How to test controller plugins

For third-party controller plugin projects it is recommended to develop their integration tests with the pool, for example, during the TDD process or for their later use in the CI process.

We recommend developing tests with pytest and sardana built-in fixtures defined in sardana.pool.test.

We believe that the approach of integration testing together with the use of pytest is good because:

  • The sardana.pool.controller API was optimized for multi-axes commands and queries, frequently making the plugin methods like: PreStartAll(), PreStartOne(), StartOne(), StartAll() not maintaining the class invariants, hence not really suitable for unit testing.

  • Some controller plugin classes, during complex actions like motion or acquisition, may even not maintain the class invariants in between the series of calls to state and read methods what is again not really suitable for unit testing.

  • For the above reasons the most efficient way of testing would be to replicate the call algorithms from the pool. But since sardana is an evolving project, the integration testing strategy with actually using the pool algorithms, could early discover eventual incompatibilities between the plugins and newer versions of sardana.

  • pytest is the most complete and widespread python testing framework, which in contrary to unittest, which follows the xUnit testing patterns, provides more pythonic and modular way of developing tests.

  • With the sardana built-in fixtures, which require minimal configuration, you could directly focus on writing your tests without carrying too much about the setup and tear down. For example, they take care of executing the calls to methods like AddDevice() and DeleteDevice(), etc.

You will need to either fine tune the fixtures using markers e.g. kwargs or create custom fixtures using “Factory as fixture” e.g. create_motor_ctrl.

Until SEP19 gets implemented, the first thing that you will need to do is to tell the pool where to look for your controller plugins i.e. set the pool_path, by either:

  • changing the behavior of the pool fixture using the attrs marker e.g. for the whole test module by setting the pytestmark module attribute:

    pytestmark = [pytest.mark.attrs({"pool": {"pool_path": ["/path/to/your/controller/classes"]}})]
    
  • or, creating your own pool fixture using the create_pool “Factory as fixture”:

    @pytest.fixture
    def pool(create_pool):
        pool = create_pool()
        pool.pool_path = ["/path/to/your/controller/classes"]
        return pool
    

Next, you will need a controller object driven by your controller plugin. Here you can either:

  • change the behavior of the motctrl01 fixture:

    pytestmark = pytest.mark.kwargs({
        "motctrl01": {
            "klass": "MyMotorController",
            "library": "MyMotorCtrl.py"
        }
    })
    
  • or, create your own motctrl01 fixture using the create_motor_ctrl “Factory as fixture”:

    @pytest.fixture
    def motctrl01(create_motor_ctrl):
        kwargs = {
            "klass": "MyMotorController",
            "library": "MyMotorCtrl.py"
        }
        return create_motor_ctrl(kwargs)
    

Important

In case you create your own fixtures, it is recommended to follow naming convention of the fixtures defined in the sardana.pool.test module, as in the above example. This way you will be able to reuse the built-in dependent fixtures e.g. by creating controller fixture called motctrl01 you can reuse mot01 built-in fixture. These fixtures could be further fine tuned with markers.

So, now you could already start testing your controller:

def test_init(motctrl01):
    assert motctrl01.is_online() == True

or axis:

def test_get_state(mot01):
    assert mot01.state == State.On

Built-in controller elements fixtures are limited to the low axis numbers. For higher axis numbers you will need to define your own fixtures. For this, you could use the sardana.pool.test.util.mot() helper function e.g.:

mot99 = pytest.fixture(mot)

Hint

You could get some inspiration on writing controller tests from the sardana-tango project.