Unit testing for Sage objects#

class sage.misc.sage_unittest.InstanceTester(instance, elements=None, verbose=False, prefix='', max_runs=4096, max_samples=None, **options)#

Bases: TestCase

A gadget attached to an instance providing it with testing utilities.

EXAMPLES:

sage: from sage.misc.sage_unittest import InstanceTester
sage: InstanceTester(instance = ZZ, verbose = True, elements = [1,2,3])
Testing utilities for Integer Ring

This is used by SageObject._tester, which see:

sage: QQ._tester()
Testing utilities for Rational Field
info(message, newline=True)#

Display user information

EXAMPLES:

sage: from sage.misc.sage_unittest import InstanceTester
sage: tester = InstanceTester(ZZ, verbose = True)

sage: tester.info("hello"); tester.info("world")
hello
world

sage: tester = InstanceTester(ZZ, verbose = False)
sage: tester.info("hello"); tester.info("world")

sage: tester = InstanceTester(ZZ, verbose = True)
sage: tester.info("hello", newline = False); tester.info(" world")
hello world
longMessage = False#
runTest()#

Trivial implementation of unittest.TestCase.runTest() to please the super class TestCase. That’s the price to pay for abusively inheriting from it.

EXAMPLES:

sage: from sage.misc.sage_unittest import InstanceTester
sage: tester = InstanceTester(ZZ, verbose = True)
sage: tester.runTest()
some_elements(S=None, repeat=None)#

Return a list (or iterable) of elements of the instance on which the tests should be run.

This is only meaningful for container objects like parents.

INPUT:

  • S – a set of elements to select from. By default this will use the elements passed to this tester at creation time, or the result of some_elements() if no elements were specified.

  • repeat – integer (default: None). If given, instead returns a list of tuples of length repeat from S.

OUTPUT:

A list of at most self._max_runs elements of S^r, or a sample of at most self._max_samples if that is not None.

EXAMPLES:

By default, this calls some_elements() on the instance:

sage: from sage.misc.sage_unittest import InstanceTester
sage: class MyParent(Parent):
....:     def some_elements(self):
....:         return [1,2,3,4,5]
...
sage: tester = InstanceTester(MyParent())
sage: list(tester.some_elements())
[1, 2, 3, 4, 5]

sage: tester = InstanceTester(MyParent(), max_runs=3)
sage: list(tester.some_elements())
[1, 2, 3]

sage: tester = InstanceTester(MyParent(), max_runs=7)
sage: list(tester.some_elements())
[1, 2, 3, 4, 5]

sage: tester = InstanceTester(MyParent(), elements=[1,3,5])
sage: list(tester.some_elements())
[1, 3, 5]

sage: tester = InstanceTester(MyParent(), elements=[1,3,5], max_runs=2)
sage: list(tester.some_elements())
[1, 3]

sage: tester = InstanceTester(FiniteEnumeratedSet(['a','b','c','d']), max_runs=3)
sage: tester.some_elements()
['a', 'b', 'c']

sage: tester = InstanceTester(FiniteEnumeratedSet([]))
sage: list(tester.some_elements())
[]

sage: tester = InstanceTester(ZZ)
sage: ZZ.some_elements()             # yikes, shamelessly trivial ...
<generator object ..._some_elements_from_iterator at 0x...>
sage: list(tester.some_elements())
[0, 1, -1, 2, -2, ..., 49, -49, 50]

sage: tester = InstanceTester(ZZ, elements = ZZ, max_runs=5)
sage: list(tester.some_elements())
[0, 1, -1, 2, -2]

sage: tester = InstanceTester(ZZ, elements = srange(100), max_runs=5)
sage: list(tester.some_elements())
[0, 1, 2, 3, 4]

sage: tester = InstanceTester(ZZ, elements = srange(3), max_runs=5)
sage: list(tester.some_elements())
[0, 1, 2]

The repeat keyword can give pairs or triples from S:

sage: list(tester.some_elements(repeat=2))
[(0, 0), (0, 1), (0, 2), (1, 0), (1, 1)]

You can use max_samples to sample at random, instead of in order:

sage: tester = InstanceTester(ZZ, elements = srange(8), max_samples = 4)
sage: all(t in srange(8) for t in tester.some_elements())
True
sage: all(s in srange(8) and t in srange(8) for s,t in tester.some_elements(repeat=2))
True

Test for github issue #15919, github issue #16244:

sage: Z = IntegerModRing(25) # random.sample, which was used pre #16244, has a threshold at 21!
sage: Z[1]                   # since #8389, indexed access is used for ring extensions
Traceback (most recent call last):
...
ValueError: variable name '1' does not start with a letter
sage: tester = InstanceTester(Z, elements=Z, max_runs=5)
sage: list(tester.some_elements())
[0, 1, 2, 3, 4]

sage: C = cartesian_product([Z]*4)
sage: len(C)
390625
sage: tester = InstanceTester(C, elements = C, max_runs=4)
sage: list(tester.some_elements())
[(0, 0, 0, 0), (0, 0, 0, 1), (0, 0, 0, 2), (0, 0, 0, 3)]
class sage.misc.sage_unittest.PythonObjectWithTests(instance)#

Bases: object

Utility class for running basis tests on a plain Python object (that is not in SageObject). More test methods can be added here.

EXAMPLES:

sage: TestSuite("bla").run()
class sage.misc.sage_unittest.TestSuite(instance)#

Bases: object

Test suites for Sage objects.

EXAMPLES:

sage: TestSuite(ZZ).run()

No output means that all tests passed. Which tests? In practice this calls all the methods ._test_* of this object, in alphabetic order:

sage: TestSuite(1).run(verbose = True)
running ._test_category() . . . pass
running ._test_eq() . . . pass
running ._test_new() . . . pass
running ._test_nonzero_equal() . . . pass
running ._test_not_implemented_methods() . . . pass
running ._test_pickling() . . . pass

Those methods are typically implemented by abstract super classes, in particular via categories, in order to enforce standard behavior and API, or provide mathematical sanity checks. For example if self is in the category of finite semigroups, this checks that the multiplication is associative (at least on some elements):

sage: S = FiniteSemigroups().example(alphabet = ('a', 'b'))
sage: TestSuite(S).run(verbose = True)
running ._test_an_element() . . . pass
running ._test_associativity() . . . pass
running ._test_cardinality() . . . pass
running ._test_category() . . . pass
running ._test_construction() . . . pass
running ._test_elements() . . .
  Running the test suite of self.an_element()
  running ._test_category() . . . pass
  running ._test_eq() . . . pass
  running ._test_new() . . . pass
  running ._test_not_implemented_methods() . . . pass
  running ._test_pickling() . . . pass
  pass
running ._test_elements_eq_reflexive() . . . pass
running ._test_elements_eq_symmetric() . . . pass
running ._test_elements_eq_transitive() . . . pass
running ._test_elements_neq() . . . pass
running ._test_enumerated_set_contains() . . . pass
running ._test_enumerated_set_iter_cardinality() . . . pass
running ._test_enumerated_set_iter_list() . . . pass
running ._test_eq() . . . pass
running ._test_new() . . . pass
running ._test_not_implemented_methods() . . . pass
running ._test_pickling() . . . pass
running ._test_some_elements() . . . pass

The different test methods can be called independently:

sage: S._test_associativity()

Debugging tip: in case of failure of some test, use %pdb on to turn on automatic debugging on error. Run the failing test independently: the debugger will stop right where the first assertion fails. Then, introspection can be used to analyse what exactly the problem is. See also the catch = False option to run().

When meaningful, one can further customize on which elements the tests are run. Here, we use it to prove that the multiplication is indeed associative, by running the test on all the elements:

sage: S._test_associativity(elements = S)

Adding a new test boils down to adding a new method in the class of the object or any super class (e.g. in a category). This method should use the utility _tester() to handle standard options and report test failures. See the code of _test_an_element() for an example. Note: Python’s testunit convention is to look for methods called .test*; we use instead ._test_* so as not to pollute the object’s interface.

Eventually, every implementation of a SageObject should run a TestSuite on one of its instances in its doctest (replacing the current loads(dumps(x)) tests).

Finally, running TestSuite on a standard Python object does some basic sanity checks:

sage: TestSuite(int(1)).run(verbose = True)
running ._test_new() . . . pass
running ._test_pickling() . . . pass

TODO:

  • Allow for customized behavior in case of failing assertion (warning, error, statistic accounting). This involves reimplementing the methods fail / failIf / … of unittest.TestCase in InstanceTester

  • Don’t catch the exceptions if TestSuite(..).run() is called under the debugger, or with %pdb on (how to detect this? see get_ipython(), IPython.Magic.shell.call_pdb, …) In the mean time, see the catch=False option.

  • Run the tests according to the inheritance order, from most generic to most specific, rather than alphabetically. Then, the first failure will be the most relevant, the others being usually consequences.

  • Improve integration with doctests (statistics on failing/passing tests)

  • Add proper support for nested testsuites.

  • Integration with unittest: Make TestSuite inherit from unittest.TestSuite? Make .run(...) accept a result object

  • Add some standard option proof = True, asking for the test method to choose appropriately the elements so as to prove the desired property. The test method may assume that a parent implements properly all the super categories. For example, the _test_commutative method of the category CommutativeSemigroups() may just check that the provided generators commute, implicitly assuming that generators indeed generate the semigroup (as required by Semigroups()).

run(category=None, skip=[], catch=True, raise_on_failure=False, **options)#

Run all the tests from this test suite:

INPUT:

  • category - a category; reserved for future use

  • skip - a string or list (or iterable) of strings

  • raise_on_failure - a boolean (default: False)

  • catch - a boolean (default: True)

All other options are passed down to the individual tests.

EXAMPLES:

sage: TestSuite(ZZ).run()

We now use the verbose option:

sage: TestSuite(1).run(verbose = True)
running ._test_category() . . . pass
running ._test_eq() . . . pass
running ._test_new() . . . pass
running ._test_nonzero_equal() . . . pass
running ._test_not_implemented_methods() . . . pass
running ._test_pickling() . . . pass

Some tests may be skipped using the skip option:

sage: TestSuite(1).run(verbose = True, skip ="_test_pickling")
running ._test_category() . . . pass
running ._test_eq() . . . pass
running ._test_new() . . . pass
running ._test_nonzero_equal() . . . pass
running ._test_not_implemented_methods() . . . pass
sage: TestSuite(1).run(verbose = True, skip =["_test_pickling", "_test_category"])
running ._test_eq() . . . pass
running ._test_new() . . . pass
running ._test_nonzero_equal() . . . pass
running ._test_not_implemented_methods() . . . pass

We now show (and test) some standard error reports:

sage: class Blah(SageObject):
....:     def _test_a(self, tester): pass
....:     def _test_b(self, tester): tester.fail()
....:     def _test_c(self, tester): pass
....:     def _test_d(self, tester): tester.fail()

sage: TestSuite(Blah()).run()
Failure in _test_b:
Traceback (most recent call last):
  ...
AssertionError: None
------------------------------------------------------------
Failure in _test_d:
Traceback (most recent call last):
  ...
AssertionError: None
------------------------------------------------------------
Failure in _test_pickling:
Traceback (most recent call last):
  ...
...PicklingError: Can't pickle <class '__main__.Blah'>: attribute
lookup ...Blah... failed
------------------------------------------------------------
The following tests failed: _test_b, _test_d, _test_pickling

sage: TestSuite(Blah()).run(verbose = True)
running ._test_a() . . . pass
running ._test_b() . . . fail
Traceback (most recent call last):
  ...
AssertionError: None
------------------------------------------------------------
running ._test_c() . . . pass
running ._test_category() . . . pass
running ._test_d() . . . fail
Traceback (most recent call last):
  ...
AssertionError: None
------------------------------------------------------------
running ._test_new() . . . pass
running ._test_not_implemented_methods() . . . pass
running ._test_pickling() . . . fail
Traceback (most recent call last):
  ...
...PicklingError: Can't pickle <class '__main__.Blah'>: attribute
lookup ...Blah... failed
------------------------------------------------------------
The following tests failed: _test_b, _test_d, _test_pickling

File "/opt/sage/local/lib/python/site-packages/sage/misc/sage_unittest.py", line 183, in run
test_method(tester = tester)

The catch=False option prevents TestSuite from catching exceptions:

sage: TestSuite(Blah()).run(catch=False)
Traceback (most recent call last):
  ...
  File ..., in _test_b
    def _test_b(self, tester): tester.fail()
  ...
AssertionError: None

In conjunction with %pdb on, this allows for the debugger to jump directly to the first failure location.

exception sage.misc.sage_unittest.TestSuiteFailure#

Bases: AssertionError

sage.misc.sage_unittest.instance_tester(instance, tester=None, **options)#

Return a gadget attached to instance providing testing utilities.

EXAMPLES:

sage: from sage.misc.sage_unittest import instance_tester
sage: tester = instance_tester(ZZ)

sage: tester.assertTrue(1 == 1)
sage: tester.assertTrue(1 == 0)
Traceback (most recent call last):
...
AssertionError: False is not true
sage: tester.assertTrue(1 == 0, "this is expected to fail")
Traceback (most recent call last):
...
AssertionError: this is expected to fail

sage: tester.assertEqual(1, 1)
sage: tester.assertEqual(1, 0)
Traceback (most recent call last):
...
AssertionError: 1 != 0

The available assertion testing facilities are the same as in unittest.TestCase [UNITTEST], which see (actually, by a slight abuse, tester is currently an instance of this class).