Contribution guidelines

This page contains various resources for those who want to get started contributing code or documentation to SEM.

SEM installation for contributors

Contributors interested in developing new functionalities, fixing bugs, running tests and updating the documentation should use the following installation method.

First of all, clone the Github repository:

git clone https://github.com/DvdMgr/sem

This module is developed using pipenv. In order to manage virtual environments and install dependencies, make sure pipenv is installed. Typically, the following is enough:

pip install pipenv

From the project root, one can then install the sem package and the requirements with the following:

pipenv install --dev

After this step, a sub-shell using the new virtual environment can be created by calling:

pipenv shell

From here, the examples in examples/ can be run and a python REPL can be started to use the library interactively. Note that, since the sem package is marked to be installed in editable mode in the Pipfile, changes to the sem/ subdirectory will be effective immediately, without requiring a new installation.

Testing framework

SEM uses pytest to perform testing. By installing with the procedure described above, you should automatically download pytest and its xdist and cov extensions, allowing to perform parallel testing and coverage.

Tests are contained in the tests/ folder. Here, different files prefixed with test_ contain test cases for the various components of the library:

  • test_cli.py uses click facilities to test the command line interface;
  • test_database.py tests result insertion, querying and generally the DatabaseManager class;
  • test_examples.py runs the examples contained in the examples/ folder;
  • test_manager.py tests the CampaignManager object, from creation to simulation running and exporting results;
  • test_runner.py tests the SimulationRunner objects;
  • test_utils.py tests the utilities provided in sem/utils.py

The conftest file, on the other hand, contains code that must be run before each test execution (i.e., downloading and compiling ns-3) and fixtures (i.e., common data structures used by tests, like temporary folders containing ns-3 installations) that are shared between test files.

From the pipenv shell, at the project root, one can run the following command to run tests for the whole library:

py.test -x .

Tests contained in a single file or in a single function can also be run using the following formats:

py.test -x tests/test_cli.py
py.test -x -s ''/home/davide/Work/sem/tests/test_cli.py::test_cli_workflow''

In the commands above, the -x flag ensures that tests will stop at the first failure or error, while the -s flag disables standard output capturing by pytest, allowing the library to print to standard output also when tests are run. To perform testing in parallel via the xdist extension, add the -n N flag, with N equal to the number of parallel processes that should be used. Considered that most tests involve I/O, this speeds up testing considerably. Coverage can be computed by adding the –cov=sem flag, and using the –cov-report html flag will generate an html folder showing detailed coverage information. Finally, docstring testing is supported through the –doctest-modules flag.

The code contained in the documentation can be tested through the following command:

py.test --doctest-glob='*.rst' docs/

Typically, one would run tests on a single file while developing, and on the whole suite before submitting a pull request. The final commands that should be run before any pull request are the following:

py.test --doctest-glob='*.rst' docs/
py.test -x --cov=sem/ --cov-report html --doctest-modules

If both commands finish successfully, and if the new coverage percentage is not lower than the previous one, the pull request should be ready for submission.