Start a new project¤
Install Cookiecutter¤
To start a new project, you should first install
Cookiecutter.
It is normally available in any Linux distribution, but some have a very old
version (for example, Ubuntu/Debian). You can check which version your distro
has on Repology. You need
at least version 2.4.0. To ensure you get an up-to-date version, you can
always use pip
and install it in a venv
:
$ python -m venv cookiecutter
$ cd cookiecutter
$ . bin/activate
[cookiecutter] $ pip install cookiecutter
Collecting cookiecutter
...
Then simply run [Cookiecutter] where you want to create the new project:
Generate the project files¤
cookiecutter gh:frequenz-floss/frequenz-repo-config-python \
--directory=cookiecutter --checkout v0.9.2
This command will prompt you for the project type, name, and other configuration options, and it will generate the entire project for you in a new subdirectory.
Warning
This command needs to be typed literally!
frequenz-floss/frequenz-repo-config-python
is the GitHub repository with
the cookiecutter template that will be downloaded, and
--directory=cookiecutter
is needed because the cookiecutter template
doesn't live at the top-level of that repository, but in a subdirectory
called cookiecutter
. The --checkout
option is provided to use
a template from a released version, otherwise the default (development)
branch will be used.
All information about your project will be prompted interactively by that command.
After completing the project and fixing the TODO
s, you can either amend the
previous commit using git commit --amend
or create a new commit for the
changes using git commit
.
Template variables reference¤
-
type
: The type of repository. It must be chosen from the list. -
name
: The name of the project. This will be used to build defaults for other inputs, such astitle
,python_package
, etc. It should be one word, using only alphanumeric characters (and starting with a letter). It can include also_
and-
which will be handled differently when building other variables from it (replaced by spaces in titles for example). -
description
: A short description of the project. It will be used as the description in theREADME.md
,pyproject.toml
,mkdocs.yml
, etc. -
title
: A human-readable name or title for the project. It will be used in theREADME.md
,CONTRIBUTING.md
, and other files to refer to the project, as well as the site title inmkdocs.yml
. -
keywords
: A comma-separated list of keywords that will be used in thepyproject.toml
file. If left untouched, it will use only some predefined keywords. If anything else is entered, it will be added to the default keywords. -
github_org
: The GitHub handle of the organization where the project will reside. This will be used to generate links to the project on GitHub. -
license
: Currently, only two options are provided:MIT
, which should be used for open-source projects, andProprietary
, which should be used for closed-source projects. This will be added to file headers and used as the license inpyproject.toml
. -
author_name
,author_email
: The name and email address of the author of the project. They will be used in the copyright notice in file headers and as the author inpyproject.toml
. -
python_package
: The Python package in which this project will reside. All files provided by this project should be located in this package. This needs to be a list of valid Python identifiers separated by dots. The source file structure will be derived from this. For example,frequenz.actor.example
will generate files insrc/frequenz/actor/example
. -
pypi_package_name
: The name of the PyPI/wheel/distribution package. This should be consistent with thepython_package
, usually replacing.
with-
. For example,frequenz-actor-example
. -
github_repo_name
: The handle of the GitHub repository where the project will reside. This will be used to generate links to the project on GitHub and as the top-level directory name. -
default_codeowners
: A space-separated list of GitHub teams (@org/team
) or users (@user
) that will be the default code owners for this project. This will be used to build theCODEOWNERS
file. Please refer to the code owners documentation for more details on the valid syntax.
Create the local environment¤
To start development, you need to make sure your environment is correctly set up. One way to do this is by using a virtual environment and installing all the dependencies there:
# requires at least Python version 3.11
python3 -m venv .venv
. .venv/bin/activate
pip install -e .[dev]
This will install your package in editable
mode, so
you can open a Python interpreter and import your package modules, picking up
any local changes without the need to reinstall. Now you can run tools
directly, like pytest
.
Verify using nox
¤
If you prefer to keep your virtual environment cleaner and avoid installing
development dependencies, you can also use nox
to create isolated
environments:
pip install -e .[dev-noxfile]
nox --install-only # Set up virtual environments once
nox -R # Run linting and testing reusing the existing virtual environments
This will only install your package in editable mode and the minimum
dependencies required to run nox
. It will then run all nox
default
sessions, which include running linters and tests.
Tip
It's much faster to use nox
with --install-only
once (each time you
change or update dependencies, you need to run it again) and then use nox
-R
to run the sessions without recreating the virtual environments.
Otherwise, nox
will create many virtual environments each time you run
it, which is very slow.
Verify the documentation works¤
To generate the documentation, you can use mkdocs
:
If the command fails, look at the log warnings and errors and fix them. If it worked, now there is a local web server serving the documentation. You can point your browser to http://127.0.0.1:8000 to have a look.
Info
For API projects, docker
is needed to generate and serve documentation,
as the easiest way to use the tool to generate the documentation from
.proto
files is using
docker
.
Initialize GitHub Pages¤
The generated documentation can be easily published via GitHub Pages, and it will be automatically updated for new pushes and releases. However, some initial setup is needed for it to work correctly:
pip install -e .[dev-mkdocs] # Not necessary if you already installed .[dev]
# Creates the branch gh-pages locally
mike deploy --update-aliases v0.1-dev v0-dev latest-dev latest
mike set-default latest # Makes the latest alias the default version
git push upstream gh-pages # Pushes the new branch upstream to publish the website
This assumes your branch is called v0.x.x
and your first release will be v0.1.0
.
Then make sure that GitHub Pages is enabled in
https://github.com/<repo-owner>/<repo-name>/settings/pages
.
If all went well, your website should be available soon via
https://<repo-owner>.github.io/<repo-name>/
.
Versions using Mike¤
The above commands create a new documentation version using Mike, which is used to keep multiple versions of the website.
The new documentation version is called v0.1-dev
, which is used as the name
for the currently in-development branch. The v0.1-dev
branch has an alias
called v0-dev
that points to the latest v0.x branch development version, an
alias latest-dev
that points to the latest in-development version, and
a latest
alias, which is set as the default.
If the website is visited without specifying an explicit version, the latest
version will be displayed. It is recommended to point latest
to the latest
stable version instead of the currently in-devepoment version, so as soon as
you make a release, you should update the alias.
Automatic generation¤
New versions of the documentation will be automatically generated and published via GitHub Actions on any push.
If a push is to a branch with semver-like format (vX.Y.Z), then the generated
version will replace the current vX.Y-dev
version (and vX-dev
and
latest-dev
alias).
If a tag is pushed instead, then the generated version will replace the current
vX.Y
version (if there was any), and the aliases vX
and latest
will be
updated to point to the new version.
To summarize, we don't create versions for patch releases, only for minor versions, and we have an alias to point to the latest version of a major version too.