Thanks for taking the time to contribute to Orange!
This document outlines our guidelines and standards of contributing to Orange. If anything is unclear, feel free to join our Discord server for a chat.
Please submit contributions in accordance with the flow explained in the GitHub Guides.
Install Orange as suggested in README. Then:
git clone https://github.com/biolab/orange3.git
cd orange3
python setup.py develop
When reporting bugs, please fill out the issue template to the best of your ability. At least, include a method to reproduce the bug (if consistently reproducible) and a screenshot (if applicable).
Roughly conform to PEP-8 style guide for Python code. Whenever PEP-8 is undefined, adhere to Google Python Style Guide.
In addition, we add the following guidelines:
-
Only ever
import *
to make objects available in another namespace, preferably in __init__.py. Everywhere else use explicit object imports. -
Use Napoleon-compatible (e.g. NumPy style) docstrings, preferably with tests.
-
When instantiating Qt widgets, pass static property values as keyword args to the constructor instead of calling separate property setters later. For example, do:
view = QListView(alternatingRowColors=True, selectionMode=QAbstractItemView.ExtendedSelection)
instead of:
view = QListView() view.setAlternatingRowColors(True) view.setSelectionMode(QAbstractItemView.ExtendedSelection)
-
Each Orange widget module, or better still, each Python module (within reason) should have a
__name__ == '__main__'
-fenced code block that shows/tests the gist of that module in a user-friendly way. -
Core library objects should represent (
__repr__
) themselves in accordance with the following statement from Python data model documentation:If at all possible, [the string returned by
__repr__
] should look like a valid Python expression that could be used to recreate an object with the same value (given an appropriate environment).To that end, use
Orange.util.Reprable
when possible.
Please ensure your commits pass code quality assurance by executing:
pip install -r requirements-dev.txt
python setup.py lint
For UI design, conform to the OS X Human Interface Guidelines. In a nutshell, use title case for titles, push buttons, menu titles and menu options. Elsewhere, use sentence case. Use title case for combo box options where the item is imperative (e.g. Initialize with Method) and sentence case otherwise.
If you contribute new code, write unit tests for it in Orange/tests or Orange/widgets/*/tests, as appropriate. Ensure the tests pass by running:
python setup.py test
Additionally, check that the tests for widgets pass:
python -m unittest -v Orange.tests \
Orange.widgets.tests
If testing on GNU/Linux, perhaps install xvfb package and prefix the above
command with xvfb-run
.
Prefer doctests for public APIs. Note, we unit-test doctests with
NORMALIZE_WHITESPACE
and ELLIPSIS
options enabled, so you can use them
implicitly.
Set these environment variables for value-added behavior:
ORANGE_DEBUG=1
- general developing and debugging. Influences stuff like DOM Inspector in QWebView right-click menu, etc.ORANGE_DEPRECATIONS_ERROR=1
- whether warnings of typeOrangeDeprecationWarning
should be raised as exceptions.
Make a separate commit for each logical change you introduce. We prefer short commit messages with descriptive titles. For a general format see Commit Guidelines. E.g.:
io: Fix reader for XYZ file format
The reader didn't work correctly in such-and-such case.
The commit title (first line) should concisely explain WHAT is the change. If the reasons for the change aren't reasonably obvious, also explain the WHY and HOW in the commit body.
The commit title should start with a tag which concisely conveys what Python package, module, or class the introduced change pertains to.
ProTip: Examine project's commit history to see examples of commit messages most probably acceptable to that project.
Implement new features in separate topic branches:
git checkout master
git checkout -b my-new-feature # spin a branch off of current branch
When you are asked to make changes to your pull request, and you add the commits that implement those changes, squash commits that fit together.
E.g., if your pull request looks like this:
d43ef09 Some feature I made
b803d26 reverts part of previous commit
77d5ad3 Some other bugfix
9e30343 Another new feature
1d5b3bc fix typo (in previous commit)
interactively rebase the commits onto the master branch:
git rebase --interactive master
and mark fixup
or squash
the commits that are just minor patches on
previous commits (interactive rebase also allows you to reword and reorder
commits). The resulting example pull request should look clean:
b432f18 some_module: Some feature I made
85d5a0a other.module: Some other bugfix
439e303 OWSomeWidget: Another new feature
Read more about squashing commits.
Documentation in located in doc folder. It is split into three parts: data-mining-library (scripting api), development (development guides), and visual-programming (widget help files). You can build it with:
cd doc/<part>
make html
# Now open build/html/index.html to see it