Provides unittest assertions in a fluent style. Translated from the Java implementation, google/truth.

PyTruth is licensed under the Apache 2.0 license.

PyTruth is not an official Google product.

Please see the guidelines for contributing before creating pull requests.

PyTruth is not an actively maintained project. No support is provided.

It is shared with the community to bring an expressive, consistent assertion style to projects that may be using a combination of unittest, abseil, googletest, mox, and mock—especially to people familiar with Java Truth.

User group: [email protected]

PyTruth can be installed using pip:

pip install pytruth

Import the truth module and alias the AssertThat() method to begin asserting things:

from truth.truth import AssertThat

Then, instead of writing

self.assertEqual(a, b)
self.assertTrue(c)
self.assertIn(a, d)
self.assertTrue(a in d and b in d)
self.assertTrue(a in d or b in d or c in d)
with self.assertRaises(Error):
  Explode()

one would write

AssertThat(a).IsEqualTo(b)
AssertThat(c).IsTrue()
AssertThat(d).Contains(a)
AssertThat(d).ContainsAllOf(a, b)
AssertThat(d).ContainsAnyOf(a, b, c)
with AssertThat(Error).IsRaised():
  Explode()

Tests should be easier to read and write, and flow more clearly.

unittest assertions accept a msg parameter to display if the assertion fails. PyTruth has no such mechanism, though its failure messages tend to be more informative.

The type of the subject under test (the parameter passed to AssertThat()) will not be known until runtime, unlike Java where the type is known at compile time. IDEs may not correctly autocomplete available predicates on an asserted subject.

In Python 2, None compares less than every other thing, except None itself. None is less than nan, and it is less than negative infinity. Therefore, use caution when a function might return None. The assertion AssertThat(Func()).IsLessThan(0) succeeds whether Func() returns a negative number or None. Instead, first check the None-ness of the return value with IsNone() or IsNotNone() before performing an inequality assertion.

In Python 3, None is no longer comparable using < > <= >=. PyTruth detects the version of the Python interpreter and compares or fails appropriately, rather than allowing Python 3's TypeError to bubble up.

If the iterator over a shared value (either expected or actual) changes that value or its underlying elements, the behavior is undefined: all, none, or some of the assertions may succeed or fail, arbitrarily.

This library is threadsafe; you may execute multiple assertions in parallel.

PyTruth supports a finer grained distinction of truthiness than unittest does. In particular, it differentiates between "is True" and "is truthy." unittest's assertTrue(x) is equivalent to assertIs(bool(x), True), and its assertFalse(x) is equivalent to assertIs(bool(x), False). PyTruth's IsTrue() and IsFalse() predicates match only the boolean subjects True and False themselves. Therefore it is important not to blindly convert assertTrue() to IsTrue(), and likewise with assertFalse() and IsFalse().

The r parameter passed to the matching functions may either be a r'raw string', or a pattern object returned from re.compile().

The cf parameter passed to IsOrderedAccordingTo() and IsStrictlyOrderedAccordingTo() should be a callable that follows the contract of cmp(x, y): it should return negative if x < y, zero if x == y, and positive if x > y.

Ordered means that the iterable's elements must increase (or decrease, depending on cf) from beginning to end. Adjacent elements are allowed to be equal. Strictly ordered means that in addition, the elements must be unique (i.e., monotonically increasing or decreasing).

IsOrdered() is equivalent to IsOrderedAccordingTo(cmp).

IsStrictlyOrdered() is equivalent to IsStrictlyOrderedAccordingTo(cmp).

By default, ContainsAll... and ContainsExactly... do not enforce that the order of the elements in the subject under test matches the that of the expected value. To do that, append InOrder() to the returned predicate.

When using InOrder(), ensure that both the subject under test and the expected value have a predictable order, otherwise the result is undefined. For example, AssertThat(list_a).ContainsExactlyElementsIn(set_a).InOrder() may or may not succeed, depending on how the set implements ordering.

When expecting an exception using the AssertThat(e).IsRaised() context, any exception raised whose type is equal to e or a subclass of e is accepted. If an exception is raised that is not a subclass of e, the assertion fails.

The e parameter in the AssertThat(e).IsRaised() context may be either an exception class like ValueError, or it may be an exception instance like ValueError('some error'). If an instance is passed, then any exception raised by the code under test must also have matching message and args properties, in addition to being a subclass of the expected exception.

The WasCalled().Once().With(...) and WasCalled().With(...).Once() assertions are subtly different. WasCalled().Once().With(...) asserts that the function was called one time ever, and that one time it was called, it was passed those arguments. WasCalled().With(...).Once() asserts that the function was passed those arguments exactly once, but it is permitted to have been called with other, irrelevant arguments. Thus, WasCalled().Once().With(...) is the stricter assertion. Consider using HasExactlyCalls() for more clarity.