SAS Kernel for Jupyter

What is this?

A SAS Kernel for Jupyter Notebooks

Dependencies

• Python3.X
• Jupyter version 4 or higher
• SAS 9.4 or higher -- This includes SAS Viya

With the latest changes in saspy it is no longer a requirement that Jupyter and SAS be installed on the same machine. SAS and Jupyter can now communicate via passwordless ssh or IOM. This is in response to issue 11. The configuration details are located in sascfg.py and documentation for configuration is here

Documentation

Here is the link to the current documentation https://sassoftware.github.io/sas_kernel/

Install

To successfully use the SAS Kernel you must have each of the following:

• SAS version 9.4 or above
• Jupyter
  • Jupyter has a number of dependencies. See the subsections for steps on installing Jupyter on your system.
• Python 3

Install for Anaconda Python (assuming SAS already installed)

1. Download and install Anaconda Python (make sure you get Python3.X). If you install Anaconda without super user rights (root or sudo) then other users on the system will not be able to access the SAS kernel. A couple notes that I've observed:

   • The default install location is the users home directory. This is fine for a single user install I would put it in a common location (/opt) if you're doing a system wide install
   • One of the prompts is to add the path to your environment. I recommend you want to answer 'yes' to that question so that all the installing user has the executables in their path. If you're doing a system wide install (root or sudo) all the other users should add that path to their environmental variables

2. Install sas_kernel. The sas_kernel has a dependency on saspy which is located here. In the command below I'm assuming that pip maps to python3 if that is not the case the you might need to use pip3 instead.

   pip install sas_kernel
   

3. Verify that the sas_kernel is installed jupyter kernelspec list

   If you installed as a superuser, your output should similar to this:

   Available kernels:
     python3    /opt/Anaconda3-2.5.0/lib/python3.5/site-packages/ipykernel/resources
     sas        /usr/local/share/jupyter/kernels/sas
   

   If you installed as a regular user (sas in this case), your output should similar to this:

   Available kernels:
     python3    /home/sas/anaconda3/lib/python3.5/site-packages/ipykernel/resources
     sas        /home/sas/.local/share/jupyter/kernels/sas
   

4. Verify SAS Executable is correct

   1. find the sascfg.py file -- it is currently located in the install location (see above) [install location]/site-packages/saspy/sascfg.py This command will search the OS for the file location: find / -name sascfg.py
   2. edit the file with the correct path the SAS executable and include any options you wish it include in the SAS invocation. See examples in the file

Install for Centos 6 (assuming SAS already installed)

These instructions assume you'll be installed system wide for all users using a superuser account (root or sudo)

1. yum packages

   sudo yum install https://centos6.iuscommunity.org/ius-release.rpm
   sudo yum install python35u gcc-c++ python35u-devel python35u-pip python35u-tools
   

2. pip

   wget https://bootstrap.pypa.io/get-pip.py
   python3.5 get-pip.py
   pip3 --version
   

3. jupyter and sas_kernel. The sas_kernel has a dependency on saspy which is located here.

   pip3.5 install jupyter
   pip3.5 install sas_kernel
   

4. Verify that the sas_kernel is installed jupyter kernelspec list

   This should produce output similar to this:

   Available kernels:
     python3    /usr/lib/python3.5/site-packages/ipykernel/resources
     sas        /usr/local/share/jupyter/kernels/sas
   

5. Verify SAS Executable is correct

   1. find the sascfg.py file -- it is currently located in the site-packages area of python install find / -name sascfg.py
   2. edit the file with the correct path the SAS executable and include any options you wish it include in the SAS invocation. See examples in the file

Improving Usability

There are a few NBExtensions that have been created to make working with Jupyter notebooks more productive. These are largely the result of pain points from my use of SAS Kernel for programming tasks. The extensions can be found here. The list includes:

• SAS Log -- which show the SAS log for the last executed cell or the entire log since the last (re)start of the notebook
• themes -- this allows you to change the color scheme for your code to match the traditional SAS color scheme

Installing the SAS Extensions

Details for installing the extensions for SAS can be found here The install experience was improved with version 1.2

Jupyter Magics for the sas_kernel

There are magics that have been written specifically for the sas_kernel to get more details see the README

Jupyterhub

The SAS kernel can be used with JupyterHub for more information look here

NBGrader

nbgrader is a system for assigning and grading notebooks and extends jupyter. I have a number of contributions that I'm currently working on in conjunction with teaching SAS programming in a classroom setting. You can see my forked repo here

FAQ

• Is there a SAS Magic that I can access from a python kernel?

  Yes! There are actually several cell magics available from SAS. They are %%SAS, %%IML, and %%OPTMODEL. To load these magics in your notebook, execute the following command %load_ext saspy.sas_magic. You can check that the magics have are successfully activated by looking at the results of %lsmagic and looking in the cell magic section. If you use multiple SAS Cell magics in the same notebook they will share a SAS session (have the same WORK libname and MACROS). There is currently no sharing of SAS Sessions between different notebooks.

• Do I need to buy SAS to use this kernel?

  The SAS Kernel is simply a gateway for Jupyter notebooks to talk to SAS, as such, if SAS is not installed this kernel won't be that helpful. For information on purchasing SAS click here

• How does Jupyter communicate with SAS?

  Behind a Jupyter notebook is a python session, that python session submits code to SAS and receives responses through socket i/o (leveraging stdin , stdout, and stderr) which has been supported in SAS for a long time

• If stdin, stdout, and stderr have been supported for so long why do I need to have SAS 9.4 or newer?

  First, SAS 9.4 was released in July 2013 so it isn't exactly a bleeding edge requirement. The reason for a prerequisite for SAS 9.4 is because that was the first release that supported the creation of HTML5 documents and that is the returned output so that we can render attractive tables and graphs automagically

• How can I see my SAS log, I only see the listing output?

  SAS is different from many other programming languages in that it has two useful information streams, the log (which details the technical details of what happened and how long it took) and the lst (which includes the tables and graphics from the analysis). The SAS Kernel attempts to show you I think you want. Here are the rules:

  LOG	LST	DISPLAYED	NOTES
  Yes	No	LOG	This happens when you run DATA Step or a PROC with the noprint option
  Yes	Yes	LST	---
  Yes (with ERROR message(s))	Yes	ERROR messages with context from the log, then the listing output	---
  Yes (with ERROR message(s))	No	LOG	---

  If you want to see the log but it was not displayed you can use SASLog NBExtension which will show the log for the last executed cell or the entire log since the last (re)start of the notebook

• Will this leave a bunch of SAS sessions hanging around?

  A SAS session is started for each notebook you have open i.e. 5 notebooks open = 5 SAS sessions. Those sessions will remain active for the life of the notebook. If you shutdown your notebook, the SAS session will also terminate. In Jupyterhub, there are configuration options to shutdown inactive sessions and the SAS kernel complies with those directives.

• I restarted my SAS Kernel and now my WORK library is now empty. What happened?

  When you restart the kernel in a notebook you are terminating the current SAS session and starting a new one. All of the temporary artifacts, data sets in the WORK library, assigned libnames, filename, WORK macros, and so on are destroyed.