BASHCRASH

A crash course in bash for the data scientist.

Unix Philosophy

"This is the Unix philosophy: Write programs that do one thing and do it well. Write programs to work together. Write programs to handle text streams, because that is a universal interface." -- Doug McIlroy

Basics

List directory contents with ls, create a directory with mkdir, copy files with cp, move and rename files with mv, remove files and directories with rm and rmdir.

Most importantly, access the Bash manual man to get help with built-in commands.

Keyboard shortcuts

Command-line keyboard shortcuts are immensely useful. For example, ctrl+a moves the cursor to the beginning of the line and ctrl+e moves the cursor to the end. Additionally, ctrl+k clears the line after the cursor, and ctrl+l clears the whole terminal window -- equivalent to clear.

See a list of shortcuts here: Wikipedia article on Bash, shortcut section

Optional matching, brace expansion, and !$

Not only can you use the * wildcard, but bash also supports optional matching. Show all tsv and csv files in the current directory:

ls -l *.[tc]sv

Expand items with brackets {}. Only list filenames with certain extensions:

ls -l *.{csv,tsv,sql,json}

Another useful feature is the ability to repeat the last argument with !$. Show all csv files, then move them to a directory, data/:

ls -l *.csv
mv !$ data/

Find

The find command has several different applications.

To search all sub-directories within a directory called zipfian for python files:

find ~/zipfian -name ".py"

The find command can also execute a given command on all of the files returned by the search using the -exec flag. To search for all python files within the zipfian directory that contain the string "RandomForestClassifier":

find ~/zipfian -name "*.py" -exec grep -l "RandomForestClassifier" {} \;

Many more examples on the find command here: Blog post by Alvin Alexander

Environment

Where does Bash search for commands?

echo $PATH

$PATH is a system variable string which contains paths separated by colons :. Bash searches the paths in $PATH, precedence goes to the command left-most path if you have more than one command in the search path.

It is very common for $PATH to be defined or appended within your shell initialization script. By default, that script is ~/.bash_profile. If you are using oh-my-zsh, then your initialization script is ~/.zshrc.

Where are commands located in the system?

The very basic system commands are located in /bin and /usr/bin/ directories.

Have a look at those commands using ls:

ls -l /usr/bin

Or, search those commands by piping the above output to grep. In the below example, I search for the word "grep" to search for all of the regular expression commands related to grep.

ls -l /usr/bin | grep "grep"

Note that the quotation marks are not necessary. Bash knows that the pattern argument passed into grep is a string.

To show the origin of a specific command, link, or alias, you can use either which or type. Also, most properly implemented commands and interpreters have a --version flag that will display the version of the command.

which python
python --version

Scripting

#! The hashbang! Also known as the shebang.

example python script: environ.py

#!/usr/bin/env python

import os

for item in os.environ.items():
    print item

Supplying #!/usr/bin/env python in first line of a script will tell Bash to pass the script as an argument to your current environment's version of python. In other words, it is equivalent to running python environ.py in your current shell.

But, there's one more step. That is to add the executable mode to your file using chmod +x.

chmod +x environ.py

Now, the file can be executed by typing out its relative or absolute path. In the current working directory, that will be ./+filename.

./environ.py

For Bash scripts, it is common to just provide the path to bash in the shebang.

example bash script: environ.sh

#!/bin/bash

cat printenv > environment.txt

echo "environment variables stored in environment.txt"

• For an example Bash script, see quakes.sh, a simple wrapper for the USGS earthquake feed.
• Learn more about shebang here: Shebang wikipedia article
• Learn more about writing scripts in the Bash language here: Robert Kline's Bash Tutorial

Process Management

; and &

To run commands in a batch, separate them by semicolons:

command1 ; command2

Or, to run both commands concurrently:

command1 & command2

Note: command1 starts running in the background, and command2 runs in the foreground.

ctrl, bg, fg, and kill

To quit a process running in the foreground with ctrl+c or in some cases, ctrl+d.

Suspend a process running in the foreground with ctrl+z. You will see something like this:

^Z
[2]  + 71826 suspended  ./energy.py

Resume a suspended process in the foreground with fg. Or you can run it in the background with bg, which is equivalent to running ./energy.py &.

Use jobs -l to list ongoing and suspended jobs with their pid (process id). The output looks something like this:

[1]  - 71979 running    ./pima.py
[2]  + 71826 suspended  ./energy.py

Finally, you can quit that suspended process with kill:

kill 71826

ps and top

To display all current shell window processes use ps, and to see all current system processes use ps ax. Or, to see what process is eating up all of your memory have a look at the top command.

purge

After many hours of uptime and usage your computer can build up inactive memory. Darwin OS's purge simulates restart conditions by flushing disk cache.

Working with CSV

Manipulation with sed and awk

When CSVs are improperly formatted, sed and awk can come to the rescue. For help on crafting regular expressions, see Regexr.

Blog posts on csv manipulation using sed and awk:

• Greg Reda: Useful commands for data science
• Adam Shinn: Sultan of Awk

CSVkit

CSVkit is a super-useful collection of commands that operate on CSVs. It is implemented in python, and can be easily installed with:

pip install csvkit

Displaying CSVs with csvlook

To display a csv in an easy-to-read format, use csvlook:

cat example.csv | csvlook

If it is a csv with many columns and rows, this may not be a pretty output. In that case, you can then pipe the output to less -S for easier viewing. The -S flag tells less not to fold lines.

cat example.csv | csvlook | less -S

Furthermore, you can choose columns with csvcut and sort on specified columns with csvsort.

Summary statistics on the command line

Much like with Pandas' describe method, you can run summary statistics on a CSV with the use of csvstat.

For more information, see: CSVkit documentation

Working with JSON

Another super-useful tool for displaying and manipulating data on the command line is JQ. You can download JQ with homebrew:

brew install jq

Or, if you don't have homebrew, you can find it here: JQ home

JQ can pretty-print, extract, and manipulate JSON data on the command line.

cat example.json | jq ".[0]"

Recommended shell, zsh

• oh-my-zsh
• my zshrc profile

Other Useful Links

• Regexr
• 7 command-line tools for data science
• Data Science at the Command Line