Developers need to memorize a whole bunch of magic command-line incantations.
These incantations may also change over time.
Often, Makefiles are used to provide aliases, but Makefiles can be convoluted, are not written in Python, and are hard to extend.
The goal of spin
is therefore to provide a simple, user-friendly, extendable interface for common development tasks.
It comes with a few common build commands out the box, but can easily be customized per project.
As a curiosity: the impetus behind developing the tool was the mass migration of scientific Python libraries (SciPy, scikit-image, and NumPy, etc.) to Meson, after distutils was deprecated. When many of the build and installation commands changed, it made sense to abstract away the nuisance of having to re-learn them.
Note: We now have experimental builds for editable installs. Most of the Meson commands listed below should work "out of the box" for those.
pip install spin
Settings are stored in .spin.toml
, spin.toml
, or your project's pyproject.toml
.
As an example, see the [tool.spin]
section of an example pyproject.toml
.
The [project]
section should contain name
.
The [tool.spin]
section should contain:
package = "pkg_importname" # name of your package
commands = [
"spin.cmds.meson.build",
"spin.cmds.meson.test"
]
See the command selection below.
Once you have several commands, it may be useful to organize them into sections.
In pyproject.toml
, instead of specifying the commands as a list, use the following structure:
[tool.spin.commands]
"Build" = [
"spin.cmds.meson.build",
"spin.cmds.meson.test"
]
"Environments" = [
"spin.cmds.meson.ipython",
"spin.cmds.meson.run"
]
These commands will then be rendered as:
Build:
build π§ Build package with Meson/ninja
test π§ Run tests
Environments:
ipython π» Launch IPython shell with PYTHONPATH set
run π Run a shell command with PYTHONPATH set
spin
or
python -m spin
Available as spin.cmds.meson.*
.
build π§ Build package with Meson/ninja
ipython π» Launch IPython shell with PYTHONPATH set
python π Launch Python shell with PYTHONPATH set
shell π» Launch shell with PYTHONPATH set
test π§ Run pytest
run π Run a shell command with PYTHONPATH set
docs π Build Sphinx documentation
gdb πΎ Execute a Python snippet with GDB
lldb πΎ Execute a Python snippet with LLDB
Build (PEP 517 builder)
Available as spin.cmds.build.*
:
sdist π¦ Build a source distribution in `dist/`
pip (Package Installer for Python)
pip
allows for editable installs, another common
development workflow.
Available as spin.cmds.pip.*
:
install π½ Build and install package using pip.
Available as spin.cmds.meta.*
:
introspect π Print a command's location and source code
spin
can invoke custom commands. These commands define their own arguments, and have access to the pyproject.toml
file for further configuration.
See, e.g., the example custom command.
Add custom commands to the commands
variable in the [tool.spin]
section of pyproject.toml
as follows:
commands = [..., '.spin/cmds.py:example']
Here, the command is stored in .spin/cmds.py
, and the function
is named example
.
Custom commands can access the pyproject.toml
as follows:
from spin import util
@click.command()
def example():
"""Command that accesses `pyproject.toml` configuration"""
config = util.get_config()
print(config["tool.spin"])
Default arguments can be overridden for any command. The custom command above, e.g., has the following signature:
@click.command()
@click.option("-f", "--flag")
@click.option("-t", "--test", default="not set")
def example(flag, test, default_kwd=None):
"""π§ͺ Example custom command.
...
"""
Use the [tool.spin.kwargs]
section to override default values for
click options or function keywords:
[tool.spin.kwargs]
".spin/cmds.py:example" = {"test" = "default override", "default_kwd" = 3}
Instead of rewriting a command from scratch, a project may want to add a flag to a built-in spin
command, or perhaps do some pre- or post-processing.
For this, we have to use an internal Click concept called a context.
Fortunately, we don't need to know anything about contexts other than that they allow us to execute commands within commands.
We proceed by duplicating the function header of the existing command, and adding our own flag:
from spin.cmds import meson
# Take this from the built-in implementation, in `spin.cmds.meson.build`:
@click.command()
@click.argument("meson_args", nargs=-1)
@click.option("-j", "--jobs", help="Number of parallel tasks to launch", type=int)
@click.option("--clean", is_flag=True, help="Clean build directory before build")
@click.option(
"-v", "--verbose", is_flag=True, help="Print all build output, even installation"
)
# This is our new option
@click.option("--custom-arg/--no-custom-arg")
# This tells spin that we will need a context, which we
# can use to invoke the built-in command
@click.pass_context
# This is the original function signature, plus our new flag
def build(ctx, meson_args, jobs=None, clean=False, verbose=False, custom_arg=False):
"""Docstring goes here. You may want to copy and customize the original."""
# Do something with the new option
print("The value of custom arg is:", custom_arg)
# The spin `build` command doesn't know anything about `custom_arg`,
# so don't send it on.
del ctx.params["custom_arg"]
# Call the built-in `build` command, passing along
# all arguments and options.
ctx.forward(meson.build)
# Also see:
# - https://click.palletsprojects.com/en/8.1.x/api/#click.Context.forward
# - https://click.palletsprojects.com/en/8.1.x/api/#click.Context.invoke
Some packages use a vendored version of Meson. The path to a custom
Meson CLI can be set in pyproject.toml
:
[tool.spin.meson]
cli = 'path/to/custom/meson'
- Running
spin
, the emojis in the command list don't show up.
Your terminal font may not include emoji characters. E.g., if you use noto on Arch Linux the emojis are installed separately:
sudo pacman -S noto-fonts-emoji
fc-cache -f -v
spin
development happens on GitHub at scientific-python/spin.
spin
tests are invoked using:
nox -s test
Other examples:
nox -s test -- -v
nox -s test -- -v spin/tests/test_meson.py
The dev.py
tool was proposed for SciPy by Ralf Gommers and implemented by Sayantika Banik, Eduardo Naufel Schettino, and Ralf Gommers (also see Sayantika's blog post).
Inspired by that implementation, spin
(this package) is a minimal rewrite by StΓ©fan van der Walt, that aims to be easily extendable so that it can be used across ecosystem libraries.
We thank Danila Bredikhin and Luca Marconato who kindly donated the spin
name on PyPi.