Anaconda at STScI - A Walkthrough¶

Step 1: Removing references to Ureka¶

Remove all references to Ureka (SSB , SSBX, SSBDEV) in your shell files. If you have never edited your bash setup files .profile, or .bash*), you can ignore these for now as we are going to delete them all later on, and just focus on your tcsh files, these could include: .cshrc, .setenv, or .scienv. You may or may not have some of these files depending on when your machine was set up originally. Accounts created after 2008 may have an additional .mysetenv file.

If you not sure what to look for, and haven’t edited your shell files before, look for references to ssbx, ssbrel or ssbdev and remove those lines. You can use the following listed commands to search for Ureka dependencies. You may see lines from your $HOME/.ureka/.default and $HOME/.bash_history files. This is expected and can be ignored.

Linux:

find $HOME -maxdepth 1 -type f -name '.*' | xargs -I'{}' grep -n -H -E --color=auto 'ssbrel|ssbx|ssbdev' "{}"

OSX:

Darwin’s BSD find does not support -maxdepth (users can install gnu findutils if they wish to execute the Linux-style command on OS X, however)

find $HOME -type f -name '.*' | xargs -I'{}' grep -n -H -E --color=auto 'ssbrel|ssbx|ssbdev' "{}"

Step 2: Backup your current bash files¶

Tar and zip your current bash setup files with the following command:

tar cfz envbackup.tar.gz .profile .bash*

Step 3: Rebuilding your bash startup files¶

Now that you have a backup of your bash files safely stored, wipe all bash startup files currently in your ~home directory.

rm ~/.profile
rm ~/.bash*

You should also delete your $HOME/.pydistutils.cfg file. This file can interefere with Anaconda’s Python install.

rm ~/.pydistutils.cfg

Now you can open up a new ~/.bash_profile file. Remember this should be a blank text file, since we just deleted the previous copy if it existed. We will also set up a ~/.bashrc file. Below is an example of a standard ~/.bash_profile and ~/.bashrc.

# File - ~/.bash_profile (or .profile)

# Common shell aliases
alias ls='ls -G'
alias ll='ls -l'
alias grep='grep --color=auto'

if [ -f $HOME/.bashrc ]; then
    source $HOME/.bashrc
fi

# EOF

# File - ~/.bashrc

# Tune your profile… these are example only
# Replace these with desired paths
export PATH="$PATH:$LOCAL_CUSTOM/bin:$PATH"
export MANPATH="$LOCAL_CUSTOM/share/man:$MANPATH"

# EOF

Using this line:

if [ -f $HOME/.bashrc ]; then
    source $HOME/.bashrc
fi

the ~/.bashrc file will get sourced by ~/.bash_profile.

Now we can start to port the environment setup information that was in the tcsh startup files over to your bash files. Most of these commands will either be setenv or alias commands. There is a syntax difference between ``tcsh`` and ``bash``. You can put these kinds of commands into your .bash_profile file. Below are some examples of how to translate tcsh to bash syntax.

Finally, you should now restart your terminal program so that these changes are applied.

Step 4: Bash as default, or temporary bash sessions¶

chsh -s /bin/bash

echo $SHELL

bash -l

Step 1: Install¶

There are several flavors of Anaconda for download. Anaconda contains the full anaconda software suite, while Miniconda contains only Conda and Python. You also have the choice of Python 3 or Python 2.7 as your default Python version. Please note, you can still access Python 3 from the Anaconda Python 2.7 version, and vice versa. This just sets up your default to be one or the other. Further details and installation instructions are available on the Anaconda webpage.

As previously mentioned, Anaconda works with the bash shell, so please make sure you are running install commands, etc. in a bash environment. To do this before you have switched to bash as your default terminal, please use

bash -l

Step 2: Make sure Anaconda was set up correctly¶

To test that your PATH variable was correctly set for Anaconda after installation, check for the following line in your ~/.bashrc file. It should be similar to the following:

Mac:
export PATH="$HOME/anaconda3/bin:$PATH"

Linux:
export PATH="/user/username/anaconda2/bin"

If this line is missing please add it, making sure to use your Anaconda folder name. This varies depending on which version of Anaconda/Miniconda you downloaded. The default install location of Anaconda/Miniconda is in your home directory. So if you’ve run the Anaconda/Miniconda install with the default settings you should now see a directory similar to this in your home directory: ~/anaconda2 (If you grabbed Anaconda2), ~/miniconda3 (if you grabbed Miniconda3), etc.

Let’s talk about Conda environments...¶

Creating and switching Conda environments is just like switching Ureka environments, but even better! We’ll cover it in a nutshell here, but for full details please see the Anaconda documentation. What so much better about it, you might ask? Well, Conda environments give you complete freedom to:

• Make as many different custom environments as you want
• Use different versions of different packages in each environment
• You decide when to update packages
• Can share your environment setup easily

Conda environments give you all the freedom to pick and choose what versions and which packages are included in that environment, without any of the hassle of having to manually install everything. It also does a great job of keeping all your different environments separate from each other. You can use Python 2.7 and an older version of Numpy in one environment, and another where you’re using Python 3 and the cutting edge version of your favorite Python libraries. This makes creating a testing environment extremely easy. Want that new Astropy function but not sure if that version of Astropy will break other parts of your code? Just make a new environment with the new Astropy version to test with. You can still keep your working environment just where you left it. Another great benefit is the ability to share an exact copy of your environment with others.

So how do you make a new Conda environment? This can be done in two simple statements. To make a new environment just use the following commands:

Example 1:

This installs a base version of a new environment. Containing just the basic Python libraries. Here we’re specifying to use Python 2 (2.7).

conda create -n myenvironment python=2

Example 2:

Here we’re creating a new environment that will be populated with the basic Python libraries and adding in the Python bokeh library.

conda create -n bokehenv python=2 bokeh

Example 3:

Alternatively, I could have omitted bokeh as an argument, or any specific packages, and populated my environment manually.

conda create -n bokehenv
source activate bokehenv
conda install bokeh

Example 4:

It is also possible to install packages into a named environment without the need to activate it first. The bokeh environment must already exist for this work

conda install -n bokehenv bokeh
source activate bokehenv

You’ll see these commands again as you walk through the AstroConda installation.

Step 3: Get AstroConda¶

AstroConda is a package repository that is built to hook into the Anaconda distribution. You can think of Anaconda (using Conda) as your environment manager, and AstroConda as an extra repository of packages and software (called a “channel” in Conda). It contains many of the tools that were built into Ureka. For more information on the AstroConda packages please see the AstroConda doc page. You will also want to reference this page if you need to include iraf in your AstroConda install.

The next step is to add the astroconda channel.

conda config --add channels http://ssb.stsci.edu/astroconda

Now we will create a new environment that contains AstroConda’s stsci metapackage

conda create -n astroconda stsci

and activate this new environment.

source activate astroconda

Make sure you are installing the stsci metapackage into a new environment and not your root Anaconda environment. If this has happened, please see the AstroConda FAQ page for instructions.

Step 3B: Add the Astropy channel¶

The Astropy Project also has a Anaconda channel which contains all of the astropy affiliated packages. While some of these are already included in astroconda, if you think you might want some that aren’t you can use the following command

conda config --add channels astropy

and you now have the astropy channel included by default.

Step 4: Quick test to see if your setup was succesful¶

Were going to simulate a fresh new bash shell for testing, so open a new terminal window and use the following commands. You can start from a tcsh terminal.

/grp/hst/ssb/blackhole/interactive.sh
source ~/.bash_profile

If successful, the PS1 variable, responsible for controlling the look and feel of your shell prompt, will be reset to the system default; often [user@host: directory]$. From here we will test Anaconda and your AstroConda install.

source activate astroconda
which python

which python should return /path/to/astroconda/bin/python, where /path/to/ will be the path to your Anaconda installation. If this test returns unexpcted results, you can contact support@stsci.edu for assistance.