Skip to content
This repository has been archived by the owner on Nov 23, 2023. It is now read-only.

Utilities for development of KNIME nodes calling Python scripts.

License

Notifications You must be signed in to change notification settings

3D-e-Chem/knime-python-wrapper

Repository files navigation

Only supported for KNIME versions <=4.3, no more changes will be made to this repo as of Nov 2023.

Abstract Python wrapper KNIME node and helpers. Used for development of KNIME nodes calling Python scripts.

Build Status Build status Quality Gate Status Coverage DOI

The nodes in Scripting>Python folder of the node repository (nodes part of the KNIME Python integration plugin) the end-user needs to paste Python code in a text area in the node dialog. Nodes derived from this repo will have a Python script included in their jar file and the dialog of the node will contain no source code text area. The included Python script is not editable by the end-user, but can read options from dialog like the input column name.

Usage

Instructions for KNIME node developers that want to call a Python script. Several steps must be performed:

  1. Add update site
  2. Add dependency
  3. Implement
  4. Write tests

Add update site

The releases of this repository are available in the https://3d-e-chem.github.io/updates update site.

Configure target platform by adding the https://3d-e-chem.github.io/updates update site with Abstract Python wrapper KNIME node and helpers software.

To make use of it in a Tycho based project, add to targetplatform/KNIME-AP-4.0.target file the following:

<location includeAllPlatforms="false" includeConfigurePhase="false" includeMode="planner" includeSource="true" type="InstallableUnit">
		<unit id="nl.esciencecenter.e3dchem.python.plugin" version="0.0.0"/>
		<repository location="https://3d-e-chem.github.io/updates"/>
</location>

Add dependency

To implement the node a dependency is needed for the plugin add tests. To do this add nl.esciencecenter.e3dchem.knime.python as a required plugin to the plugin/META-INF/MANIFEST.MF and tests/META-INF/MANIFEST.MF file.

Implement

Config

Create your node config class extended from the nl.esciencecenter.e3dchem.python.PythonWrapperNodeConfig class.

Overwrite the constructor to add required Python modules with the addRequiredModule("<module name>") method.

PythonWrapperNodeConfig class are configured for a single input table called input_table and a single output table called output_table. To change the the number and names of input and/or output tables, override the constructor.

Dialog

In your nodes dialog the Python options panel should be added by adding the following to the dialog constructor

pythonOptions = new PythonOptionsPanel<PredictMetabolitesConfig>();
addTab("Python options", pythonOptions);

To save the Python options to disk you must call the pythonOptions.saveSettingsTo(config) followed by config.saveToInDialog(settings) in the save*To() method of the dialog. To load the Python options from disk you must call the config.loadFromInDialog(settings) followed by pythonOptions.loadSettingsFrom(config) in the load*From() methods of the dialog.

Python script

Inside Python script the following variables are special:

  • options, dictionary filled from Java with PythonWrapperNodeConfig.getOptionsValues() method, to read from
  • input_table, Pandas Dataframe with input data table, to read from
  • output_table, Pandas Dataframe with output data table, to assign with value
  • flow_variables, dictionary of flow variables, to get or put key/value pairs
    • flow_variables['warning_message'], if key exists then value will be set as warning message of node

The Python script should be located in same directory as the model.

Model

Create your node model class extended from the nl.esciencecenter.e3dchem.python.PythonWrapperNodeModel class. Overwrite the python_code_filename fields in the constructor, this is the name of the Python script.

Write tests

To run tests which execute the node a call to PythonWrapperTestUtils.init() is required this will add the KNIME-Python utility scripts to the Python path and configure it to use the python3 executable in current PATH.

Build

To build the plugin and run the tests run the following command:

mvn verify

An Eclipse update site will be made in p2/target/repository repository. The update site can be used to perform a local installation.

Development

Steps to get development environment setup based on https://github.com/knime/knime-sdk-setup#sdk-setup:

  1. Install Java 8
  2. Install Eclipse for RCP and RAP developers
  3. Configure Java 8 inside Eclipse Window > Preferences > Java > Installed JREs
  4. Import this repo as an Existing Maven project
  5. Activate target platform by going to Window > Preferences > Plug-in Development > Target Platform and check the KNIME Analytics Platform (4.0) - nl.esciencecenter.e3dchem.python.targetplatform/KNIME-AP-4.0.target target definition.
  6. A KNIME Analytics Platform instance can be started by right clicking on the targetplatform/KNIME\ Analytics\ Platform.launch file and selecting Run As → KNIME Analytics Platform. The KNIME instance will contain the target platform together with all extensions defined in the workspace.

During import the Tycho Eclipse providers must be installed.

New release

  1. Update versions in pom files with mvn org.eclipse.tycho:tycho-versions-plugin:set-version -DnewVersion=<version>-SNAPSHOT command.
  2. Commit and push changes
  3. Create package with mvn package, will create update site in p2/target/repository
  4. Append new release to an update site
  5. Make clone of an update site repo
  6. Append release to the update site with mvn install -Dtarget.update.site=<path to update site>
  7. Commit and push changes in this repo and update site repo.
  8. Make nodes available to 3D-e-Chem KNIME feature by following steps at https://github.com/3D-e-Chem/knime-node-collection#new-release
  9. Create a GitHub release
  10. Update CITATION.cff file to reflect DOI which was generated by GitHub release on https://zenodo.org