icontract-hypothesis 1.1.4

Library

There are two ways to integrate icontract-hypothesis in your tests as a library.

Only assume. First, you can use it for defining the assumptions of the test based on the precondition:

>>> from hypothesis import given
>>> import hypothesis.strategies as st

>>> import icontract
>>> import icontract_hypothesis

>>> @icontract.require(lambda x: x > 0)
... @icontract.ensure(lambda result: result > 0)
... def some_func(x: int) -> int:
...     return x - 1000

>>> assume_preconditions = icontract_hypothesis.make_assume_preconditions(
...     some_func)

>>> @given(x=st.integers())
... def test_some_func(x: int) -> None:
...    assume_preconditions(x)
...    some_func(x)

>>> test_some_func()
Traceback (most recent call last):
    ...
icontract.errors.ViolationError: File <doctest README.rst[4]>, line 2 in <module>:
result > 0:
result was -999
x was 1

The function assume_preconditions created by icontract_hypothesis.make_assume_preconditions will reject all the input values which do not satisfy the pre-conditions of some_func.

Infer strategy. Second, you can automatically infer the strategy and test the function:

>>> import icontract
>>> import icontract_hypothesis

>>> @icontract.require(lambda x: x > 0)
... @icontract.ensure(lambda result: result > 0)
... def some_func(x: int) -> int:
...     return x - 1000

>>> icontract_hypothesis.test_with_inferred_strategy(some_func)
Traceback (most recent call last):
    ...
icontract.errors.ViolationError: File <doctest README.rst[10]>, line 2 in <module>:
result > 0:
result was -999
x was 1

Which approach to use depends on how you want to write your tests. The first approach, using assume_preconditions, is practical if you already defined your search strategy and you only want to exclude a few edge cases. The second approach, automatically inferring test strategies, is useful if you just want to test your function without specifying any particular search strategy.

Icontract-hypothesis guarantees that the inferred strategy must satisfy the preconditions. If it does not, you should consider it a bug in which case please create an issue so that we can fix it.

If you want to inspect the strategy or further refine it programmatically, use icontract_hypothesis.infer_strategy:

>>> import math

>>> import icontract
>>> import icontract_hypothesis

>>> @icontract.require(lambda x: x > 0)
... @icontract.require(lambda x: x > math.sqrt(x))
... def some_func(x: float) -> int:
...     pass

>>> icontract_hypothesis.infer_strategy(some_func)
fixed_dictionaries({'x': floats(min_value=0, exclude_min=True).filter(lambda x: x > math.sqrt(x))})

Testing Tool

We provide pyicontract-hypothesis test command-line tool which you can use to automatically test a module.

usage: pyicontract-hypothesis test [-h] -p PATH
                                   [--settings [SETTINGS [SETTINGS ...]]]
                                   [--inspect] [-i [INCLUDE [INCLUDE ...]]]
                                   [-e [EXCLUDE [EXCLUDE ...]]]

optional arguments:
  -h, --help            show this help message and exit
  -p PATH, --path PATH  Path to the Python file to test
  --settings [SETTINGS [SETTINGS ...]]
                        Specify settings for Hypothesis

                        The settings are assigned by '='.
                        The value of the setting needs to be encoded as JSON.

                        Example: max_examples=500
  --inspect             Only show the strategy and the settings

                        No tests are executed.
  -i [INCLUDE [INCLUDE ...]], --include [INCLUDE [INCLUDE ...]]
                        Regular expressions, lines or line ranges of the functions to process

                        If a line or line range overlaps the body of a function,
                        the function is considered included.

                        Example 1: ^do_something.*$
                        Example 2: 3
                        Example 3: 34-65
  -e [EXCLUDE [EXCLUDE ...]], --exclude [EXCLUDE [EXCLUDE ...]]
                        Regular expressions, lines or line ranges of the functions to exclude

                        If a line or line range overlaps the body of a function,
                        the function is considered excluded.

                        Example 1: ^do_something.*$
                        Example 2: 3
                        Example 3: 34-65

Note that pyicontract-hypothesis test can be trivially integrated with your IDE if you can pass in the current cursor position and the current file name.

Ghostwriting Tool

Writing property-based tests by hand is tedious and can be partially automated. To that end, we implemented a ghostwriter utility pyicontract-hypothesis ghostwrite that generates a first draft based on pre-conditions that you manually refine further:

usage: pyicontract-hypothesis ghostwrite [-h] (-m MODULE | -p PATH)
                                         [-o OUTPUT] [--explicit] [--bare]
                                         [-i [INCLUDE [INCLUDE ...]]]
                                         [-e [EXCLUDE [EXCLUDE ...]]]

optional arguments:
  -h, --help            show this help message and exit
  -m MODULE, --module MODULE
                        Module to ghostwrite the unit tests for
  -p PATH, --path PATH  Path to the module to ghostwrite the unit tests for.

                        If the file represents a module reachable through
                        sys.path, use the qualified module name in
                        the unit test.

                        Otherwise, the module is represented as the stem
                        of the path with all non-identifier characters
                        replaced with an underscore ("_").
  -o OUTPUT, --output OUTPUT
                        Path to the file where the output should be written.

                        If '-', writes to STDOUT.
  --explicit            Write the inferred strategies explicitly

                        This is practical if you want to tune and
                        refine the strategies and just want to use
                        ghostwriting as a starting point.

                        Mind that pyicontract-hypothesis does not
                        automatically fix imports as this is
                        usually project-specific. You have to fix imports
                        manually after the ghostwriting.
  --bare                Print only the body of the tests and omit header/footer
                        (such as TestCase class or import statements).

                        This is useful when you only want to inspect a single test or
                        include a single test function in a custom test suite.
  -i [INCLUDE [INCLUDE ...]], --include [INCLUDE [INCLUDE ...]]
                        Regular expressions, lines or line ranges of the functions to process

                        If a line or line range overlaps the body of a function,
                        the function is considered included.

                        Example 1: ^do_something.*$
                        Example 2: 3
                        Example 3: 34-65
  -e [EXCLUDE [EXCLUDE ...]], --exclude [EXCLUDE [EXCLUDE ...]]
                        Regular expressions, lines or line ranges of the functions to exclude

                        If a line or line range overlaps the body of a function,
                        the function is considered excluded.

                        Example 1: ^do_something.*$
                        Example 2: 3
                        Example 3: 34-65

The examples of ghostwritten tests are available at: tests/pyicontract_hypothesis/samples

Running Tools as Module

If for some reason you want to run the tools as a module, just invoke:

usage: icontract_hypothesis [-h] {test,ghostwrite} ...

Combine property-based testing with contracts of a Python module.

positional arguments:
  {test,ghostwrite}  Commands
    test             Test the functions automatically by inferring search
                     strategies from preconditions
    ghostwrite       Ghostwrite the unit tests with inferred search strategies

optional arguments:
  -h, --help         show this help message and exit

Classes

Hypothesis automatically builds composite input arguments (classes, dataclasses, named tuples etc.). If your class enforces pre-conditions in the constructor method (__init__), make sure that it inherits from icontract.DBC.

That way icontract-hypothesis will use hypothesis.strategies.register_type_strategy to register your class with Hypothesis and consider pre-conditions when building its instances.

It is important that you should not use hypothesis.strategies.builds with the classes using contracts in their constructors as builds will disregard the registered strategy. You should use hypothesis.strategies.from_type instead. See this comment on an Hypothesis issue and the corresponding answer.

Many times default inferred strategies for the constructors should be enough, though you are of course not restricted to them. You can register your own strategies with hypothesis.strategies.register_type_strategy . Icontract-hypothesis will respect the previous registrations and will not overwrite them.

Benchmarks

We run benchmarks against deal and dpcontracts libraries as part of our continuous integration.

We benchmark against functions using 1, 2 and 3 arguments, respectively, with the precondition that the argument should be positive (e.g., a > 0). We sampled 100 inputs per each run.

The following scripts were run:

• benchmarks/compare_with_others.py

The benchmarks were executed on Intel(R) Xeon(R) E-2276M CPU @ 2.80GHz. We used Python 3.8.5, icontract 2.4.1, deal 4.4.0 and dpcontracts 0.6.0.

The following tables summarize the results.

Benchmarking Hypothesis testing:

Argument count: 1

Argument count: 2

Argument count: 3