Thanks a lot for using Py4J and wanting to contribute! Your help is greatly appreciated!
The following is a set of guidelines for contributing to Py4J. These are just guidelines, not rules, use your best judgment and feel free to propose changes to this document in a pull request.
These guidelines have evolved since Py4J started and past commits may not always follow them.
In general, if your contribution does not follow these guidelines, we will review your contribution, but expect that we will ask you for some changes and refer you to specific parts of this guide!
Contributions to Py4J are always welcome. To accelerate the review process, make sure that your contribution includes:
We follow pep8 rather stricly:
Code must be compatible with from Python 2.6 to the newest released version of Python.
If external libraries must be used, they should be wrapped in a mechanism that by default does not require them (e.g., conditional imports, graceful degradation, etc.).
Libraries used for testing and contributing (flake8, nose, tox, Sphinx) can be installed with pip:
pip install -r requirements.txt # Run flake8 flake8
You can format your code using gradle and Java 8:
# To check Java coding conventions: ./gradlew spotlessCheck # To format your code: ./gradlew spotlessJavaApply
We use FindBugs on the main source code (not the test source code) and any warnings must be corrected before a pull request will be accepted.
# To run findbugs ./gradlew findbugsMain
On the Python side, we use nose to run the test suite and tox to run the test suite across the supported python versions.
# make sure that the jar file is created cd py4j-java ./gradlew clean ./gradlew assemble # install test requirements cd py4j-python pip install -r requirements.txt # Run the full test suite nosetests # Run only one particular test nosetests py4j.tests.java_gateway_test:GatewayLauncherTest.testRedirectToDeque # Run all tests on all supported pythons. # Typically only do this if the automated build failed # on one version of python. tox # Run flake8 checks flake8
New code or bug fix should ideally be accompanied by a test case.
Because we start a JVM for most test cases, it may happen that some test fails because the process was not ready to receive a request. This is a problem we have been working on for some time now and it has been a few months since the test suite failed because of synchronization issues. If your tests constantly fail, then something is wrong with your test!
We use JUnit to write test cases.
cd py4j-java ./gradlew clean ./gradlew test
We have started to keep track of code coverage on both the Python and Java libraries. Unfortunately, most of the Java code is being tested by the Python library so the reported code coverage is smaller than the actual coverage.
When reviewing a contribution, we will always require that both the Java code and the Python code are tested, but it does not matter if the Java code is tested by the Python code.
In general, it is best to first open an issue and then, refer to the issue in your commit:
Your branch should have the format
issue-XYZ-branch-name where XYZ is
the issue number and branch-name is a short description.
Commits that refer to an issue will have the format
refs #XYZ -- message
where XYZ is the issue number.
Once a pull request is approved, we ask you to rebase your changes against the master branch and squash your commits into one meaningful commit (see below for tips on how to do this). The format of the commit would be:
fixes #XYZ -- short description below 72 characters Longer description that lists all the changes that occured on multiple lines of 79 characters.
To squash your commits, you can use the git rebase command:
# Squash the last three commits into 1 git rebase -i HEAD~3 # An editor will open. Change the word "pick" to "squash" # except for the very first commit at the top of the list. # After you save and exit, you will be prompted again to # change the commit message of the squashed commit.
Py4J does not have an official Contributor License Agreement (CLA), but it is assumed that as soon as you make a contribution (patch, code suggestion through any medium, pull requests) to Py4J, you accept that your code will be redistributed under the current license used by Py4J, i.e., the new BSD license. This is an irrevocable right to ensure that developers can use Py4J without the fear of seeing parts removed in the future.
You maintain the full copyrights for your contributions: you are only providing a license to distribute your code without further restrictions.
The copyright statement in the License has been standardized to:
Copyright (c) 2009-2016, Barthelemy Dagenais and individual contributors. All
Individual contributors are identified in the AUTHORS file. If you have contributed to Py4J and your name is not in AUTHORS, please open a pull request!
If you are working for a company while contributing to Py4J, make sure that the code is yours or that your company agrees with this implied CLA.
This approach is heavily inspired from the Django Contributor License Agreement..
If you have any question, do not hesitate to contact the founder of the project, Barthelemy.