Let's write a shell script

Now that you have a $PATH let’s start using it! Today we’ll write a simple shell script in your personal bin directory.

Let’s start by picking the language the shell script will be written in. You can choose any scripting language that you have available on your machine. In our script we’ll tell the shell what command to execute to run the script.

The shebang

The first line of a shell script is called the “shebang” or “hashbang”. It tells your shell that the way to run the script is to run that program with the script file as an argument.

The shebang has a specific syntax: #!/path/to/program/to/execute

For example:

#!/bin/bash

Says to run this script as an argument to /bin/bash. If we had our script in our home directory’s bin folder then the bash command would be given the path to that file as an argument.

/bin/bash /Users/sdball/bin/shell_script_file

This is a really awesome bit of tooling! It means we can write our “shell” scripts in anything that we can call directly.

Making a script executable

A key step when making your own shell script is that you need to allow your shell to execute the script. You give specific files the permission to run as a script with the special command chmod.

chmod can do a wide range of permissions modifications, but for our local scripts we can keep it (relatively) simple. To make your scripts executable run:

chmod +x script_filename

Experiment!

Time to experiment! Let’s make a file called ls_script.

#!/bin/ls

# this won't happen because ls doesn't know how to run commands
echo "hello"

This silly script will try to run with the ls command as its parser. If what we’ve learned about the shebang is correct then what will actually be run is /bin/ls /Users/sdball/bin/ls_script. Since that file does really exist then we should expect to see the ls command print out that file.

$ chmod +x ls_script
$ ls_script
/Users/sdball/bin/ls_script

It works! And interestingly we can see that the script actually runs slightly differently depending on how we call it.

$ ls_script
/Users/sdball/bin/ls_script

$ cd $HOME/bin
$ ./ls_script
./ls_script

So we can give our script file to any command we can run. That means we can do things like put ruby in the shebang and we’ll end up having our shell automatically run ruby our_awesome_script.

Get weird!

How weird can we get? Let’s change our ls_script file:

#!/bin/rm -f

$ ls_script
$ ls_script
zsh: command not found: ls_script

Well that worked. We just told our script to call rm -f with itself when it’s run. A neat vanishing act that can only happen once!

Writing a real script

We’ve gotten enough of a grounding to kinda know what’s going on in a shell script. Let’s write one!

But what should it do? How about we keep it simple and make a script called uppercase that converts whatever we call it with into uppercase letters?

$ uppercase hello
HELLO

$ uppercase Hello, Shell
HELLO, SHELL

$ uppercase pi: 3.14159
PI: 3.14159

There is a unix command called tr that will be doing the bulk of the work for us in this script. Our script will simply be wrapping a named command around one use of it.

Like all the great unix tools, tr is designed to do one simple job. The classic Unix idea being that you take a bunch of these tools that do one thing and compose them into a data pipeline. It’s a surprisingly effective concept, but that’s a topic for a future post.

The tr command “translates characters”. When given a set of characters like “hello” it can be told to turn all the “l” characters into “1”s, or to delete all the vowels, or shift each character up by one letter.

$ echo "hello" | tr l 1
he11o

$ echo "hello" | tr -d aeiou
hll

$ echo "hello" | tr a-z b-za
ifmmp

The tr command typically takes two arguments: the characters to translate (i.e. replace), followed by the character to translate into. Each character in the first argument matches with its corresponding character in the second argument.

# l translates into 1
$ echo "hello" | tr l 1
he11o

# e translates into 3
# l translates into 1
# o translates into 0
$ echo "hello" | tr elo 310
h3110

# the range of a-y characters translates into b-z
# z translates into a
# (both of these commands are the same translation)
$ echo "hello abc xyz" | tr a-z b-za
ifmmp bcd yza

$ echo "hello abc xyz" | tr a-yz b-za
ifmmp bcd yza

In our script, we’ll want tr to replace all lowercase letters with uppercase.

$ echo "hello" | tr a-z A-Z
HELLO

Let’s write it! Open up your favorite text editor and create a new file called uppercase:

#!/bin/bash

echo "$@" | tr '[:lower:]' '[:upper:]'

Using [:lower:] and [:upper:] supports accents and non-English alphabets.

That’s it! Let’s run it first and then we’ll step through it.

$ chmod +x uppercase

$ ./uppercase hello 123
HELLO 123

$ ./uppercase Hello, Shell
HELLO, SHELL

It works! What does it do? Let’s look at each part:

#!/bin/bash

The shebang says to run this file with the /bin/bash command. That means we’re writing a bash script.

echo $@

echo $@ says to echo out all of the the arguments that the command is given. If we’d used, say, echo $1 then only the first word would be processed.

|

The | character says to “pipe” the data from the echo command into the following command. At a low level: STDOUT from the first command is being redirected into the STDIN of the second command.

tr '[:lower:]' '[:upper:]'

Finally, our trusty tr command is being told to turn all the lowercase characters into their corresponding uppercase characters.

And we’re done! You’ve written a bash shell script that actually does a thing! It’s a simple thing, sure, but it’s better to have a bunch of scripts that do one simple thing than to have one gigantic script that can do anything. After all, if we had just one gigantic script then we might as well just use the tr command directly.

Never underestimate the power of giving a name (e.g. uppercase) to a simple piece of work.