Developer Guide#
The hvPlot library is a project that provides a wide range of data interfaces and an extensible set of plotting backends, which means the development and testing process involves a broad set of libraries.
This guide describes how to install and configure development environments.
If you have any problems with the steps here, please reach out in the dev
channel on Discord or on Discourse.
TL;DR#
Open an issue on Github if needed
Fork and clone hvPlot’s Github repository
Install
pixi
Run
pixi run setup-dev
to create your development environmentMake some changes and run:
pixi run test-unit
if you updated the source code to run the unit testspixi run test-example
if you updated the notebooks to run thempixi run docs-build
if you need to build the website locally
Open a Pull Request
Preliminaries#
Basic understanding of how to contribute to Open Source#
If this is your first open-source contribution, please study one or more of the below resources.
Git#
The hvPlot source code is stored in a Git source control repository. The first step to working on hvPlot is to install Git onto your system. There are different ways to do this, depending on whether you use Windows, Mac, or Linux.
To install Git on any platform, refer to the Installing Git section of the Pro Git Book.
To contribute to hvPlot, you will also need Github account and knowledge of the fork and pull request workflow.
Pixi#
Developing all aspects of hvPlot requires a wide range of packages in different environments, but for new contributors the default
environment will be more than enough.
To make this more manageable, Pixi manages the developer experience. To install Pixi, follow this guide.
Glossary#
Tasks: A task is what can be run with
pixi run <task-name>
. Tasks can be anything from installing packages to running tests.Environments: An environment is a set of packages installed in a virtual environment. Each environment has a name; you can run tasks in a specific environment with the
-e
flag. For example,pixi run -e test-core test-unit
will run thetest-unit
task in thetest-core
environment.Lock-file: A lock-file is a file that contains all the information about the environments.
For more information, see the Pixi documentation.
Note
The first time you run pixi
, it will create a .pixi
directory in the source directory.
This directory will contain all the files needed for the virtual environments.
The .pixi
directory can be large, so it is advised not to put the source directory into a cloud-synced directory.
Installing the Project#
Cloning the Project#
The source code for the hvPlot project is hosted on GitHub. The first thing you need to do is clone the repository.
Run in your terminal:
git clone https://github.com/<Your Username Here>/hvplot
The instructions for cloning above created a hvplot
directory at your file system location.
This hvplot
directory is the source checkout for the remainder of this document, and your current working directory is this directory.
Start developing#
To start developing, run the following command, this will create an environment called default
(in .pixi/envs
), install hvPlot in editable mode, download test datasets, and install pre-commit
:
pixi run setup-dev
Note
The first time you run it, it will create a pixi.lock
file with information for all available environments.
This command will take a minute or so to run.
All available tasks can be found by running pixi task list
, the following sections will give a brief introduction to the most common tasks.
Developer Environment#
The default
environment is meant to provide all the tools needed to develop hvPlot.
This environment is created by running pixi run setup-dev
. Run pixi shell
to activate it; this is equivalent to source venv/bin/activate
in a Python virtual environment or conda activate
in a conda environment.
If you need to run a command directly instead of via pixi
, activate the environment and run the command (e.g. pixi shell
and pytest hvplot/tests/<somefile.py>
).
VS Code#
This environment can also be selected in your IDE. In VS Code, this can be done by running the command Python: Select Interpreter
and choosing {'default': Pixi}
.
To confirm you are using this dev environment, check the bottom right corner:
Jupyter Lab#
You can launch Jupyter lab with the default
environment with pixi run lab
. This can be advantageous when you need to edit the documentation or debug an example notebook.
Linting#
hvPlot uses pre-commit
to lint and format the source code. pre-commit
is installed automatically when running pixi run setup-dev
; it can also be installed with pixi run lint-install
.
pre-commit
runs all the linters when a commit is made locally. Linting can be forced to run for all the files with:
pixi run lint
Note
Alternatively, if you have pre-commit
installed elsewhere you can run:
pre-commit install # To install
pre-commit run --all-files # To run on all files
Testing#
To help keep hvPlot maintainable, all Pull Requests (PR) with code changes should typically be accompanied by relevant tests. While exceptions may be made for specific circumstances, the default assumption should be that a Pull Request without tests will not be merged.
There are three types of tasks and five environments related to tests.
Unit tests#
Unit tests are usually small tests executed with pytest. They can be found in hvplot/tests/
.
Unit tests can be run with the test-unit
task:
pixi run test-unit
Advanced usage
The task is available in the following environments: test-39
, test-310
, test-311
, test-312
, and test-core
. Where the first ones have the same environments except for different Python versions, and test-core
only has a core set of dependencies.
You can run the task in a specific environment with the -e
flag. For example, to run the test-unit
task in the test-39
environment, you can run:
pixi run -e test-39 test-unit
Advanced usage
Currently, an editable install needs to be run in each environment. So, if you want to install in the test-core
environment, you can add --environment
/ -e
to the command:
pixi run -e test-core install
Example tests#
hvPlot’s documentation consists mainly of Jupyter Notebooks. The example tests execute all the notebooks and fail if an error is raised. Example tests are possible thanks to nbval and can be found in the doc/
folder.
Example tests can be run with the following command:
pixi run test-example
Documentation#
The documentation can be built with the command:
pixi run docs-build
As hvPlot uses notebooks for much of the documentation, this takes a little while. You can disable:
Executing all the notebooks by setting the environment variable
HVPLOT_EXECUTE_NBS
tofalse
Building the gallery with
HVPLOT_REFERENCE_GALLERY="false"
Running the user guide notebooks with
HVPLOT_EXECUTE_NBS_USER_GUIDE="false"
Running the getting started notebooks with
HVPLOT_EXECUTE_NBS_GETTING_STARTED="false"
A development version of hvPlot can be found here. You can ask a maintainer if they want to make a dev release for your PR, but there is no guarantee they will say yes.
Link to hvPlot objects#
{meth}`hvplot.hvPlot.scatter`
{meth}`<obj>.scatter() <hvplot.hvPlot.scatter>`
Intersphinx#
The Sphinx Intersphinx extension allows linking to references in other projects that use this extension. For example:
Run this command to find all the references of the HoloViews site
python -m sphinx.ext.intersphinx https://holoviews.org/objects.inv
Extend
intersphinx_mapping
inconf.py
Link to the
Scatter
element with:
:class:`holoviews:holoviews.element.Scatter`
Build#
hvPlot has two build tasks to build a Python (for pypi.org) and a Conda package (for anaconda.org).
pixi run build-pip
pixi run build-conda
Continuous Integration#
Every push to the main
branch or any PR branch on GitHub automatically triggers a test build with GitHub Actions.
You can see the list of all current and previous builds at this URL
Etiquette#
GitHub Actions provides free build workers for open-source projects. A few considerations will help you be considerate of others needing these limited resources:
Run the tests locally before opening or pushing to an opened PR.
Group commits to meaningful chunks of work before pushing to GitHub (i.e., don’t push on every commit).