setuptools-conda
Build a conda package from a setuptools project.
Table of Contents
- Installation and usage
- build and install setuptools-conda from source
- Help text of setuptools-conda
- Help text of setuptools-conda build command
- Help text of setuptools-conda install-requirements command
- Help text of python setup.py dist_conda distutils command
Installation and usage
To install:
$ conda install -c labscript-suite setuptools-condaTo make a conda package: in your project directory, run:
$ setuptools-conda build .The resulting conda package will be in conda_packages/<architecture>/ and can be
installed with conda install conda_packages/<architecture>/<pkgfile> or uploaded to
your account on anaconda.org with (you'll need to install anaconda-client):
$ anaconda upload conda_packages/<architecture>/<pkgfile>setuptools-conda build installs the project's build dependencies, as declared in a
pyproject.toml, setup.cfg or setup.py, and then runs python setup.py dist_conda,
where dist_conda is a distutils command added by setuptools-conda. You can also run
python setup.py dist_conda yourself. See below for the full documentation of the
setuptools-conda build command and the dist_conda command to setup.py.
setuptools-conda also provides a command setuptools-conda install-requirements that
will install a project or projects build and runtime requirements with conda. This
allows one to then make an editable install with pip install --no-deps -e . whilst
ensuring all of a project's dependencies are without pip installing missing
dependencies from PyPI into a conda environment, which can be problematic.
setuptools-conda aims to be as comprehensive as possible, allowing you to generate
conda packages with setuptools-conda build or python setup.py dist_conda, with as
much information as possible determined automatically. More powerful than
conda-build's bdist_conda - this package can build for multiple Python versions,
allows you to specify any package names that differ between conda and PyPI, and converts
a wider range of environment markers in dependencies such as:
INSTALL_REQUIRES = [
"pyzmq >=18.0",
"ipaddress; python_version == '2.7'",
"subprocess32; python_version == '2.7'",
"enum34; python_version == '2.7'",
"pywin32; sys_platform == 'win32'",
"windows-curses; sys_platform == 'win32'",
]
In most cases, you should not need to modify a project's code at all to use
setuptools-conda, though if it's your project you're building, you may wish to add
configurations settings to setup.cfg in order to avoid passing them all as command
line arguments at build time.
You may customise dist_conda to build for multiple Python versions at once, or many
other options - see the full list of options in the help text of the dist_conda
command below. Options may be passed in on the command line, or in a setup.cfg file in
the [dist_conda] section:
[dist_conda]
pythons = 3.6, 3.7
noarch = TrueThey can also be set if you need to with the command_options argument in setup.py:
setup(
...
command_options = (
{
'dist_conda': {
'pythons': (__file__, '3.6, 3.7'),
'noarch': (__file__, True),
}
},
)
)build and install setuptools-conda from source
setuptools-conda can build a conda package of itself. To make a conda package for your
Python version and platform, in an activated conda environment, run the following:
$ git clone https://github.com/chrisjbillington/setuptools-conda
$ cd setuptools-conda/
$ python -m setuptools_conda build .
$ conda install conda_packages/<your_architecture>/setuptools-conda-<version>-<build>.tar.bz2Help text of setuptools-conda
$ python setuptools-conda -h
usage: setuptools-conda [-h] {build,install-requirements} ...
positional arguments:
{build,install-requirements}
Action to perform, either "build" or "install-requirements". For
help on arguments accepted by a given command, run
'setuptools-conda <command> -h'
build Build a conda package from a setuptools project.
Installs the build requirements of the project with conda, and
then runs 'python setup.py dist_conda', passing remaining
arguments. This is similar to 'pip wheel' etc in the presence of
a pyproject.toml file (but without build isolation - the
dependencies will be installed in the current environment). A
typical way to call this script would be as 'setuptools-conda
build [args] .' from the working directory of a project, where
'[args]' are any additional arguments you want to pass to the
'dist_conda' command. See 'python setup.py dist_conda -h' for a
full list of accepted arguments.
Build requirements are searched for in the places in order,
stopping on the first found:
1. --setup-requires passed in on the command line
2. [dist_conda]/setup_requires in the project's setup.cfg
3. [build-system]/requires in the project's pyproject.toml
4. [options]/setup_requires in the project's setup.cfg
This the same way the 'dist_conda' command gets build
dependencies.
Additional conda channels to enable to install the build
requirements can be passed in with the '--channels' argument or
set in [dist_conda]/channels in setup.cfg, any any PyPI:conda
name differences can be passed in with the
'--conda-name-differences' argument or configured in
[dist_conda]/conda_name_differences in setup.cfg.
install-requirements
Install the requirements of the given project(s). This will
install both the build and runtime requirements of all packages
given. Build dependencies are determined from the same sources
as the 'build' command.
Runtime requirements are searched for in the places in order,
stopping on the first found:
1. --install-requires passed in on the command line
2. [dist_conda]/install_requires in the project's setup.cfg
4. [options]/install_requires in the project's setup.cfg or
setup.py (obtained via 'python setup.py egg_info')
any any PyPI:conda name differences can be passed in with the
'--conda-name-differences' argument or configured in
[dist_conda]/conda_name_differences in setup.cfg.
Any runtime requirements that themselves are in the list of
projects to install requirements for will not be installed. This
is intended to facilitate running `pip install -e --no-deps` to
create editable installs for a set of projects, for which one
would not want to install those projects normally in addition to
in editable mode.
optional arguments:
-h, --help show this help message and exit
Help text of setuptools-conda build command
$ python setuptools-conda build -h
usage: setuptools-conda build [-h] [setup_args [setup_args ...]] project_path
positional arguments:
setup_args Arguments to pass to setup.py as 'python setup.py dist_conda
[setup_args]'; e.g. '--noarch'
project_path Path to project; e.g. '.' if the project's `setup.py`,
`pyproject.toml` or `setup.cfg` are in the current working
directory.
optional arguments:
-h, --help show this help message and exit
Help text of setuptools-conda install-requirements command
$ python setuptools-conda install-requirements -h
usage: setuptools-conda install-requirements [-h] [--setup-requires SETUP_REQUIRES]
[--install-requires INSTALL_REQUIRES]
[--conda-name-differences CONDA_NAME_DIFFERENCES]
[--channels CHANNELS]
projects [projects ...]
positional arguments:
projects Project directories to install dependencies for
optional arguments:
-h, --help show this help message and exit
--setup-requires SETUP_REQUIRES
Build requirements override. 'See python setup.py dist_conda -h'
--install-requires INSTALL_REQUIRES
Install requirements override. 'See python setup.py dist_conda
-h'
--conda-name-differences CONDA_NAME_DIFFERENCES
PyPI:conda name differences override. 'See python setup.py
dist_conda -h'"
--channels CHANNELS Channels to search for build requires. 'See python setup.py
dist_conda -h'
Help text of python setup.py dist_conda distutils command
$ python setup.py dist_conda -h
<snip>
Options for 'dist_conda' command:
--pythons Minor Python versions to build for, as a comma-
separated list e.g. '2.7, 3.6'. Also accepts a
list of strings if passed into `setup()` via
`command_options`. Defaults to the current Python
version.
--build-number (-n) Conda build number. Defaults to zero
--license-file (-l) License file to include in the conda package.
Defaults to any file in the working directory
named 'LICENSE', 'COPYING', or 'COPYRIGHT', case
insensitive and ignoring extensions. Set to 'None'
to not include a license file even if one of the
above is present.
--build-string (-s) Conda build string.
--setup-requires Build dependencies, as a comma-separated list in
standard setuptools format, e.g. 'foo >= 2.0;
sys_platform=="win32",bar==2.3'. Also accepts a
list of strings if passed into `setup()` via
`command_options`. Defaults to any requirements
listed in a `pyproject.toml` under [build-
system]/requires, or if none, any requirements
listed in the `setup_requires` setuptools
configuration option. Can be be omitted if the
build dependencies when building for conda do not
differ.
--install-requires Runtime dependencies, as a comma-separated list in
standard setuptools format, e.g. 'foo >= 2.0;
sys_platform=="win32",bar==2.3'. Also accepts a
list of strings if passed into `setup()` via
`command_options`. Defaults to the
`install_requires` argument to `setup()`, and can
therefore be omitted if the runtime dependencies
when running in conda do not differ.
--ignore-run-exports Comma-separated list of conda packages that should
*not* be considered runtime dependencies, even if
they are declared in run_exports of a build
dependency. run_exports declared by build
dependencies are normally automatically considered
run dependencies, for example libraries that were
linked against at build-time - but this can be
undesirable when it creates a brittle dependency
on a specific version of a library which is not
actually required at runtime. Also accepts a list
of strings if passed into `setup()` via
`command_options`.
--channels (-c) Additional channels to search for build
requirements during the build, as a comma-
separated list, or a list of strings if passed in
via setup.py. Defaults to [tools.setuptools-
conda]/channels listed in a `pyproject.toml` file,
if any.
--conda-name-differences Mapping of PyPI package names to conda package
names, as a comma-separated list of colon-
separated names, e.g.
'PyQt5:pyqt,beautifulsoup4:beautiful-soup'. Also
accepts a dict if passed into `setup()` via
`command_options`. Conda packages usually share a
name with their PyPI equivalents, but use this
option to specify the mapping when they differ. If
the only difference is lowercasing or conversion
of underscores into hyphens, no entry is needed -
these changes are made automatically.
--link-scripts Comma-separated list of link scripts to include,
such as post-link.sh, pre-unlink.bat etc. These
will be placed in the recipe directory before
building. If passed to `setup()` via
`command_options`, this shound instead be a
dictionary mapping link script filenames to their
contents.
--noarch Build a platform-independent package. Only set
this if your dependencies are the same on all
platforms and Python versions you support, and you
have no compiled extensions.
--from-wheel Whether to build a wheel before invoking conda-
build. By default setuptools-conda invokes conda-
build on an sdist such that any compilation of
extensions will be done in the conda build
environment. However, if your extensions are not
able to be compiled with conda's compiler
configuration, you might set this option to pass
conda-build a wheel that has been pre-compiled
with the system configuration. In this case,
setuptools-conda will only produce a conda package
for the current Python version.
--from-downloaded-wheel Whether to avoid local building at all and
download a wheel from PyPI before invoking conda-
build. For projects with tricky build environment
requirements, this can be a way to essentially
repackage an existing wheel without having to any
building at all. Requires that the exact version
as understood by setuptools is availalble on PyPI
as a wheel. In this case, setuptools-conda will
only produce a conda package for the current
Python version.
--croot Value of --croot to pass to conda-build, used as
its build directory. Defaults to
./conda_build/conda-bld. Setting this to a very
short path can be useful on Windows, where conda-
build sometimes chokes on very long filepaths.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent