Skip to main content

Test Scaffolding

What is pytest?

Pytest is a unit testing framework intended to test the Python code you write.

Install pytest

To install pytest run this command:

pip install pytest

You can install a specific pytest version, for example:

pip install pytest==5.4.1

We use version 5.4.1 internally, but these tests should run on later versions of pytest as well.

Run pytest

Once you install pytest, you should be able to run your tests from the workspace root via:

ayx_plugin_cli test

or

pytest .\backend

Pytest looks through this directory for any Python functions that fit the pattern test_*.

IDE Integration

One nice thing about pytest is that it works well with all widely used integrated development environments (IDEs), like PyCharm, and VSCode. You can discover and run tests from these IDEs with the touch of a button, and get test results delivered in a nice breakdown. However, there might be some setup involved to get this integration working.

PyCharm

Go to File > Settings > Tools > Integrated Python Tools, and make sure that under Testing, the default test runner is set to pytest.

Once this is set, PyCharm takes a moment to index your files and discover the tests. When PyCharm is finished, a green run button displays to the left of each test you can run. To run all of the tests in a file, select and hold (or right-click) on the file and select the Run pytest in <filename> option.

VSCode

When you open VSCode, you should see a Testing tab on the leftmost side. Select this tab, and it should indicate that no tests have been found yet and that you can configure a test framework. Select Configure Python Tests, and then select pytest. Use the backend folder as the directory that contains the tests. The discovered tests should populate in that window.

Notes

If test discovery fails, and the error message given is an import error, this could be due to an environment issue. Check the bottom-right corner of the IDE to make sure it's set to the conda environment where your ayx_python_sdk and ayx_plugin_cli are installed.

You can set breakpoints in both IDEs in roughly the same way. Hover over the line numbers on the left to see a little red dot appear. Select and run the test in debug mode, and you'll be able to stop execution at that point and examine the values of local variables.

Generate Test

Any plugins generated with ayx_plugin_cli version 1.0.3+ should also come with unit test scaffolds. However, you can generate these scaffolds for early v2 plugins as well with the new ayx_plugin_cli generate-tests command.

Command Usage

To generate tests for all tools in a workspace, run this command at the workspace root:

ayx_plugin_cli generate-tests

To specify which tools you want to generate tests for, use:

ayx_plugin_cli generate-tests --tool-name Tool1 --tool-name Tool2

Note that you can only generate tools this way if the tool already exists in your ayx_workspace.json, using the same name as the tool’s tool_class_name attribute.

Also note that if you generate tests for tools that already exist, the tests need to be updated to reflect your plugin code.

Write Tests

To discover tests (by default), pytest searches for files that match the patterns test_*.py or *_test.py. Any functions that match the pattern test_*() are marked as tests within these files.

Use assert statements to check your code's values against an expected outcome, for example:

assert "value" in ["expected", "value"]

When a test runs, there are 3 possible outcomes:

  • The test code runs to the end without issue.

  • An unexpected error is raised.

  • The test fails on an assertion check.

The goal of unit testing is to discover these test failures ahead of time and debug your logic until your output matches the expected output.

Interact with the Plugin Service

SdkToolTestService is a middleware layer that mocks out some of Alteryx Designer's functionality and allows you to test your plugin's callbacks in an isolated environment. By default, we generate a pytest fixture that wraps and returns an instance of SdkToolTestService. The class contains several helper methods and attributes to make your testing experience easier:

io_stream

This attribute mocks out Alteryx Designer's Messages window—basically, any strings that are sent over through provider.io calls. When you run your test, you can examine plugin_service_fixture.io_stream to see which messages were captured, and compare them against a list of expected messages. Note that io_stream is a list of strings, prepended with the provider.io call in question. This is roughly the format they follow:

"<INFO|WARN|ERROR>:<message>"

For example, this `provider.io` call in the plugin code:

self.provider.io.info("Test Code") shows up as "INFO:Test Code" in io_stream.

data_streams

This attribute mocks out the plugin's output anchor. In other words, any data that shows up in a Browse tool placed after the plugin in Designer, should show up here. In the plugin code, this is any data that is written to provider.io.write_to_anchor(). When you run your test, you can examine data_streams to ensure that the correct output data was written to the output anchor, and compare the captured record batches against a list of expected record batches.

For the purpose of simplicity, the completed stream is represented by a dictionary, typed like this:

{ "<Output Anchor name>": [pa.RecordBatch] }

If provider.io.write_to_anchor is never called, the data_streams attribute should be an empty dictionary.

run_on_record_batch

This method runs your plugin's on_record_batch method. Pass it an input anchor and a corresponding record batch, and it should run the method and capture the I/O and data stream outputs.

run_on_incoming_connection_complete

This method runs your plugin's on_incoming_connection_complete method, on the specified input anchor, and captures the data and I/O output.

run_on_complete

This method runs your plugin's on_complete method and captures the data and I/O output.

Autogenerated Tests

By default, we generate these 4 tests:

  • test_init

  • test_on_record_batch

  • test_on_incoming_connection_complete

  • test_on_complete

However, you can add as many, or as few, as needed. By default, these run the corresponding SdkToolTestService methods and compare them to the default output.

One thing to note is that the on_record_batch test is parametrized, and runs 3 times by default, one for each batch named in the list argument.

These batches are defined in conftest.py. Edit, rename, and change these to suit your testing needs.

Examples

@pytest.mark.parametrize("record_batch", ["small_batch", "medium_batch", "large_batch"])
def test_on_record_batch(plugin_service_fixture, anchor, record_batch, request):
    record_batch = request.getfixturevalue(record_batch)
    plugin_service_fixture.run_on_record_batch(record_batch, anchor)
    #  In this case, since the tool is a simple passthrough, the input data should match the output data, 1-1.
    assert plugin_service_fixture.data_streams["Output"] == [record_batch]
    #  In this case, there are no calls being made to provider.io, so the io_stream for on_record_batch should be empty.
    assert plugin_service_fixture.io_stream == []
def test_on_incoming_connection_complete(plugin_service_fixture, anchor):
    plugin_service_fixture.run_on_incoming_connection_complete(anchor)
    #  In this case, no data was written to any of the output anchors, so the streams should be empty.
    assert plugin_service_fixture.data_streams == {}
    #  In this case, the only call being made is "Connection connection on Anchor anchor" as an info message.
    assert plugin_service_fixture.io_stream == [f"INFO:Connection {anchor.connection} on Anchor {anchor.name}"]
def test_on_complete(plugin_service_fixture):
    plugin_service_fixture.run_on_complete()
    #  In this case, no data was written to any of the output anchors, so the streams should be empty.
    assert plugin_service_fixture.data_streams == {}
    #  In this case, the only call being made is "Pass through tool done" as an info message.
    assert plugin_service_fixture.io_stream == ["INFO:Pass through tool done"]