Python Gradle Plugin

Now, you can run Python scripts with Java and Gradle builds only.

This Gradle Plugin automatically downloads and installs specific Conda tool to run python scripts or other executables (pip, conda etc.) from virtual env.

Python project configuration can be fully automated by Gradle tasks now.

Requirements

• Java 11 or higher
• Gradle 6.2 or higher project

Purpose

Running python scripts or projects by executing single task which will download and install Python virtual environment.

Additional Python configuration (pip/conda packages installation etc.) can be done by defining Gradle tasks in build.gradle.kts file.

Usage

Steps to run python script from Gradle

1. Apply a plugin to a project as described on gradle portal.

2. Configure a plugin by specifying desired python version in build script:

   pythonPlugin {
       pythonVersion = "3.8.2"
   }

3. Define a task to run desired python script, for example to run quicksort.py script in main dir add the following task configuration to build script:

   tasks {
       register<VenvTask>("runQuickSort") {
               workingDir = projectDir.resolve("main")
               args = listOf("quicksort.py")
       }
   }

4. Run python script from gradle:

   # Linux
   ./gradlew runQuickSort
   # Windows
   gradlew.bat runQuickSort

5. Wait until Conda is installed and environment set up.

6. Enjoy :)

Python Plugin properties

Plugin default behavior can be adjusted by specifying the following properties:

• pythonVersion -> Python environment version, default 3.13.0, the available ones can be checked at https://anaconda.org/conda-forge/python/
• condaVersion -> Miniconda or Anaconda version, default py312_24.9.2-0, the available ones can be checked at https://repo.anaconda.com/miniconda/ or https://repo.anaconda.com/archive/ (Anaconda)
• condaInstaller -> Conda environment installer name, default is Miniconda3, for Anaconda change to Anaconda3
• condaRepoUrl -> repository URL which should be used to download binaries, default https://repo.anaconda.com/miniconda, for Anaconda https://repo.anaconda.com/archive/
• condaRepoUsername -> username for the basic auth if needed, absent by default
• condaRepoPassword -> password for the basic auth, used if condaRepoUsername is specified, should not be passed directly in script file, can be supplied by gradle properties , absent by default
• condaRepoHeaders -> additional optional headers used for connection, empty map by default
• useHomeDir -> when true the default install directory will be the one from user.home system property, false by default
• installDir -> property to customize conda installation directory, equals to <rootProjectDir>/.gradle/python by default or user home if useHomeDir = true
• systemArch -> operating system architecture, default is got from os.arch system property
• ideaDir -> target .idea directory to detect Intellij project, equals to <rootProjectDir>/.idea by default

Sample extension configuration inside of build.gradle.kts file:

pythonPlugin {
    pythonVersion = "3.7.0"
    condaVersion = "2022.05"
    condaInstaller = "Anaconda3"
    condaRepoUrl = "https://nexus.com/repositories/conda"
    condaRepoUsername = "user"
    condaRepoPassword = extra["conda.repo.pass"].toString()
    condaRepoHeaders = mapOf(
        "CUSTOM_HEADER_1" to "headerValue1",
        "CUSTOM_HEADER_2" to "headerValue2"
    )
    installDir = file(layout.buildDirectory.dir("python"))
    systemArch = "arm64"
}

VenvTask task properties

All tasks which should be executed in virtual env can be customized as well by the following fields:

• venvExec -> name of executable from virtual env which will be executed, python by default
• inputFile -> optional input file, none by default
• outputFile -> optional output file, none by default
• args -> list of arguments for a venvExec executable, empty by default
• workingDir -> working directory, project directory by default
• environment -> map with environment variables to apply during the execution, empty by default

Sample VenvTask configuration inside of build.gradle.kts file:

register<VenvTask>("runPythonScript") {
    venvExec = "python"
    inputFile = file("inputFile.txt")
    outputFile = file("outputFile.txt")
    args = listOf("--some-flag", "arg1")
    workingDir = projectDir.resolve("main")
    environment = mapOf("ENV_VAR_TO_PRINT" to "sampleEnvVar")
}

Additional examples alongside with sample PipTasks configurations can be found in examples module in this project.

Intellij setup

Auto import installed Python SDK:

• Install SDK-Import Intellij Plugin.
• Execute gradle envSetup task.
• Choose from "Tools" -> "Reimport SDK" to import installed Python SDK with plugin.

Manual way:

• To have autocomplete and modules properly recognized in Intellij Idea point to Conda environment as described in: https://www.jetbrains.com/help/idea/configuring-python-sdk.html
• To have properly recognized imported source modules in tests, right click on sources directory (for example main) -> Mark Directory as -> as Sources root.

Conda install directories

• Linux - <installDir>/.gradle/python/Linux/<condaInstaller>-<condaVersion>

• Windows - <installDir>/.gradle/python/Windows/<condaInstaller>-<condaVersion>

• MacOSX - <installDir>/.gradle/python/MacOSX/<condaInstaller>-<condaVersion>

Where <installDir> is the root catalog where the Conda will be installed specified by installDir property, <condaInstaller> is Conda installer e.g. Miniconda3 and <condaVersion> is Conda installer version e.g. py38_4.8.3

If you are familiar with conda you can also execute conda commands like conda deactivate or conda install directly with the binaries from the catalogs above.

Notes

• It may be required to unset PYTHONPATH in system before running any tasks (https://stackoverflow.com/a/31841132)

• You can also run some simple inline Python scripts inside build files like this:

  

  Intellij 'inject language' feature can be useful in such scenario :)

Known Issues

• /usr/bin/env: ‘python’: No such file or directory when executing envSetup task -> It is related to the shebang char limit which is 128.

  When installing the conda if the prefix path is longer than the limit the default shebang (#!/usr/bin/env python) is being used in the installed conda script file (condabin/conda). Since no python binary is accessible by this path the exception is being thrown.

  The easiest solution is to store the root project at the shortest possible path or use the installDir to specify shorter path per particular troublesome subproject.