Testing Instrument Functionality

Overview

When developing new instrument classes, or adding functionality to existing instruments, it is important to also add automated checks for the correctness of the new functionality. Such tests serve two distinct purposes:

• Ensures that the protocol for each instrument is being followed correctly, even with changes in the underlying InstrumentKit behavior.

• Ensures that the API seen by external users is kept stable and consistent.

The former is especially important for instrument control, as the developers of InstrumentKit will not, in general, have access to each instrument that is supported— we rely on automated testing to ensure that future changes do not cause invalid or undesired operation.

For InstrumentKit, we rely heavily on pytest, a mature and flexible unit-testing framework for Python. When run from the command line via pytest, or when run by Travis CI, pytest will automatically execute functions and methods whose names start with test in packages, modules and classes whose names start with test or Test, depending. (Please see the pytest documentation for full details, as this is not intended to be a guide to pytest so much as a guide to how we use it in IK.) Because of this, we keep all test cases in the instruments.tests package, under a subpackage named for the particular manufacturer, such as instruments.tests.test_srs. The tests for each instrument should be contained within its own file. Please see current tests as an example. If the number of tests for a given instrument is numerous, please consider making modules within a manufacturer test subpackage for each particular device.

Below, we discuss two distinct kinds of unit tests: those that check that InstrumentKit functionality such as Property Factories work correctly for new instruments, and those that check that existing instruments produce correct protocols.

tox Based Testing

When submitting a PR, tests are run through the tool tox. It helps to provide some isolation between your source code and test code, and gives you quick access to separate dedicated test venvs.

While tox will setup and manage its virtual environments automatically, you will need a copy of each major version of Python you wish to test against. I suggest you use pyenv to do this. You can install this tool via the pyenv-installer tool. After installation, each Python version can be installed via the following pattern:

$ pyenv install 3.8.13

Afterwards, you can use tox to run the tests under that specific Python environment:

$ tox -e py38,py38-numpy

Here we have specified two tox environments that will be run: both under Python 3.8, one with numpy installed and the other without.

py38 can be subsituted for other supported versions of Python, assuming they have been installed.

You can also run all defined tox environments by simply running:

$ tox

Mock Instruments

TODO

Expected Protocols

As an example of asserting correctness of implemented protocols, let’s consider a simple test case for instruments.srs.SRSDG645`:

def test_srsdg645_output_level():
    """
    SRSDG645: Checks getting/setting unitful ouput level.
    """
    with expected_protocol(ik.srs.SRSDG645,
            [
                "LAMP? 1",
                "LAMP 1,4.0",
            ], [
                "3.2"
            ],
            sep="\n"
    ) as ddg:
        unit_eq(ddg.output['AB'].level_amplitude, u.Quantity(3.2, "V"))
        ddg.output['AB'].level_amplitude = 4.0

Here, we see that the test has a name beginning with test_, has a simple docstring that will be printed in reports on failing tests, and then has a call to expected_protocol(). The latter consists of specifying an instrument class, here given as ik.srs.DG645, then a list of expected outputs and playback to check property accessors.

Note that expected_protocol() acts as a context manager, such that it will, at the end of the indented block, assert the correct operation of the contents of that block. In this example, the second argument to expected_protocol() specifies that the instrument class should have sent out two strings, "LAMP? 1" and LAMP 1,4.0, during the block, and should act correctly when given an answer of "3.2" back from the instrument. The third parameter, sep specifies what will be appended to the end of each lines in the previous parameters. This lets you specify the termination character that will be used in the communication without having to write it out each and every time.

Protocol Assertion Functions