Developer guide

Pyranges was originally written by Endre Bakken Stovner under supervision of Pål Sætrom. It is now mainly developed by Endre Bakken Stovner and by the Comparative Genomics lab of Marco Mariotti. It follows the guidelines for open source software, and external contributors are welcome. The code is centralized on github, at https://github.com/pyranges/pyranges1

Bugs and feature requests can be reported as github issues. You may also contribute by submitting your own code edits, either to deal with issues or to add new functionalities. Please discuss new features with the team before making a PR. The code will be reviewed by the core development team, which may integrate it in the main repository. Contributions are tracked by github and are publicly visible.

Below, we sketch a guide to contribute to Pyranges. It assumes familiarity with python and with the terminal, and minimal experience with git/github. Before the actual list of steps follow (Task sequence), we go over some essential concepts used in the “continuous integration” system in place to maintain and evolve Pyranges.

Tests

Tests are an essential part of continuous integration. Briefly, they ensure that code edits do not break existing functions. Various layers of tests are implemented in Pyranges:

  • unit tests: quick and compulsory tests about the main Pyranges functionalities

  • doctest: quick and compulsory tests that ensures that the code in the documentation (docstrings examples, tutorial and how-to-pages) gives the expected results

If the code submitted to Pyranges does not pass the compulsory tests, it will not be integrated. Therefore, we highly recommend developers to run tests before code submissions, as explained further below.

Documentation: docstrings

Python docstrings are widely used to document the rationale, input arguments, and returned values of all functions and methods. The use of a consistent docstring style allows the automatic generation of API documentation, as seen in Pyranges documentation at https://pyranges1.readthedocs.io/, built through the Sphynx software.

Pyranges adopts the NumPy/SciPy-style: https://numpydoc.readthedocs.io/en/latest/format.html. It is important that code contributors who edit any function also update their docstrings to reflect how it works; and that all new functions contain an appropriate docstring. Follow the link above and inspect existing Pyranges code to write good docstrings.

Code formatting and linting

Pyranges code follows strict guidelines about its formatting and non-redundancy. This burden is not upon the developer: instead, this is achieved by running dedicated software that polishes, formats, and “lints” the code before its integration in the main repository. The tool ruff is used, for both code linting and formatting.

Type checking

Pyranges is strict about type checking to ensure code reliability and maintainability. Type checking involves specifying and verifying the data types of variables and functions, a practice that may be unfamiliar to some Python developers.

These annotations allow tools like pyright, which we use extensively, to check types. Consequently, if add is called with non-integer arguments or if it returns a non-integer value, the type checker flags an error. This preemptive error detection, happening at compile time (before the code is run), significantly reduces runtime issues and enhances code quality.

For example, consider this basic function:

def add(a, b):
    return a + b

With type annotations, it becomes:

def add(a: int, b: int) -> int:
    return a + b

For more detailed guidance on typing in Python, see the official Python documentation.

We encourage contributions to Pyranges, even if they involve partial typing. If you’re new to typing or have any questions, feel free to ask for help. We’re committed to supporting our community in enhancing Pyranges together.

Task sequence

1. Create and download your own Pyranges fork

The easiest way to do this is through github. Login into the github website if you aren’t already, then visit the Pyranges page on github, click “Fork” on the top right. Fill the form and confirm. In the page of your new fork, find the <> Code button, and copy the https address. On your computer, create a new folder dedicated to the project, then clone your fork inside it:

mkdir mypyranges1
cd mypyranges1
git clone PASTE_GITHUB_HTTPS

2. Set up your developer environment and install Pyranges

We recommend creating an environment dedicated to the development of pyranges1:

conda create -n prdev python pip
conda activate prdev

Next, let’s install pyranges1 and its dependencies: cd into your pyranges1 fork, and install it locally with pip as shown below. By using pip option -e, your installation is in “editable” mode: any changes you make to your pyranges1 code will be immediately reflected in your environment. In other words, you won’t need to re-run pip install every time you change something in the code.

cd pyranges1
pip install -e .

Next, let’s install optional dependencies (necessary to run certain functions and thus required to pass tests) and all developer dependencies (necessary to run tests, lint code etc). Their list, like most of Pyranges configuration, is found in the pyproject.toml file. You can install all you need with:

pip install -e '.[all]'

3. Edit the code

Now, you’re ready to edit the code in the pyranges1/ folder.

To run your code to see that it behaves as intended, we recommend using a separate script that imports pyranges1, making sure you’re in the prdev conda environment.

4. Run tests

For each of the test commands, inspect their output: warnings are acceptable, but errors must be fixed. At the time of writing, the test commands are these (omitting formatting, linting, and type-checking tests by ruff and pyright):

python tests/run_doctest_tutorial_howto.py
pytest --doctest-modules pyranges1
pytest tests/unit

The up-to-date test commands can be found in the tox section of the repo pyproject.toml configuration file.

5. Format, lint, type-checking code

Next, let’s format code with ruff:

ruff format pyranges1

Then, let’s lint code, also with ruff:

ruff check pyranges1

If the ruff check above shows any error, you must fix them before you proceed. If errors are deemed ‘fixable’, you may simply run ruff check --fix pyranges1. If not, you must delve into the code. Note, ChatGPT/Copilot are your friends!

After fixing any errors, the corresponding tests (see pyproject.toml file content) should pass.

Lastly, let’s use pyright to ensure correct type-checking:

pyright

Again, any error in the pyright must be amended before proceeding. Note that if you edit the code, you may want to format and lint code again with ruff.

6. Test on all supported python and package versions

Next, we use tox to test whether the code works across all the versions of python and main dependencies that we intend to support. This step internally runs steps 4 and 5 for every such version defined in pyproject.toml. (Advanced users may actually directly run 6 instead of 4-5). For this, run:

tox

If any errors emerge, correct them (or ask us for help).

7. Inspect the Sphynx documentation

Your code edits may warrant edits in the Pyranges docstrings. In this case, it is compelling to locally check that the automatically generated documentation is built appropriately. Inside the pyranges1/ folder, run these commands:

cd docs
make html
cd -

If the “make” command has no major errors, it will generate the full pyranges1 documentation in the form of html pages, identical to https://pyranges1.readthedocs.io/. Open the file docs/build/html/index.html with a browser to inspect all the parts that may have been affected by your changes, and fix any potential problems. To know more about its inner workings, read about the Sphynx system.

8. Log your changes

At this stage, you are ready to submit your code for integration into the main Pyranges repository; that is to say, to open a “pull request”. Before you can do that, you have to update your remote repository, i.e. your Pyranges fork at github.

First, bump the version number in the file pyproject.toml.

Then, it’s essential to document your changes in the CHANGE_LOG.txt file. This log should provide a clear and concise summary of the modifications, additions, and fixes made in each version of your project. Include relevant details such as feature enhancements, bug fixes, and any other notable alterations to help maintain a transparent and informative record of your project’s evolution.

Also, edit the README.md file to add the most recent changes in the corresponding section.

9. Commit and push to your remote repository

Run this command to list all the local files you modified:

git status

You must tell git which of these files have to be synchronized, i.e. “git add” them. You can do this by explicitly providing the list of files with:

git add file1 file2 ... fileN

Alternatively to the previous command, if you want to add ALL edited files, you can use:

git add . --dry-run

to check the list of all modified files, then this to actually add them:

git add .

After adding files, you have to commit your changes locally with:

git commit -m"Include an informative message here"

Finally, push to your remote repository, i.e. update your online fork at github, with:

git push

You will be requested your github credentials. Note that your online password may not work; in this case, google how to set up a github token that you can use.

10. Open a pull request

The easiest way to open a pull request is through the github website. Go to your Pyranges fork on github, then find the “Contribute” button (near the <> Code button). Click it, and select Open pull request.

In the newly opened page, carefully check that source and destination are correctly selected. The Base repository should be pyranges1/pyranges1 (i.e. the main pyranges1 repo), and the Head repository should be your fork. If you worked on non-master git branches, select them here.

In the comments, write a summary of the introduced changes and their rationale, tagging any related github issues (i.e. paste their http address). On the rest of the page, you are presented with a list of the code edits. When you’re ready, click “Open pull request”.

Github will run a “check” workflow which basically replicates the steps above. If all checks are ok, Pyranges administrators will inspect the pull request, comment it if necessary, and potentially accept it.

11. Core team only: upload to PyPI

Every now and then, the core development team considers that a new pyranges1 version should be released. To do so:

  • Update the version number in the pyproject.toml file

  • Find the “Build and upload to PyPI” workflow in the left menu of the github actions at https://github.com/pyranges/pyranges1/actions/

  • Click the “Run workflow” button on the right

Next, check that everything worked correctly, by confirming that a new pyranges1 installation via pip selects the new version.

Finally, the pyranges1 conda package at Bioconda is updated automatically upon pip upload. Check that this is updated correctly.