Custom nodes

maize-contrib is set up as a namespace package, with all custom nodes under maize/steps in appropriate subfolders.

Structure

To add your own nodes, find or create an appropriate subfolder under maize/steps/mai (or maize/steps/ext for projects external to the Molecular AI group). Add a python file with your custom node definition, and expose it using an __init__.py file. Every node should have its own test function as part of the same module, and all test data should be contained in the data child directory.

Node creation

Creating custom nodes is described in the user guide. Nodes made available through maize-contrib should be documented using the reStructuredText (reST) syntax (see the Sphinx docs for more information) and contain references to the literature if appropriate. A common strategy will be wrapping existing scripts that may require custom modules not part of the maize core. In this case, creating an additional file with this code and importing it in the run() method of your node is a useful pattern.

If you require rdkit functionality, you should probably use the chem module instead. The included Isomer class is a thin wrapper around the rdkit.Chem.Mol object, providing improved error handling and a more pythonic interface. Feel free to add additional functionality to the module as required.

Testing

All tests are implemented using pytest. Every node should have a test function in the same file. If there are multiple related tests, they should ideally be combined in a single test class. Some useful predefined fixtures are temp_working_dir, to perform the test in a temporary directory and test_config for a configuration file (specified as an additional pytest option in pyproject.toml).

To create a test for a node, use the TestRig utility class to wrap your node (with a custom configuration). You can then use the setup_run() convenience method to specify both parameters and inputs to your node. Note that inputs takes a list of items that will be sent to the node separately, to allow for tests of nodes that read from the same input multiple times. Node results can then be retrieved by accessing the corresponding output in the returned dictionary and calling the get() method:

def test_node(temp_working_dir, test_config):
    rig = TestRig(YourNode, config=test_config)
    result = rig.setup_run(parameters={"value": 42}, inputs={"inp": [17]})
    assert result["out"].get() == 59