nti.testing 2.2.1

Support for testing code

nti.testing

Support for writing tests, particularly in a Zope3/ZTK environment, using zope.testing (nose2 may also work, but is not recommended).

Complete documentation is hosted at https://ntitesting.readthedocs.io/

Installation

nti.testing can be installed using pip, either from the git repository or from PyPI:

pip install nti.testing

PyHamcrest

nti.testing provides a group of PyHamcrest matchers. There are both general-purpose matchers and matchers that are of use to users of zope.interface and zope.schema.

Matchers can be imported from the nti.testing.matchers module.

Basic Matchers

is_true and is_false check the bool value of a supplied object (we’re using literals for explanation purposes, but it obviously makes more sense, and reads better, when the matched object is a variable, often of a more complex type):

>>> from hamcrest import assert_that, is_
>>> from nti.testing.matchers import is_true, is_false
>>> assert_that("Hi", is_true())
>>> assert_that(0, is_false())

Interface Matchers

Next we come to matchers that support basic use of zope.interface.

We can check that an object provides an interface and that a factory implements it:

>>> from zope.interface import Interface, Attribute, implementer
>>> class IThing1(Interface):
...     pass
>>> class IThing2(Interface):
...     pass
>>> class IThings(IThing1, IThing2):
...     got_that_thing_i_sent_you = Attribute("Did you get that thing?")
>>> @implementer(IThings)
... class Thing(object):
...     pass

>>> from nti.testing.matchers import provides, implements
>>> assert_that(Thing(), provides(IThings))
>>> assert_that(Thing, implements(IThings))

The attentive reader will have noticed that IThings defines an attribute that our implementation doesn’t actually provide. This is where the next stricter check comes in. verifiably_provides uses the interface machinery to determine that all attributes and methods specified by the interface are present as described:

>>> from nti.testing.matchers import verifiably_provides
>>> assert_that(Thing(), verifiably_provides(IThing2, IThing1))
>>> assert_that(Thing(), verifiably_provides(IThings))
Traceback (most recent call last):
...
AssertionError:...
Expected: object verifiably providing IThings
     but: <class 'Thing'> failed to provide attribute "got_that_thing_i_sent_you" from IThings
<BLANKLINE>

zope.interface can only check whether or not an attribute or method is present. To place (arbitrary) tighter constraints on the values of the attributes, we can step up to zope.schema and the validly_provides matcher:

>>> from zope.schema import Bool
>>> class IBoolThings(IThing1, IThing2):
...     got_that_thing_i_sent_you = Bool()
>>> @implementer(IBoolThings)
... class BoolThing(object):
...     pass

validly_provides is a superset of verifiably_provides:

>>> from nti.testing.matchers import validly_provides
>>> assert_that(BoolThing(), validly_provides(IThing1, IThing2))
>>> assert_that(BoolThing(), validly_provides(IBoolThings))
Traceback (most recent call last):
...
AssertionError:...
Expected: (object verifiably providing IBoolThings and object validly providing <InterfaceClass ....IBoolThings>)
     but: object verifiably providing IBoolThings <class 'BoolThing'> failed to provide attribute "got_that_thing_i_sent_you" from IBoolThings
<BLANKLINE>

For finer grained control, we can compare data against schema fields:

>>> from nti.testing.matchers import validated_by, not_validated_by
>>> field = IBoolThings.get('got_that_thing_i_sent_you')
>>> assert_that(True, is_(validated_by(field)))
>>> assert_that(None, is_(not_validated_by(field)))

Parent/Child Relationships

The aq_inContextOf matcher uses the concepts from Acquisition to check parent/child relationships:

>>> from nti.testing.matchers import aq_inContextOf
>>> class Parent(object):
...     pass
>>> class Child(object):
...     __parent__ = None
>>> parent = Parent()
>>> child = Child()
>>> child.__parent__ = parent

>>> assert_that(child, aq_inContextOf(parent))

Test Fixtures

Support for test fixtures can be found in nti.testing.base and nti.testing.layers. The base package includes fully-fleshed out base classes for direct use, while the layers package includes mixins that can be used to construct your own test layers.

The base package makes a distinction between “normal” and “shared” fixtures. Normal fixtures are those that are used for a single test case. They are established via setUp and torn down via tearDown.

In contrast, shared fixtures are expected to endure for the duration of all the tests in the class or all the tests in the layer. These are best used when the fixture is expensive to create. Anything that extends from base.AbstractSharedTestBase creates a shared fixture. Through the magic of metaclasses, such a subclass can also be assigned as the layer property of another class to be used as a test layer that can be shared across more than one class.

The most important bases are base.ConfiguringTestBase and base.SharedConfiguringTestBase. These are both fixtures for configuring ZCML, either from existing packages or complete file paths. To use these, subclass them and define class attributes set_up_packages and (if necessary) features:

>>> from nti.testing.base import ConfiguringTestBase
>>> import zope.security
>>> class MyConfiguringTest(ConfiguringTestBase):
...     set_up_packages = (
...         'zope.component', # the default configuration by name
...          # a named file in a named package
...          ('ftesting.zcml', 'zope.traversing.tests'),
...          # an imported module
...          zope.security,
...          # Our own package; in a test, this will mean the parent
...          # package
...          ".")

We would then proceed to write our test methods. The packages that we specified will be set up and torn down around every test method. In addition, the zope.testing cleanup functions will also run around every test method.

Time

Having a clock that’s guaranteed to move in a positive increasing way in every call to time.time is useful. nti.testing.time provides a decorator to accomplish this that ensures values always are at least the current time and always are increasing. (It is not thread safe.) It can be applied to functions or methods, and optionally takes a granularity argument:

>>> from nti.testing.time import time_monotonically_increases
>>> from nti.testing.time import reset_monotonic_time
>>> @time_monotonically_increases(0.1) # increment by 0.1
... def test():
...     import time
...     t1 = time.time()
...     t2 = time.time()
...     assert t2 == t1 + 0.1, (t2, t1)

>>> test()

Changes

2.2.1 (2019-09-10)

• Make transaction cleanup safer if the default transaction manager has been made explicit.

  Also, reset the default transaction manager to implicit.

  See issue 17.

2.2.0 (2018-08-24)

• Add support for Python 3.7.

• Make time_monotonically_increases also handle time.gmtime and add a helper for using it in layers.

2.1.0 (2017-10-23)

• Make Acquisition an optional dependency. If it is not installed, the aq_inContextOf matcher will always return False.

• Remove dependency on fudge. Instead, we now use unittest.mock on Python 3, or its backport mock on Python 2. See issue 11.

• Refactor ZCML configuration support to share more code and documentation. See issue 10.

• The layer ConfiguringLayerMixin and the base class SharedConfiguringTestBase now default to running configuration in the package the subclass is defined in, just as subclasses of ConfiguringTestBase did.

2.0.1 (2017-10-18)

• The validation matchers (validated_by and not_validated_by) now consider it a failure (by default) if the validate method raises anything other than zope.interface.exceptions.Invalid (which includes the zope.schema exceptions like WrongType). Previously, they accepted any exception as meaning the object was invalid, but this could hide bugs in the actual validation method itself. You can supply the invalid argument to the matchers to loosen or tighten this as desired. (Giving invalid=Exception will restore the old behaviour.) See issue 7.

2.0.0 (2017-04-12)

• Add support for Python 3.6.

• Remove unicode_literals.

• Substantially rework time_monotonically_increases for greater safety. Fixes issue 5.

1.0.0 (2016-07-28)

• Add Python 3 support.

• Initial PyPI release.