Installation

Installing astropy

If you are new to Python and/or do not have familiarity with Python virtual environments, then we recommend starting by installing the Anaconda Distribution. This works on all platforms (linux, Mac, Windows) and installs a full-featured scientific Python in a user directory without requiring root permissions.

Using pip

Warning

Users of the Anaconda Python distribution should follow the instructions for Using Conda.

To install astropy with pip, run:

pip install astropy

If you want to make sure none of your existing dependencies get upgraded, you can also do:

pip install astropy --no-deps

On the other hand, if you want to install astropy along with recommended or even all of the available optional dependencies, you can do:

pip install astropy[recommended]

or:

pip install astropy[all]

In most cases, this will install a pre-compiled version (called a wheel) of astropy, but if you are using a very recent version of Python, if a new version of astropy has just been released, or if you are building astropy for a platform that is not common, astropy will be installed from a source file. Note that in this case you will need a C compiler (e.g., gcc or clang) to be installed (see Building from source below) for the installation to succeed.

If you get a PermissionError this means that you do not have the required administrative access to install new packages to your Python installation. In this case you may consider using the --user option to install the package into your home directory. You can read more about how to do this in the pip documentation.

Alternatively, if you intend to do development on other software that uses astropy, such as an affiliated package, consider installing astropy into a virtualenv.

Do not install astropy or other third-party packages using sudo unless you are fully aware of the risks.

Using Conda

To install astropy using conda run:

conda install astropy

astropy is installed by default with the Anaconda Distribution. To update to the latest version run:

conda update astropy

There may be a delay of a day or two between when a new version of astropy is released and when a package is available for conda. You can check for the list of available versions with conda search astropy.

If you want to install astropy along with recommended or all of the available optional dependencies, you can do:

conda install -c conda-forge -c defaults scipy matplotlib

or:

conda install -c conda-forge -c defaults scipy matplotlib \
  h5py beautifulsoup4 html5lib bleach pandas sortedcontainers \
  pytz setuptools mpmath bottleneck jplephem asdf pyarrow

To also be able to run tests (see below) and support Building Documentation use the following. We use pip for these packages to ensure getting the latest releases which are compatible with the latest pytest and sphinx releases:

pip install pytest-astropy sphinx-astropy

Warning

Attempting to use pip to upgrade your installation of astropy itself may result in a corrupted installation.

Testing an Installed astropy

See the latest documentation on how to test your installed version of astropy.

Requirements

astropy has the following strict requirements:

astropy also depends on a number of other packages for optional features. The following are particularly recommended:

The further dependencies provide more specific features:

  • h5py: To read/write Table objects from/to HDF5 files.

  • BeautifulSoup: To read Table objects from HTML files.

  • html5lib: To read Table objects from HTML files using the pandas reader.

  • bleach: Used to sanitize text when disabling HTML escaping in the Table HTML writer.

  • xmllint: To validate VOTABLE XML files. This is a command line tool installed outside of Python.

  • pandas: To convert Table objects from/to pandas DataFrame objects. Version 0.14 or higher is required to use the Pandas I/O functions to read/write Table objects.

  • sortedcontainers for faster SCEngine indexing engine with Table, although this may still be slower in some cases than the default indexing engine.

  • pytz: To specify and convert between timezones.

  • jplephem: To retrieve JPL ephemeris of Solar System objects.

  • setuptools: Used for discovery of entry points which are used to insert fitters into astropy.modeling.fitting.

  • mpmath: Used for the ‘kraft-burrows-nousek’ interval in poisson_conf_interval.

  • asdf >=2.10.0 or later: Enables the serialization of various Astropy classes into a portable, hierarchical, human-readable representation.

  • bottleneck: Improves the performance of sigma-clipping and other functionality that may require computing statistics on arrays with NaN values.

  • certifi: Useful when downloading files from HTTPS or FTP+TLS sites in case Python is not able to locate up-to-date root CA certificates on your system; this package is usually already included in many Python installations (e.g., as a dependency of the requests package).

  • pyarrow >=5.0.0 or later: To read/write Table objects from/to Parquet files.

  • fsspec >=2022.8.2 or later: Enables access to subsets of remote FITS files without having to download the entire file.

  • s3fs >=2022.8.2 or later: Enables access to files hosted in AWS S3 cloud storage.

However, note that these packages require installation only if those particular features are needed. astropy will import even if these dependencies are not installed.

The following packages can optionally be used when testing:

Building from Source

Prerequisites

You will need a compiler suite and the development headers for Python in order to build astropy. You do not need to install any other specific build dependencies (such as Cython) since these are declared in the pyproject.toml file and will be automatically installed into a temporary build environment by pip.

Prerequisites for Linux

On Linux, using the package manager for your distribution will usually be the easiest route to making sure you have the prerequisites to build astropy. In order to build from source, you will need the Python development package for your Linux distribution, as well as pip.

For Debian/Ubuntu:

sudo apt-get install python3-dev python3-numpy-dev python3-setuptools cython3 python3-pytest-astropy

For Fedora/RHEL:

sudo yum install python3-devel python3-numpy python3-setuptools python3-Cython python3-pytest-astropy

Note

Building the developer version of astropy may require newer versions of the above packages than are available in your distribution’s repository. If so, you could either try a more up-to-date distribution (such as Debian testing), or install more up-to-date versions of the packages using pip or conda in a virtual environment.

Prerequisites for Mac OS X

On MacOS X you will need the XCode command line tools which can be installed using:

xcode-select --install

Follow the onscreen instructions to install the command line tools required. Note that you do not need to install the full XCode distribution (assuming you are using MacOS X 10.9 or later).

The instructions for building NumPy from source are a good resource for setting up your environment to build Python packages.

Obtaining the Source Packages

Source Packages

The latest stable source package for astropy can be downloaded here.

Development Repository

The latest development version of astropy can be cloned from GitHub using this command:

git clone https://github.com/astropy/astropy.git

If you wish to participate in the development of astropy, see the Developer Documentation. The present document covers only the basics necessary to installing astropy.

Building and Installing

To build and install astropy (from the root of the source tree):

pip install .

If you install in this way and you make changes to the code, you will need to re-run the install command for changes to be reflected. Alternatively, you can use:

pip install -e .

which installs astropy in develop/editable mode – this then means that changes in the code are immediately reflected in the installed version.

Troubleshooting

If you get an error mentioning that you do not have the correct permissions to install astropy into the default site-packages directory, you can try installing with:

pip install . --user

which will install into a default directory in your home directory.

External C Libraries

The astropy source ships with the C source code of a number of libraries. By default, these internal copies are used to build astropy. However, if you wish to use the system-wide installation of one of those libraries, you can set environment variables with the pattern ASTROPY_USE_SYSTEM_??? to 1 when building/installing the package.

For example, to build astropy using the system’s expat parser library, use:

ASTROPY_USE_SYSTEM_EXPAT=1 pip install -e .

To build using all of the system libraries, use:

ASTROPY_USE_SYSTEM_ALL=1 pip install -e .

The C libraries currently bundled with astropy include:

  • wcslib see cextern/wcslib/README for the bundled version. To use the system version, set ASTROPY_USE_SYSTEM_WCSLIB=1.

  • cfitsio see cextern/cfitsio/changes.txt for the bundled version. To use the system version, set ASTROPY_USE_SYSTEM_CFITSIO=1.

  • expat see cextern/expat/README for the bundled version. To use the system version, set ASTROPY_USE_SYSTEM_EXPAT=1.

Installing pre-built Development Versions of astropy

Most nights a development snapshot of astropy will be compiled. This is useful if you want to test against a development version of astropy but do not want to have to build it yourselves. You can see the available astropy dev snapshots page to find out what is currently being offered.

Installing these “nightlies” of astropy can be achieved by using pip:

$ pip install -U -i https://pypi.anaconda.org/astropy/simple astropy --pre

The extra index URL tells pip to check the pip index on pypi.anaconda.org, where the nightlies are stored, and the --pre command tells pip to install pre-release versions (in this case .dev releases).

Building Documentation

Note

Building the documentation is in general not necessary unless you are writing new documentation or do not have internet access, because the latest (and archive) versions of Astropy’s documentation should be available at docs.astropy.org .

Dependencies

Building the documentation requires the astropy source code and some additional packages. The easiest way to build the documentation is to use tox as detailed in Building. If you are happy to do this, you can skip the rest of this section.

On the other hand, if you wish to call Sphinx manually to build the documentation, you will need to make sure that a number of dependencies are installed. If you use conda, the easiest way to install the dependencies is with:

conda install -c conda-forge sphinx-astropy

Without conda, you install the dependencies by specifying [docs] when installing astropy with pip:

pip install -e '.[docs]'

You can alternatively install the sphinx-astropy package with pip:

pip install sphinx-astropy

In addition to providing configuration common to packages in the Astropy ecosystem, this package also serves as a way to automatically get the main dependencies, including:

  • Sphinx - the main package we use to build the documentation

  • astropy-sphinx-theme - the default ‘bootstrap’ theme used by astropy and a number of affiliated packages

  • sphinx-automodapi - an extension that makes it easy to automatically generate API documentation

  • sphinx-gallery - an extension to generate example galleries

  • numpydoc - an extension to parse docstrings in NumPyDoc format

  • pillow - used in one of the examples

  • Graphviz - generate inheritance graphs (available as a conda package or a system install but not in pip)

Note

Both of the pip install methods above do not include Graphviz. If you do not install this package separately then the documentation build process will produce a very large number of lengthy warnings (which can obscure bona fide warnings) and also not generate inheritance graphs.

Building

There are two ways to build the Astropy documentation. The easiest way is to execute the following tox command (from the astropy source directory):

tox -e build_docs

If you do this, you do not need to install any of the documentation dependencies as this will be done automatically. The documentation will be built in the docs/_build/html directory, and can be read by pointing a web browser to docs/_build/html/index.html.

Alternatively, you can do:

cd docs
make html

Note

If you have a multi-core processor, and wish to leverage this for building documentation, you can do so as follows:

cd docs
SPHINXOPTS="-j N" make html

where N is the number of processes over which to distribute the build, as described in the sphinx-build Documentation.

The documentation will be generated in the same location. Note that this uses the installed version of astropy, so if you want to make sure the current repository version is used, you will need to install it with e.g.:

pip install -e .[docs]

before changing to the docs directory.

In the second way, LaTeX documentation can be generated by using the command:

make latex

The LaTeX file Astropy.tex will be created in the docs/_build/latex directory, and can be compiled using pdflatex.

Reporting Issues/Requesting Features

As mentioned above, building the documentation depends on a number of Sphinx extensions and other packages. Since it is not always possible to know which package is causing issues or would need to have a new feature implemented, you can open an issue in the core astropy package issue tracker. However, if you wish, you can also open issues in the repositories for some of the dependencies:

  • For requests/issues related to the appearance of the docs (e.g. related to the CSS), you can open an issue in the astropy-sphinx-theme issue tracker.

  • For requests/issues related to the auto-generated API docs which appear to be general issues rather than an issue with a specific docstring, you can use the sphinx-automodapi issue tracker.

  • For issues related to the default configuration (e.g which extensions are enabled by default), you can use the sphinx-astropy issue tracker.

Testing a Source Code Build of astropy

See the latest documentation on how to run the tests in a source checkout of astropy