integrate - Testing framework for integration tests

The usual testing frameworks you can find for Python are so-called unit-testing-frameworks. As this collides with some goals of integration testing this framework was built.

Changelog

1.0.0

• First release, supports Python >= 3.3

1.1.0

• Backport to work on Python 2.7
• If you want to use Python 3 use at least 3.3

1.2.0

• Bugfix: Python 2 issue (Re #4)
• Ignore Test Base classes (Thanks withrocks: https://github.com/withrocks)
• Testrunner.run() now returns stats (Thanks nooploop: https://github.com/nooploop)

What's the difference between integration testing and unit-testing?

In unit testing you should only test the smallest possible thing in your application (or library) and all test should be independent and the ordering of tests is not of particular interest. To make that possible it is very common to stub or mock data models or other external dependencies and only test the functionality of your data transforms. You define fixtures and setup and teardown functions to prepare the environment for a unit test and destroy that environment immediately after running the test.

Integration testing tests the complete system and has no direct influence on the data in the system so naturally the ordering of tests matter and tests may have dependencies (one test creates a DB record which another one modifies, etc.)

As you should test one thing in an unit-test only, you should only use a single assert in a unit test. The default behaviour of most assert methods in unit-testing-frameworks is to raise an exception. This means everything after the failed assert is skipped and the test is marked as failed. In integration-testing you probably want to check multiple things at once and get a detailed error report which checks failed exactly.

What does this framework better than others?

1. You have full control over which tests run in what order (if you want)
2. You can mark tests as dependent on others, so they will be skipped when a dependency fails
3. You can log as many errors as you want in a single test. Every error saves a stack trace and the error message for further debugging
4. The test runner is very flexible (Your python files that contain the tests do not need to be in python modules)

How to use it aka Quickstart

1. Install:

   pip install integrate
   

2. Write a test case class:

   from integrate import TestCase, test
   
   class Test(TestCase):
           "Simple test case"
   
           @test(skip_if_failed=["other_test"])
           def simple_test(self, check):
                   check.equal(1,2)
   
           @test()
           def other_test(self, check):
                   "Always failing test"
                   check.fail("Always fails")
   

3. Write a test-runner (we assume you put the test case into tests/test.py):

   #!/usr/bin/env python
   
   from integrate import TestRunner
   TestRunner(dirs=["tests"], pattern="*.py").run()
   

4. Make it executable and run it (or run with python testrunner.py):

   chmod a+x testrunner.py
   ./testrunner.py
   

Short documentation

All tests have to be in a class that is derived from TestCase, you may put anything in that class that you want, functions that should be called by the test runner have to be decorated with @test().

TestCase class

The TestCase class is the workhorse of the integrate framework. It has some functions you may override in a subclass in addition to adding test functions:

• def setup_all(self): This function is run at the beginning of the test case class before any test is started
• def teardown_all(self): This function is run at the end of the test case class after all tests have finished
• def setup_test(self): This function is run before each test in the test case
• def teardown_test(self): This function is run after each test in the test case

You can run the test case class by it's own by calling YourTestCase().run() or rely on the test runner

There are some interesting initialization parameters:

• verbosity, how verbose the test output should be (min: 0, max: 2, default: 2)
• checker, which Check subclass to use, usually you will use the default Check class, but you may extend that to add methods to the check object all tests receive

@test decorator

All functions that should be picked up by the TestCase class have to be decorated with @test() (notice the parentheses!), the decorator has some optional parameters:

• skip boolean, defaults to False, set to True to skip a test
• skip_if_failed list of strings, names of test functions that have to succeed (not fail or be skipped) in order for this function to run, defaults to an empty list
• depends list of strings, names of test functions that should be run before this function, defaults to an empty list
• expect_fail boolean, set to true if you expect this test to fail (just for logging purposes)

The test functions have 2 parameters: self and check, for the description of check read on.

Check class

All errors that surface in a test should be found and logged by an instance of the Check class. You may subclass this class to add additional checker functions and insert it into the TestCase or TestRunner initializer.

The assertion API looks like the following, if there is a message parameter it usually is optional and may be left out. User messages are prepended to an error message:

• equal(a, b, message=None) Check if two values are equal
• not_equal(a, b, message=None) Check if two values are not equal
• is_none(a, message=None) Check if a value is None
• is_not_none(a, message=None) Check if a value is not None
• is_true(a, message=None) Check if a value is True
• is_false(a, message=None) Check if a value is False
• fail(message) Just log an error message
• raises(exception_type, function, *args, **kwargs) Check if a function raises a specified exception type, args and kwargs are forwarded to the function
• does_not_raise(function, *args, **kwargs) Check if a function does not raise an exception, args and kwargs are forwarded to the function

All check functions should return True if the check succeeded and False if it failed if they don't have to return any other result (like the raises and does_not_raise functions which return the result of the function or None)

Exceptions in test functions will still cancel the test function and log the exception to the error log if you don't wrap it with a raises() call. The traceback of an exception caught by the toplevel will be not of much use though if you can't pinpoint the location based on the exception type. If you just want to catch all exceptions use check.raises(Exception, myFunc, "myParam")

For extending the Check class there is a, rather small, extension API:

• log_error(error, message, detail=None, strip=4) Use this function to add an error to the list, a corresponding stack trace is appended automatically. The error parameter is a textual one line description of the error, the message parameter is a user message. Use the detail parameter to give a detailed error description if needed. Only modify the strip parameter if your stacktrace gets entries after the error location in the test, by default it strips the last 4 stack frames as these are in the testing framework and just clobber the stack traces.
• error_message() Use this for debugging, this function joins all error messages into one string

TestRunner class

The TestRunner class is the entry point for automatically discovering tests in a project and running them. It has some initialization parameters:

• verbosity verbosity of test output (min: 0, max: 2, default: 2)
• dirs list of directories to scan for tests, defaults to current directory. Directories are scanned recursively.
• pattern file name pattern to search (argument to fnmatch) defaults to the python best practice *_test.py
• checker the Check subclass to send to the tests, if you have subclassed the Check class put your class here, defaults to the unmodified Check class

To start the tests instanciate the test runner and call the run() function:

from integrate import TestRunner
TestRunner().run()

If you want to run just some tests of your test suite you may either run the tests directly by calling run() on the TestCase subclass or by supplying a filter to the run() function of the test runner like so:

TestRunner().run(only='special.')

This example would only run tests which have a module name that starts with special. the module names are generated by replacing all slashes of the python file path with a dot and removing the .py extension.

A test file that is stored in the path special/tests/runme.py will get a module name of special.tests.runme.

If you only want to look at what the Test runner would actually do use the plan() function, this just displays a list of test that would be executed and the order of execution instead of really running the tests. It has the same parameters as the run() function.