Bash Shell Scripting/Shell Functions

A shell function is a special type of variable that is essentially a script-within-a-script. This feature allows us to group a sequence of commands into a single named command, which is particularly useful if the sequence of commands needs to be run from many places within the script. A shell function can even consist of just a single command; this may be useful if the command is particularly complicated, or if its significance would not immediately be obvious to a reader. (That is, shell functions can serve two purposes: they can save typing, and they can allow for more readable code by creating intuitively-named commands.) Consider the following script:

# Usage:     get_password VARNAME
# Asks the user for a password; saves it as $VARNAME.
# Returns a non-zero exit status if standard input is not a terminal, or if the
# "read" command returns a non-zero exit status.
get_password() {
  if [[ -t 0 ]] ; then
    read -r -p 'Password:' -s "$1" && echo
    return 1

get_password PASSWORD && echo "$PASSWORD"

The above script creates a shell function called get_password that asks the user to type a password, and stores the result in a specified variable. It then runs get_password PASSWORD to store the password as $PASSWORD; and lastly, if the call to get_password succeeded (as determined by its exit status), the retrieved password is printed to standard output (which is obviously not a realistic use; the goal here is simply to demonstrate the behavior of get_password).

The function get_password doesn't do anything that couldn't be done without a shell function, but the result is much more readable. The function invokes the built-in command read (which reads a line of user input and saves it in one or more variables) with several options that most Bash programmers will not be familiar with. (The -r option disables a special meaning for the backslash character; the -p option causes a specified prompt, in this case Password:, to be displayed at the head of the line; and the -s option prevents the password from being displayed as the user types it in. Since the -s option also prevents the user's newline from being displayed, the echo command supplies a newline.) Additionally, the function uses the conditional expression -t 0 to make sure that the script's input is coming from a terminal (a console), and not from a file or from another program that wouldn't know that a password is being requested. (This last feature is debatable; depending on the general functionality of the script, it may be better to accept a password from standard input regardless of its source, under the assumption that the source was designed with the script in mind.) The overall point is that giving sequence of commands a name — get_password — makes it much easier for a programmer to know what it does.

Within a shell function, the positional parameters ($1, $2, and so on, as well as $@, $*, and $#) refer to the arguments that the function was called with, not the arguments of the script that contains the function. If the latter are needed, then they need to be passed in explicitly to the function, using "$@". (Even then, shift and set will only affect the positional parameters within the function, not those of the caller.)

A function call returns an exit status, just like a script (or almost any command). To explicitly specify an exit status, use the return command, which terminates the function call and returns the specified exit status. (The exit command cannot be used for this, because it would terminate the entire script, just as if it were called from outside a function.) If no exit status is specified, either because no argument is given to the return command or because the end of the function is reached without having run a return command, then the function returns the exit status of the last command that was run.

Incidentally, either function or ( ) may be omitted from a function declaration, but at least one must be present. Instead of function get_password ( ), many programmers write get_password(). Similarly, the { … } notation is not exactly required, and is not specific to functions; it is simply a notation for grouping a sequence of commands into a single compound command. The body of a function must be a compound command, such as a { … } block or an if statement; { … } is the conventional choice, even when all it contains is a single compound command and so could theoretically be dispensed with.