This section describes the testing framework and format standards for tests in sunpy. Here we have heavily adapted the Astropy version, and it is worth reading that link.
The testing framework used by sunpy is the pytest framework, accessed through the
pytest project was formerly called
py.test, and you may
see the two spellings used interchangeably.
Dependencies for testing¶
Since the testing dependencies are not actually required to install or use sunpy, they are not included in “install_requires” in “setup.cfg”.
Developers who want to run the test suite will need to install the testing packages using pip:
$ pip install -e .[tests]
If you want to see the current test dependencies, you check “extras_require” in “setup.cfg”.
There are currently two different ways to invoke the sunpy tests.
However, we strongly suggest using
tox as the default one.
Each method uses the widely-used
pytest framework and are detailed below.
The primary method is to use tox, which is a generic virtualenv management and test command line tool. We have several environments within our “tox.ini” file and you can list them:
$ tox -v -l
Then you can run any of them doing:
$ tox -e <name of env>
This will create a test environment in “.tox” and build, install sunpy and runs the entire test suite. This is the method that our continuous integration uses.
The test suite can be run directly from the native
In this case, it is important for developers to be aware that they must manually rebuild any extensions by running
python setup.py build_ext before testing.
To run the entire suite with
will use the settings in
If you want to run one specific test file:
$ pytest sunpy/map/tests/test_mapbase.py
or one specific test in a test file:
$ pytest sunpy/map/tests/test_mapbase.py::<test_name>
(This does not work with
tox and is a known issue.)
If a test errors, you can use
pdb to create a debugging session at the moment the test fails:
$ pytest --pdb
Another method is to use
import sunpy sunpy.self_test()
You will see something like the following in your terminal:
Starting sunpy self test... Checking for packages needed to run sunpy: All required and optional sunpy dependencies are installed. Starting the sunpy test suite: ... The tests will run and will report any fails. You can report these through the `sunpy issue tracker <https://github.com/sunpy/sunpy/issues>`__ and we will strive to help.
It is possible to run this command in a situation where not all packages are installed. If this is the case, you will see the following when you run the test suite:
Starting sunpy self test... Checking for packages needed to run sunpy: The following packages are not installed for the sunpy[database] requirement: * sqlalchemy ... You do not have all the required dependencies installed to run the sunpy test suite. If you want to run the sunpy tests install the 'tests' extra with `pip install "sunpy[all,tests]"`
This does not mean sunpy is broken, but you will need to install the extra packages to ensure a “complete” installation of sunpy and run the entire test suite. It is quite likely that you will run into not having the tests dependencies installed.
By default, no online tests are selected and so to run the online tests you have to:
$ tox -e py38-online
$ pytest --remote-data=any
In order to avoid changes in figures due to different package versions, we recommend using tox to run the figure tests:
$ tox -e py38-figure
This will ensure that any figures created are checked using the package versions that were used to create the original figure hashes. Running this will create a folder, “figure_test_images”, within your work folder (“<local clone location>/figure_test_images”), which is ignored by git. Inside this folder will be all the images created, as well as a json file with the hashes of the figures created by the test run. The current hashes are located within “sunpy/tests/figure_hashes_mpl_<ver>_ft_<ver>_astropy_<ver>.json” and this will be where you will need to update old hashes or create new figure entries if anything changes. The filenames are the versions of Matplotlib, freetype and astropy used. If these versions differ to your local setup, the figure tests will not run. In theory, the Python version does not change the results as we have pinned the packages that cause the hash to vary.
Running tests in parallel¶
pip install pytest-xdist
Once installed, tests can be run in parallel using the
--parallel commandline option.
For example, to use 4 processes:
$ tox -e <name of environment> -- -n=4
$ pytest -n 4 ./sunpy
$ pip install pytest-cov
To generate a test coverage report, use:
$ pytest --cov ./sunpy
This will print to the terminal a report of line coverage of our test suite. If you want to create a report in html, you can run:
$ pytest --cov-report xml:cov.xml --cov ./sunpy $ coverage html
pytest has the following test discovery rules:
* ``test_*.py`` or ``*_test.py`` files * ``Test`` prefixed classes (without an ``__init__`` method) * ``test_`` prefixed functions and methods
We use the first one for our test files,
test_*.py and we suggest that developers follow this.
A rule of thumb for unit testing is to have at least one unit test per public function.
The following example shows a simple function and a test to test this function:
def func(x): """Add one to the argument.""" return x + 1 def test_answer(): """Check the return value of func() for an example argument.""" assert func(3) == 5
If we place this in a
test.py file and then run:
$ pytest test.py
The result is:
============================= test session starts ============================== python: platform darwin -- Python 3.8.3 -- pytest-3.2.0 test object 1: /Users/username/tmp/test.py test.py F =================================== FAILURES =================================== _________________________________ test_answer __________________________________ def test_answer(): > assert func(3) == 5 E assert 4 == 5 E + where 4 = func(3) test.py:5: AssertionError =========================== 1 failed in 0.07 seconds ===========================
Sometimes the output from the test suite will have
xfail meaning a test has passed although it has been marked as
skipped meaning a test that has been skipped due to not meeting some condition (online and figure tests are the most common).
You need to use the option
-rs for skipped tests and
-rx for xfailed tests, respectively.
-rxs for detailed information on both skipped and xfailed tests.
Where to put tests¶
Each package should include a suite of unit tests, covering as many of the public methods/functions as possible. These tests should be included inside each package, e.g:
“tests” directories should contain an
__init__.py file so that the tests can be imported.
There are some tests for functions and methods in sunpy that require a working connection to the internet.
pytest is configured in a way that it iterates over all tests that have been marked as
pytest.mark.remote_data and checks if there is an established connection to the internet.
If there is none, the test is skipped, otherwise it is run.
Marking tests is pretty straightforward, use the decorator
@pytest.mark.remote_data to mark a test function as needing an internet connection:
@pytest.mark.remote_data def func(x): """Add one to the argument.""" return x + 1
Tests that create files¶
Tests may often be run from directories where users do not have write permissions so tests which create files should always do so in temporary directories. This can be done with the pytest tmpdir function argument or with Python’s built-in tempfile module.
Tests that use test data¶
We store test data in “sunpy/data/test” as long as it is less than about 100 kB.
These data should always be accessed via the
This way you can use them when you create a test.
You can also use our sample data but this will have to be marked as an online test (see above):
import sunpy.data.sample @pytest.mark.remote_data def func(): """Returns the file path for the sample data.""" return sunpy.data.sample.AIA_131_IMAGE
Generally we do not run the tests on our sample data, so only do this if you have a valid reason.
Figure unit tests¶
The figure tests and the hashes they use are only checked on Linux and might be different on other platforms. We should suggest if you do not use a Linux, to add a fake hash to the json files and then CircleCi (ran on a PR) will tell you the real hash to use.
You can write sunpy unit tests that test the generation of Matplotlib figures by adding the decorator
Here is a simple example:
import matplotlib.pyplot as plt from sunpy.tests.helpers import figure_test @figure_test def test_simple_plot(): plt.plot([0,1])
The current figure at the end of the unit test, or an explicitly returned figure, has its hash (currently
SHA256) compared against an established hash collection (more on this below).
If the hashes do not match, the figure has changed, and thus the test is considered to have failed.
If you are adding a new figure test you will need to generate a new hash library:
$ tox -e py38-figure -- --mpl-generate-hash-library=sunpy/tests/figure_hashes_mpl_332_ft_261_astropy_42.json
The filename changes if the version of astropy or Matplotlib or freetype gets updated. So you might need to adjust this command. For the development figure tests:
$ tox -e py38-figure-devdeps -- --mpl-generate-hash-library=sunpy/tests/figure_hashes_mpl_dev_ft_261_astropy_dev.json
This will run the figure test suite and update the hashes stored.
If you want to check what the images look like, you can do:
$ tox -e py38-figure -- --mpl-generate-path=baseline
The images output from the tests will be stored in a folder called
baseline in the sunpy folder, so you can double check the test works as you expected.
Code examples in the documentation will also be run as tests and this helps to validate that the documentation is accurate and up to date. sunpy uses the same system as Astropy, so for information on writing doctests see the astropy documentation.
You do not have to do anything extra in order to run any documentation tests.
setup.cfg file we have set default options for
pytest, such that you only need to run:
$ pytest <file to test>
to run any documentation test.
In addition to writing unit tests new functionality, it is also a good practice to write a unit test each time a bug is found, and submit the unit test along with the fix for the problem. This way we can ensure that the bug does not re-emerge at a later time.