Collaborative Github Workflow¶
For developers interested in expanding pymatgen for their own purposes, we recommend forking pymatgen directly from the pymatgen GitHub repo. Here’s a typical workflow (adapted from http://www.eqqon.com/index.php/Collaborative_Github_Workflow):
Ignore the Github fork queue. Let the maintainer of pymatgen worry about the fork queue.
Create a free GitHub account (if you don’t already have one) and perform the necessary setup (e.g., install SSH keys etc.).
Fork the pymatgen GitHub repo, i.e., go to the main pymatgen GitHub repo and click fork to create a copy of the pymatgen code base on your own Github account.
Install git on your local machine (if you don’t already have it).
Clone your forked repo to your local machine. You will work mostly with your local repo and only publish changes when they are ready to be merged:
git clone firstname.lastname@example.org:YOURNAME/pymatgen.git
Note that the entire Github repo is fairly large because of the presence of test files, but these are absolutely necessary for rigorous testing of the code.
It is highly recommended you install all the optional dependencies as well.
Code (see Coding Guidelines). Commit early and commit often. Keep your code up to date. You need to add the main repository to the list of your remotes. Let’s name the upstream repo as mpmaster (materialsproject master).
git remote add mpmaster git://github.com/materialsproject/pymatgen.git
Make sure your repository is clean (no uncommitted changes) and is currently on the master branch. If not, commit or stash any changes and switch to the master.
git checkout master
Then you can pull all the new commits from the main line
git pull mpmaster master
Remember, pull is a combination of the commands fetch and merge, so there may be merge conflicts to be manually resolved.
Publish your contributions. Assuming that you now have a couple of commits that you would like to contribute to the main repository. Please follow the following steps:
If your change is based on a relatively old state of the main repository, then you should probably bring your repository up-to-date first to see if the change is not creating any merge conflicts.
Check that everything compiles cleanly and passes all tests. The pymatgen repo comes with a complete set of tests for all modules. If you have written new modules or methods, you must write tests for the new code as well (see Coding Guidelines). Install and run nosetest in your local repo directory and fix all errors before continuing further. There must be no errors for the nosetest.
If everything is ok, publish the commits to your github repository.
git push origin master
Now that your commit is published, it doesn’t mean that it has already been merged into the main repository. You should issue a merge request to pymatgen’ maintainers. They will pull your commits and run their own tests before releasing.
Given that pymatgen is intended to be long-term code base, we adopt very strict quality control and coding guidelines for all contributions to pymatgen. The following must be satisfied for your contributions to be accepted into pymatgen.
- Unittests are required for all new modules and methods. The only way to minimize code regression is to ensure that all code are well-tested. If the maintainer cannot test your code, the contribution will be rejected.
- Python PEP 8 code style. We allow a few exceptions when they are well-justified (e.g., Element’s atomic number is given a variable name of capital Z, in line with accepted scientific convention), but generally, PEP 8 must be observed.
- Python 3. All code should seamless work with Python 2.7 and Python 3.x. Please read Python’s official guidelines on how to write Python 3.x compatible code, including the usage of the “six” package. It is recommended that you install the “python-modernize” package and run it before submitting any pull requests.
- Documentation required for all modules, classes and methods. In particular, the method docstrings should make clear the arguments expected and the return values. For complex algorithms (e.g., an Ewald summation), a summary of the alogirthm should be provided, and preferably with a link to a publication outlining the method in detail.
- IDE. We highly recommend the use of Pycharm or Eclipse + PyDev. You should also set up pylint and pep8 and turn those on within the IDE setup. This will warn of any issues with coding styles.
For the above, if in doubt, please refer to the core classes in pymatgen for examples of what is expected.
A word on coding for Python 3 compatibility¶
With effect from version 3.0, all pymatgen code must be both Python 2.7+ and 3 compatible. Specifically, we have adopted the following practices throughout pymatgen.
Unicode-always. Unless you are absolutely sure you need byte literals (rare for pymatgen), always use unicode. In particular, the following should be the first line of all pymatgen modules:
from __future__ import division, print_function, unicode_literals
Future division means that 1/2 returns a float (0.5), which is more intuitive scientifically, instead of 0 (default integer division in Python 2). print_function ensures that print() is used instead of the print statement. And unicode_literals makes it such that all strings are treated as unicode by default. If you need to use bytes, those should be marked up explicitly as b’byte literal’.
Use of the six package. Where necessary, use the six package to handle interoperability between Python 2 and 3. Examples include the six.moves functions (common ones are zip, filter, map), and six.stringtypes (testing for string types, which should be rarely done).
Python-modernize. Use python-modernize to check your code for any potential changes that need to be made.
Unit testing. The entire pymatgen code base is continuously being tested in both Python 2.7 and >=3.3. If your code fails either of the tests, you need to fix it.