. The usage
  check can be disabled with the environment variable MOCKITO_CONTEXT_MANAGERS_CHECK_USAGE="0".

• The between matcher now supports open ranges, e.g. between=(0,) to assert that at least
  0 interactions occurred.

• Calling thenAnswer() without arguments is now allowed and is treated like
  thenReturn() without arguments: the stubbed method will return None.

• Allow ... in fixed argument positions as an ad-hoc any matcher.
  Trailing positional ... keeps its existing "rest" semantics.

• Expanded mock({...}) constructor shorthands. Refer https://mockito-python.readthedocs.io/en/latest/mock-shorthands.html

• Added a first-class InOrder API via mockito.InOrder. The legacy in-order mode only supported one mock at a time;
  the new API supports true cross-mock order verification.

  Migration (old limited style -> new style)::

  # Before (legacy, single-mock order only)
  from mockito import inorder
  inorder.verify(cat).meow()
  inorder.verify(cat).purr()
  
  # Now (preferred, explicit cross-mock order)
  from mockito import InOrder
  in_order = InOrder(cat, dog)
  in_order.verify(cat).meow()
  in_order.verify(dog).bark()
  

• The legacy in-order verification mode (inorder.verify(...))
  is deprecated in favor of InOrder(...).

• Added first-class async/await stubbing support: async callables now preserve
  awaitable behavior for thenReturn, thenRaise, and thenAnswer (including
  sync and async answer callables), with parity across when, when2,
  patch, and expect.
  Note that async introspection metadata (e.g. inspect.iscoroutinefunction)
  for stub wrappers is currently implemented only on Python 3.12+.

• Added first-class property/descriptor stubbing support, including class-level property
  stubbing via when(F).p.thenReturn(...) and thenCallOriginalImplementation() support for
  property stubs (including chained answers like
  thenReturn(...).thenCallOriginalImplementation()). Stubbing instance properties now fails
  fast with clear guidance to use class-level stubbing (when(F).p...).

• Added chained stubbing and expectations across call/property hops, e.g.
  when(cat).meow().purr().thenReturn(...), when(User).query.filter_by(...).first(), and
  expect(cat, times=1).meow().purr(), including cleanup that preserves sibling chain branches.

• Extend captor to be used as a rest matcher

  E.g., support

  args = captor()
  when(freud).says(*args).thenReturn("Yes.")
  assert freud.says("Are", "you", "dreaming?") == "Yes."
  assert args.value == ("Are", "you", "dreaming?")
  # or
  kwargs = captor()
  when(freud).does(**kwargs).thenReturn(False)
  

• Added call_captor

  call = call_captor()
  when(mock).do(call).thenReturn("ok")
  mock.do(1, 2, x=3)
  assert call.value == ((1, 2), {"x": 3})
  

• Added patch_attr and patch_dict for non-callable monkeypatch-style use cases
  (e.g. sys.stdout, sys.argv, and environment/config dictionaries) with
  context-manager support and restoration through unstub.

  E.g.

  patch_attr("sys.argv", ["foo", "bar"])
  with patch_attr("sys.stdout", StringIO()) as stdout: ...
  with patch_dict(os.environ, {"user": "bob"}): ...
  

• Added explicit partial-unstub targeting by host + attribute name.
  This complements method-reference partial unstub (e.g. unstub(cat.meow))
  and supports tuple form and shorthand form, including multiple attributes
  in one call.

  E.g.

  unstub(cat.meow)
  unstub((cat, "meow"))
  unstub(cat, "meow")
  unstub((cat, "meow"), (os.path, "exists"))
  

• Also implemented unstub("os.path")

2.0.0a6

• Extend captor to be used as a rest matcher

  E.g., support

  args = captor()
  when(freud).says(*args).thenReturn("Yes.")
  assert freud.says("Are", "you", "dreaming?") == "Yes."
  assert args.value == ("Are", "you", "dreaming?")
  # or
  kwargs = captor()
  when(freud).does(**kwargs).thenReturn(False)
  

• Added call_captor

  call = call_captor()
  when(mock).do(call).thenReturn("ok")
  mock.do(1, 2, x=3)
  assert call.value == ((1, 2), {"x": 3})
  

• Added patch_attr and patch_dict for non-callable monkeypatch-style use cases
  (e.g. sys.stdout, sys.argv, and environment/config dictionaries) with
  context-manager support and restoration through unstub.

  E.g.

  patch_attr("sys.argv", ["foo", "bar"])
  with patch_attr("sys.stdout", StringIO()) as stdout: ...
  with patch_dict(os.environ, {"user": "bob"}): ...
  

• Added explicit partial-unstub targeting by host + attribute name.
  This complements method-reference partial unstub (e.g. unstub(cat.meow))
  and supports tuple form and shorthand form, including multiple attributes
  in one call.

  E.g.

  unstub(cat.meow)
  unstub((cat, "meow"))
  unstub(cat, "meow")
  unstub((cat, "meow"), (os.path, "exists"))
  

• Also implemented unstub("os.path")

2.0.0a5

• Bug fixes and iteration on the chain feature now supporting deep property chaining as well.

• Rollback specificity ordering from a4. Turns out that this less predictable then the last one takes it all rule.

2.0.0a4

• Allow ... in fixed argument positions as an ad-hoc any matcher.
  Trailing positional ... keeps its existing "rest" semantics.

• Expanded mock({...}) constructor shorthands. Refer https://mockito-python.readthedocs.io/en/latest/mock-shorthands.html

• Added chained stubbing and expectations across call/property hops, e.g.
  when(cat).meow().purr().thenReturn(...), when(User).query.filter_by(...).first(), and
  expect(cat, times=1).meow().purr(), including cleanup that preserves sibling chain branches.

• BREAKING: Stubs for the same method are now sorted by specificity. #110 Rolled back in a5.

  Same-ish are for example

  when(os.path).exists(...).thenCallOriginalImplementation()
  when(os.path).exists('.flake8').thenReturn(False)
  

2.0.0a3

• Added first-class async/await stubbing support: async callables now preserve
  awaitable behavior for thenReturn, thenRaise, and thenAnswer (including
  sync and async answer callables), with parity across when, when2,
  patch, and expect.
  Note that async introspection metadata (e.g. inspect.iscoroutinefunction)
  for stub wrappers is currently implemented only on Python 3.12+.

• Added first-class property/descriptor stubbing support, including class-level property
  stubbing via when(F).p.thenReturn(...) and thenCallOriginalImplementation() support for
  property stubs (including chained answers like
  thenReturn(...).thenCallOriginalImplementation()). Stubbing instance properties now fails
  fast with clear guidance to use class-level stubbing (when(F).p...).

2.0.0a2

• Added a first-class InOrder API via mockito.InOrder (also available as
  mockito.inorder.InOrder). The legacy in-order mode only supported one mock at a time;
  the new API supports true cross-mock order verification.

  Migration (old limited style -> new style)::

  # Before (legacy, single-mock order only)
  from mockito import inorder
  inorder.verify(cat).meow()
  inorder.verify(cat).purr()
  
  # Now (preferred, explicit cross-mock order)
  from mockito import InOrder
  in_order = InOrder(cat, dog)
  in_order.verify(cat).meow()
  in_order.verify(dog).bark()
  

• The legacy in-order verification mode (inorder.verify(...))
  is deprecated in favor of InOrder(...).

2.0.0a1

• Calling thenAnswer() without arguments is now allowed and is treated like
  thenReturn() without arguments: the stubbed method will return None.
• Deprecate verifyNoMoreInteractions in favor of ensureNoUnverifiedInteractions.
• Deprecate verifyNoUnwantedInteractions in favor of verifyExpectedInteractions.
• Context managers now check usage and any expectations (set via expect) on exit. The usage
  check can be disabled with the environment variable MOCKITO_CONTEXT_MANAGERS_CHECK_USAGE="0".
• The between matcher now supports open ranges, e.g. between=(0,) to assert that at least
  0 interactions occurred.

1.5.4

Surprisingly deepcopy(mock()) returned None. Fix that! For dumb mocks, let deepcopy(m) return an actual deep copy of m. For strict mocks, let it throw if it is an unexpected usage.

You can quickly define mock({"__deepcopy__": None}) to expect the usage and get the default deepcopy algorithm. Note that expectations and recorded usages are shared by default even for copied mocks. You can fix that DIY, e.g when(copy).deepcopy(m).thenReturn(n). (Or the other way around: when(m).__deepcopy__(...).thenReturn(n).