Unit Test Quickstart

From Gramps
Jump to: navigation, search

Recipes and tips for writing unit tests in the Gramps source code tree.

First, some general procedural and structural notes

  • unit tests are usually created to verify correct functional behavior of a single module of source code.
  • test code goes in a test subdirectory of the subject module's directory
  • the test module should be named with a suffix _test so that it can be found by automatic regression test tools.
  • the leading part of the test will commonly be named the same as the basename (or a variant) of the module under test, leading, for example, to the following files
  • the test subdirectory can contain data or helper programs as required
  • the test and such supplementary elements that are persistent will be maintained as part of the source control system.
  • the test module can create and delete test data during execution. Commonly this would go in a deeper subdir, named to avoid collision with other test programs.

This article's content is incomplete or a placeholder stub.
Please update or expand this section.

Running Unit Tests

A single unit test can be run from the top-level Gramps directory.

GRAMPS_RESOURCES=. python -m unittest <test_module>

for example:

GRAMPS_RESOURCES=. python -m unittest gramps.gen.lib.test.date_test

To run all the unit tests use:

GRAMPS_RESOURCES=. python -m unittest discover -p '*_test.py'

For verbose output use the -v flag:

GRAMPS_RESOURCES=. python -m unittest discover -p '*_test.py' -v

For convenience, all the unit tests can be run from the setup.py script using:

python setup.py test

For verbose output use:

python setup.py --verbose test

Simple Code Recipe

There are only a few firm requirements to fit into the framework. Here is a simple test module that may be considered something of a template.

import unittest

# import the module under test
import ..MyModule                # use your module name
# find the data directory
this_dir = os.path.dirname(__file___)
data_dir = os.path.join(this_dir, 'data')
# unittest requires a TestCase class containing test function members 
# optional setUp() and tearDown() functions can perform pre/post test housekeeping
class Test_top(unittest.TestCase):
     def setUp(self):
     def tearDown(self):

     def test_action_x_leads_to_y(self):
         ..do stuff..
         self.assertTrue(expression, "message to display on failure")
         #see other assert and fail functions in the unittest module

     ..more defs for more tests, more classes for possible grouping logic..

if __name__ == "__main__":

Using Your Unit Test

  • create a module_test.py if and whenever it seems like it might be useful
  • create a test subdir if one is not already there
  • one practice might be as follows
cd ...gramps/src
export PYTHONPATH=`pwd`
cd A/B/test
gvim +ba ../module.py module_test.py
  • repeat
    • do some editing
    • do some testing as follows (optional -v shows extra progress messages)
 python module_test.py -v
  • until happy
  • check-in your module code and test code

This works in gramps40/trunk from the toplevel dir to run the existing exportvcard_test.py:

GRAMPS_RESOURCES=$PWD PYTHONPATH=$PWD:$PWD/gramps python gramps/plugins/export/test/exportvcard_test.py

See also Testing Gramps.

Mocking external dependencies with unittest.mock

When you develop a class or a method with external dependencies, you often would like to provide test coverage just to the developed code proper, not the external dependencies it exercised. For instance, if your code calls sys.stderr.write, there is no need to check that sys.stderr.write actually outputs anything on the system standard error file descriptor. Rather, you'd like to know that you code calls the write method and what arguments get passed.

Python3.3 onwards provides the unittest.mock module for such tasks. For Python 3.2 and earlier you can install the "mock" module that back-ports this functionality, available as "python-mock" package on Ubuntu or from https://pypi.python.org/pypi/mock

Gramps tests using mocks will be skipped if the module is not available.

See also