In this document we describe several methods for installing yt. The method that
will work best for you depends on your precise situation:
All-in-One Installation Script
Because installation of all of the interlocking parts necessary to install yt
itself can be time-consuming, yt provides an all-in-one installation script
which downloads and builds a fully-isolated Python + NumPy + Matplotlib + HDF5 +
Mercurial installation. Since the install script compiles yt’s dependencies from
source, you must have C, C++, and optionally Fortran compilers installed.
The install script supports UNIX-like systems, including Linux, OS X, and most
supercomputer and cluster environments. It is particularly suited for deployment
in environments where users do not have root access and can only install
software into their home directory.
Since the install is fully-isolated in a single directory, if you get tired of
having yt on your system, you can just delete the directory and yt and all of
its dependencies will be removed from your system (no scattered files remaining
throughout your system).
Running the Install Script
To get the installation script, download it from:
wget https://bitbucket.org/yt_analysis/yt/raw/stable/doc/install_script.sh
By default, the bash install script will install an array of items, but there
are additional packages that can be downloaded and installed (e.g. SciPy, enzo,
etc.). The script has all of these options at the top of the file. You should be
able to open it and edit it without any knowledge of bash syntax. To execute
it, run:
Because the installer is downloading and building a variety of packages from
source, this will likely take a while (e.g. 20 minutes), but you will get
updates of its status at the command line throughout.
If you receive errors during this process, the installer will provide you
with a large amount of information to assist in debugging your problems. The
file yt_install.log will contain all of the stdout and stderr from
the entire installation process, so it is usually quite cumbersome. By looking
at the last few hundred lines (i.e. tail -500 yt_install.log), you can
potentially figure out what went wrong. If you have problems, though, do not
hesitate to contact us for assistance.
Activating Your Installation
Once the installation has completed, there will be instructions on how to set up
your shell environment to use yt by executing the activate script. You must
run this script in order to have yt properly recognized by your system. You can
either add it to your login script, or you must execute it in each shell session
prior to working with yt.
source <yt installation directory>/bin/activate
If you use csh or tcsh as your shell, activate that version of the script:
source <yt installation directory>/bin/activate.csh
If you don’t like executing outside scripts on your computer, you can set
the shell variables manually. YT_DEST needs to point to the root of the
directory containing the install. By default, this will be yt-<arch>, where
<arch> is your machine’s architecture (usually x86_64 or i386). You
will also need to set LD_LIBRARY_PATH and PYTHONPATH to contain
$YT_DEST/lib and $YT_DEST/python2.7/site-packages, respectively.
Updating yt and Its Dependencies
With many active developers, code development sometimes occurs at a furious
pace in yt. To make sure you’re using the latest version of the code, run
this command at a command-line:
Additionally, if you want to make sure you have the latest dependencies
associated with yt and update the codebase simultaneously, type this:
Removing yt and Its Dependencies
Because yt and its dependencies are installed in an isolated directory when
you use the script installer, you can easily remove yt and all of its
dependencies cleanly. Simply remove the install directory and its
subdirectories and you’re done. If you really had problems with the
code, this is a last defense for solving: remove and then fully
re-install from the install script again.
Installing yt Using Anaconda
Perhaps the quickest way to get yt up and running is to install it using the
Anaconda Python Distribution,
which will provide you with a easy-to-use environment for installing Python
packages.
If you do not want to install the full anaconda python distribution, you can
install a bare-bones Python installation using miniconda. To install miniconda,
visit http://repo.continuum.io/miniconda/ and download a recent version of the
Miniconda-x.y.z script (corresponding to Python 2.7) for your platform and
system architecture. Next, run the script, e.g.:
bash Miniconda-3.3.0-Linux-x86_64.sh
Make sure that the Anaconda bin directory is in your path, and then issue:
which will install yt along with all of its dependencies.
Recipes to build conda packages for yt are available at
https://github.com/conda/conda-recipes. To build the yt conda recipe, first
clone the conda-recipes repository
git clone https://github.com/conda/conda-recipes
Then navigate to the repository root and invoke conda build:
cd conda-recipes
conda build ./yt/
Note that building a yt conda package requires a C compiler.
Installing yt Using pip or from Source
To install yt from source, you must make sure you have yt’s dependencies
installed on your system. These include: a C compiler, HDF5, python,
Cython, NumPy, matplotlib, sympy, and h5py. From here, you
can use pip (which comes with Python) to install the latest stable
version of yt:
The source code for yt may be found at the Bitbucket project site and can also
be utilized for installation. If you prefer to install the development version
of yt instead of the latest stable release, you will need mercurial to clone
the official repo:
hg clone https://bitbucket.org/yt_analysis/yt
cd yt
hg update yt
python setup.py install --user --prefix=
Note
If you maintain your own user-level python installation separate from the OS-level python
installation, you can leave off --user --prefix=, although you might need
sudo depending on where python is installed. See This StackOverflow
discussion
if you are curious why --prefix= is neccessary on systems.
This will install yt into a folder in your home directory
($HOME/.local/lib64/python2.7/site-packages on Linux,
$HOME/Library/Python/2.7/lib/python/site-packages/ on OSX) Please refer to
the setuptools documentation for the additional options.
If you choose this installation method, you do not need to run any activation
script since this will install yt into your global python environment.
If you will be modifying yt, you can also make the clone of the yt mercurial
repository the “active” installed copy:
hg clone https://bitbucket.org/yt_analysis/yt
cd yt
hg update yt
python setup.py develop --user --prefix=
As above, you can leave off --user --prefix= if you want to install yt into the default
package install path. If you do not have write access for this location, you
might need to use sudo.
Keeping yt Updated via Mercurial
If you want to maintain your yt installation via updates straight from the
Bitbucket repository or if you want to do some development on your own, we
suggest you check out some of the development docs,
especially the sections on Mercurial and
building yt from source.
You can also make use of the following command to keep yt up to date from the
command line:
This will detect that you have installed yt from the mercurial repository, pull
any changes from bitbucket, and then recompile yt if necessary.
The yt Project
3.1