!pwd
/Users/jhoward/Documents/GitHub/pshnb
psh
magicspshnb
pshnb
adds %psh
and %%psh
functions to Jupyter and IPython, which execute expressions in a persistent shell.
Install pshnb with:
pip install pshnb
Once that’s complete, you can install the magics to all IPython and Jupyter sessions automatically by running in your terminal:
pshnb_install
In jupyter and ipython, you can run a shell command using the !
prefix:
However, each time you run this command, a new shell is created and then removed, so for instance, cd
doesn’t actually do anything if you run another command afterwards:
As you see from the !pwd
output, the directory hasn’t actually changed!
%psh
, on the other hand, creates a persistent shell, which solves this problem:
With %psh
, you can implement, and document in notebooks, multi-step stateful shell interactions, including setting environment variables, sourcing scripts, and changing directories.
You can use the %%psh
cell magic to run multi-line shell commands, such as here-docs. For instance:
This creates a file called tmp
containing two lines. Let’s check it worked, and then remove it – as you see, you can also use the cell magic to run multiple commands:
You can pipe commands together just like in a regular shell, and use standard unix utilities like head
to process the output. For instance, here we show just the first 3 lines of the directory listing:
You can use Python variables in your shell commands by prefixing them with @{}
. For instance, here we create a variable n
and then display it using echo
:
Here we use n
to show just the first two entries from the directory listing:
You can run commands in the background in the shell by adding &
at the end of a command. The parentheses (...)
group commands together to run as one unit. In this example, we first print “starting”, and then create a background process that will wait for 1 second (using sleep 1
) and then print “finished”. The shell immediately shows us “starting” and tells us it created background process number 1 (with a process ID):
For this demonstration, we wait for 1.1 seconds (slightly longer than the background process needs). During this time, the background process will complete in the background. But we won’t see any output from it yet.
The next time we run any psh
magic we will also see any output that has occurred in our persistent shell since the last command. Run %psh
by itself to only see those updates, e.g here we see that “finished” was printed, and the shell tells us that background job 1 completed successfully.
You can get help on the %psh
magic’s options using -h
.
::
%psh [-h] [-r] [-o] [-x] [-X] [-s] [-S] [-t TIMEOUT] [command ...]
Run line or cell in persistent shell
positional arguments:
command The command to run
options:
-h, --help Show this help
-r, --reset Reset the shell interpreter
-o, --obj Return this magic object
-x, --expand Enable variable expansion
-X, --no-expand Disable variable expansion
-s, --sudo Enable sudo
-S, --no-sudo Disable sudo
-t TIMEOUT, --timeout TIMEOUT
Set timeout in seconds
You can reset the shell to its initial state using the -r
flag. Let’s first check our current directory:
Now let’s reset the shell:
As you can see, after resetting we’re back in our starting directory:
The -s
flag enables sudo
mode, which runs commands as the root user, and -S
disables it. For instance, here we first enable sudo
mode:
Then we can check which user we’re running as:
As you can see, we’re now running as root
. We can disable sudo
mode:
And when we check again, we’re back to our regular user:
You can set a timeout (in seconds) using the -t
flag, which will raise a TIMEOUT
exception if a command takes too long. For instance, here we set a 1-second timeout:
Then we try to run a command that sleeps for 2 seconds – since this is longer than our timeout, we’ll get a timeout error: