Installation

Note

Fermipy is only compatible with Science Tools v10r0p5 or later. If you are using an earlier version, you will need to download and install the latest version from the FSSC. Note that it is recommended to use the non-ROOT binary distributions of the Science Tools.

These instructions assume that you already have a local installation of the Fermi Science Tools (STs). For more information about installing and setting up the STs see Installing the Fermi Science Tools. If you are running at SLAC you can follow the Running at SLAC instructions. For Unix/Linux users we currently recommend following the Installing with Anaconda Python instructions. For OSX users we recommend following the Installing with pip instructions. The Installing with Docker instructions can be used to install the STs on both OSX and Linux machines that are new enough to support Docker.

Installing the Fermi Science Tools

The Fermi STs are a prerequisite for fermipy. To install the STs we recommend using one of the non-ROOT binary distributions available from the FSSC. The following example illustrates how to install the binary distribution on a Linux machine running Ubuntu Trusty:

$ curl -OL http://fermi.gsfc.nasa.gov/ssc/data/analysis/software/tar/ScienceTools-v10r0p5-fssc-20150518-x86_64-unknown-linux-gnu-libc2.19-10-without-rootA.tar.gz
$ tar xzf ScienceTools-v10r0p5-fssc-20150518-x86_64-unknown-linux-gnu-libc2.19-10-without-rootA.tar.gz
$ export FERMI_DIR=ScienceTools-v10r0p5-fssc-20150518-x86_64-unknown-linux-gnu-libc2.19-10-without-rootA/x86_64-unknown-linux-gnu-libc2.19-10
$ source $FERMI_DIR/fermi-init.sh

More information about installing the STs as well as the complete list of the available binary distributions is available on the FSSC software page.

Installing with pip

These instructions cover installation with the pip package management tool. This method will install fermipy and its dependencies into the python distribution that comes with the Fermi Science Tools. First verify that you’re running the python from the Science Tools

$ which python

If this doesn’t point to the python in your Science Tools install (i.e. it returns /usr/bin/python or /usr/local/bin/python) then the Science Tools are not properly setup.

Before starting the installation process, you will need to determine whether you have setuptools and pip installed in your local python environment. You may need to install these packages if you are running with the binary version of the Fermi Science Tools distributed by the FSSC. The following command will install both packages in your local environment:

$ curl https://bootstrap.pypa.io/get-pip.py | python -

Check if pip is correctly installed:

$ which pip

Once again, if this isn’t the pip in the Science Tools, something went wrong. Now install fermipy by running

$ pip install fermipy

To run the ipython notebook examples you will also need to install jupyter notebook:

$ pip install jupyter

Finally, check that fermipy imports:

$ python
Python 2.7.8 (default, Aug 20 2015, 11:36:15)
[GCC 4.2.1 Compatible Apple LLVM 6.0 (clang-600.0.56)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from fermipy.gtanalysis import GTAnalysis
>>> help(GTAnalysis)

Installing with Anaconda Python

Note

The following instructions have only been verified to work with binary Linux distributions of the Fermi STs. If you are using OSX or you have installed the STs from source you should follow the Installing with pip thread above.

These instructions cover how to use fermipy with a new or existing anaconda python installation. These instructions assume that you have already downloaded and installed the Fermi STs from the FSSC and you have set the FERMI_DIR environment variable to point to the location of this installation.

If you already have an existing anaconda python installation then fermipy can be installed from the conda-forge channel as follows:

$ conda config --append channels conda-forge
$ conda install fermipy

If you do not have an anaconda installation, the condainstall.sh script can be used to create a minimal anaconda installation from scratch. First download and source the condainstall.sh script from the fermipy repository:

$ curl -OL https://raw.githubusercontent.com/fermiPy/fermipy/master/condainstall.sh
$ source condainstall.sh

If you do not already have anaconda python installed on your system this script will create a new installation under $HOME/miniconda. If you already have anaconda installed and the conda command is in your path the script will use your existing installation. After running condainstall.sh fermipy can be installed with conda:

$ conda install fermipy

Alternatively fermipy can be installed from source following the instructions in Building from Source.

Once fermipy is installed you can initialize the ST/fermipy environment by running condasetup.sh:

$ curl -OL https://raw.githubusercontent.com/fermiPy/fermipy/master/condasetup.sh
$ source condasetup.sh

If you installed fermipy in a specific conda environment you should switch to this environment before running the script:

$ source activate fermi-env
$ source condasetup.sh

Installing with Docker

Note

This method for installing the STs is currently experimental and has not been fully tested on all operating systems. If you encounter issues please try either the pip- or anaconda-based installation instructions.

Docker is a virtualization tool that can be used to deploy software in portable containers that can be run on any operating system that supports Docker. Before following these instruction you should first install docker on your machine following the installation instructions for your operating system. Docker is currently supported on the following operating systems:

  • macOS 10.10.3 Yosemite or later
  • Ubuntu Precise 12.04 or later
  • Debian 8.0 or later
  • RHEL7 or later
  • Windows 10 or later

Note that Docker is not supported by RHEL6 or its variants (CentOS6, Scientific Linux 6).

These instructions describe how to create a docker-based ST installation that comes preinstalled with anaconda python and fermipy. The installation is fully contained in a docker image that is roughly 2GB in size. To see a list of the available images go to the fermipy Docker Hub page. Images are tagged with the release version of the STs that was used to build the image (e.g. 11-05-00). The latest tag points to the image corresponding to the most recent ST release.

To install the latest image first download the image file:

$ docker pull fermipy/fermipy
$ docker tag fermipy/fermipy fermipy

This will create an image called fermipy. Now change to the directory where you plan to do your analysis and run the following command to launch a docker container instance:

$ docker run -it --rm -p 8888:8888 -v $PWD:/workdir -w /workdir fermipy

This will start an ipython notebook server that will be attached to port 8888. Once the server is running you can start a notebook session by navigating to the URL http://localhost:8888/. The -v $PWD:/workdir argument mounts the current directory to the working area of the container. Additional directories may be mounted by adding more volume arguments -v with host and container paths separated by a colon.

The same docker image may be used to launch python, ipython, or a bash shell by passing the command as an argument to docker run:

$ docker run -it --rm -v $PWD:/workdir -w /workdir fermipy ipython
$ docker run -it --rm -v $PWD:/workdir -w /workdir fermipy python
$ docker run -it --rm -v $PWD:/workdir -w /workdir fermipy /bin/bash

By default interactive graphics will not be enabled. The following code can be inserted at the top of your analysis script to fall-back to a non-interactive backend:

from fermipy.utils import init_matplotlib_backend
init_matplotlib_backend()

The following commands can be used to enable X11 forwarding for interactive graphics on an OSX machine. This requires you to have installed XQuartz 2.7.10 or later. First enable remote connections by default and start the X server:

$ defaults write org.macosforge.xquartz.X11 nolisten_tcp -boolean false
$ open -a XQuartz

Now check that the X server is running and listening on port 6000:

$ lsof -i :6000

If you don’t see X11 listening on port 6000 then try restarting XQuartz.

Once you have XQuartz configured you can enable forwarding by setting DISPLAY environment variable to the IP address of the host machine:

$ export HOST_IP=`ifconfig en0 | grep "inet " | cut -d " " -f2`
$ xhost +$HOST_IP
$ docker run -it --rm -e DISPLAY=$HOST_IP:0 -v $PWD:/workdir -w /workdir fermipy ipython

Running at SLAC

This section provides specific installation instructions for running in the SLAC computing environment. First download and source the slacsetup.sh script:

$ wget https://raw.githubusercontent.com/fermiPy/fermipy/master/slacsetup.sh -O slacsetup.sh
$ source slacsetup.sh

To initialize the ST environment run the slacsetup function:

$ slacsetup

This will setup your GLAST_EXT path and source the setup script for one of the pre-built ST installations (the current default is 11-05-00). To manually override the ST version you can provide the release tag as an argument to slacsetup:

$ slacsetup XX-XX-XX

Because users don’t have write access to the ST python installation all pip commands that install or uninstall packages must be executed with the --user flag. After initializing the STs environment, install fermipy with pip:

$ pip install fermipy --user

This will install fermipy in $HOME/.local. You can verify that the installation has succeeded by importing GTAnalysis:

$ python
Python 2.7.8 |Anaconda 2.1.0 (64-bit)| (default, Aug 21 2014, 18:22:21)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
Anaconda is brought to you by Continuum Analytics.
Please check out: http://continuum.io/thanks and https://binstar.org
>>> from fermipy.gtanalysis import GTAnalysis

Upgrading

By default installing fermipy with pip or conda will get the latest tagged released available on the PyPi package respository. You can check your currently installed version of fermipy with pip show:

$ pip show fermipy

or conda info:

$ conda info fermipy

To upgrade your fermipy installation to the latest version run the pip installation command with --upgrade --no-deps (remember to also include the --user option if you’re running at SLAC):

$ pip install fermipy --upgrade --no-deps
Collecting fermipy
Installing collected packages: fermipy
  Found existing installation: fermipy 0.6.6
    Uninstalling fermipy-0.6.6:
      Successfully uninstalled fermipy-0.6.6
Successfully installed fermipy-0.6.7

If you installed fermipy with conda the equivalent command is:

$ conda update fermipy

Building from Source

These instructions describe how to install fermipy from its git source code repository using the setup.py script. Installing from source can be useful if you want to make your own modifications to the fermipy source code or test features in an untagged commit. Note that non-expert users are recommended to install a tagged release of fermipy following the Installing with pip or Installing with Anaconda Python instructions above.

First clone the fermipy git repository and cd to the root directory of the repository:

$ git clone https://github.com/fermiPy/fermipy.git
$ cd fermipy

To install the latest commit in the master branch run setup.py install from the root directory:

# Install the latest commit
$ git checkout master
$ python setup.py install --user

A useful option if you are doing active code development is to install your working copy of the package. This will create an installation in your python distribution that is linked to the copy of the code in your local repository. This allows you to run with any local modifications without having to reinstall the package each time you make a change. To install your working copy of fermipy run with the develop argument:

# Install a link to your source code installation
$ python setup.py develop --user

You can later remove the link to your working copy by running the same command with the --uninstall flag:

# Install a link to your source code installation
$ python setup.py develop --user --uninstall

You also have the option of installing a previous release tag. To see the list of release tags run git tag. To install a specific release tag, run git checkout with the tag name followed by setup.py install:

# Checkout a specific release tag
$ git checkout X.X.X
$ python setup.py install --user

Issues

If you get an error about importing matplotlib (specifically something about the macosx backend) you might change your default backend to get it working. The customizing matplotlib page details the instructions to modify your default matplotlibrc file (you can pick GTK or WX as an alternative). Specifically the TkAgg and macosx backends currently do not work on OSX if you upgrade matplotlib to the version required by fermipy. To get around this issue you can switch to the Agg backend at runtime before importing fermipy:

>>> import matplotlib
>>> matplotlib.use('Agg')

However note that this backend does not support interactive plotting.

If you are running OSX El Capitan or newer you may see errors like the following:

dyld: Library not loaded

In this case you will need to disable the System Integrity Protections (SIP). See here for instructions on disabling SIP on your machine.

In some cases the setup.py script will fail to properly install the fermipy package dependecies. If installation fails you can try running a forced upgrade of these packages with pip install --upgrade:

$ pip install --upgrade --user numpy matplotlib scipy astropy pyyaml healpy wcsaxes ipython jupyter