Using Arguments And Variables That Contain Spaces

Take a second look at the script from the previous section:

#!/bin/sh

FIRST_ARGUMENT="$1"

echo "Hello, world $FIRST_ARGUMENT!"

Notice that the echo statement is followed by a string surrounded by quotation marks. Normally, the shell uses spaces to separate arguments to commands. Outside of quotation marks, the shell would treat “Hello,” and “world” as separate arguments to echo.

By surrounding the string with double quote marks, the shell treats the entire string as a single argument to echo even though it contains spaces.

To see how this works, save the script above as test.sh (if you haven’t already), then type the following commands:

./test.sh leaders and citizens

./test.sh "leaders and citizens"

The first line above prints “Hello, world leaders!” because the space after “leaders” ends the first argument ($1). Inside the script, the variable $1 contains “leaders”, $2 contains “and”, and $3 contains “citizens”.

The second line above prints “Hello, world leaders and citizens!” because the quotation marks on the command line cause everything within them to be grouped as a single argument.

Notice also that there are similar quotation marks on the right side of the assignment statement:

FIRST_ARGUMENT="$1"

With most modern shells, these double quotation marks are not required for this particular assignment statement (because there are no literal spaces on the right side), but they are a good idea for maximum compatibility. See Historical String Parsing in Historical Footnotes and Arcana to learn why.

When assigning literal strings (rather than variables containing strings) to a variable, however, you must surround any spaces with quotation marks. For example, the following statement does not do what you might initially suspect:

STRING2=This is a test

If you type this statement, the Bourne shell gives you an error like this:

sh: is: command not found

The reason for this seemingly odd error is that the assignment statement ends at the first space, so the next word after that statement is interpreted as a command to execute. See Overriding Environment Variables for Child Processes (Bourne Shell) for more details.

Instead, write this statement as:

STRING2="This is a test"

Using quotation marks is particularly important when working with variables that contain filenames or paths. For example, type the following commands:

mkdir "/tmp/My Folder"

FILENAME="/tmp/My Folder"

ls "$FILENAME"

ls $FILENAME

The above example creates a directory in /tmp called “My Folder”. (Don’t worry about deleting it because /tmp gets wiped every time you reboot.) It then attempts to list the files in that directory. The first time, it uses quotation marks. The second time, it does not. Notice that the shell misinterprets the command the second time as being an attempt to list the files in /tmp/My and the files in Folder.

Handling Quotation Marks in Strings

In modern Bourne shells, expansion of variables, occurs after the statement itself is fully parsed by the shell. (See Historical String Parsing in Historical Footnotes and Arcana for more information.) Thus, as long as the variable is enclosed in double quote marks, you do not get any execution errors even if the variable’s value contains double-quote marks.

However, if you are using double quote marks within a literal string, you must quote that string properly. For example:

MYSTRING="The word of the day is \"sedentary\"."

This quoting technique also applies to literal strings within commands entered on the command line. For example, using the script from earlier in Shell Variables and Printing, the command:

./test.sh "\"leaders\""

prints the phrase “Hello, world “leaders”!”

The details of quotes as they apply to variable expansion are explained in Parsing, Variable Expansion, and Quoting. (Variable safety with shells that predate this behavior is generally impractical. Fortunately, the modern behavior has been the norm since the mid-1990s.)

Shell scripts also allow the use of single quote marks. Variables between single quotes are not replaced by their contents. Be sure to use double quotes unless you are intentionally trying to display the actual name of the variable. You can also use single quotes as a way to avoid the shell interpreting the contents of the string in any way. These differences are described further in Parsing, Variable Expansion, and Quoting.

Using the export Builtin (Bourne Shell)

Generally speaking, the first time you assign a value to an environment variable such as the PATH variable, the Bourne shell creates a new, local copy of this shell variable that is specific to your script. Any tool executed from your script is passed the original value of PATH inherited from whatever script, tool, or shell that launched it.

With the BASH shell, however, any variable inherited from the environment is automatically exported by the shell. Thus, in some versions of OS X, if you modify inherited environment variables (such as PATH) in a script, your local changes will be seen automatically by any tool or script that your script executes. Thus, in these versions of OS X, you do not have to explicitly use the export statement when modifying the PATH variable.

Because different Bourne shell variants handle these external environment variables differently (even among different versions of OS X), this creates two minor portability problems:

• A script written without the export statement may work on some versions of OS X, but will fail on others. You can solve this portability problem by using the export builtin, as described in this section.

• A shell script that changes variables such as PATH will alter the behavior of any script that it executes, which may or may not be desirable. You can solve this problem by overriding the PATH environment variable when you execute each individual tool, as described in Overriding Environment Variables for Child Processes (Bourne Shell).

To guarantee that your modifications to a shell variable are passed to any script or tool that your shell script calls, you must use the export builtin. You do not have to use this command every time you change the value; the variable remains exported until the shell script exits.

For example:

export PATH="/usr/local/bin:$PATH"

# or

PATH="/usr/local/bin:$PATH"

export PATH

Either of these statements has the same effect—specifically, they export the local notion of the PATH environment variable to any command that your script executes from now on. There is a small catch, however. You cannot later undo this export to restore the original global declaration. Thus, if you need to retain the original value, you must store it somewhere yourself.

In the following example, the script stores the original value of the PATH environment variable, exports an altered version, executes a command, and restores the old version.

ORIGPATH="$PATH"

PATH="/usr/local/bin:$PATH"

export PATH

# Execute some command here---perhaps a

# modified ls command....

ls

PATH="$ORIGPATH"

If you need to find out whether an environment variable (whether inherited by your script or explicitly set with the export directive) was set to empty or was never set in the first place, you can use the printenv command to obtain a complete list of defined variables and use grep to see if it is in the list. (You should note that although printenv is a csh builtin, it is also a standalone command in /usr/bin.)

For example:

DEFINED=`printenv | grep -c '^VARIABLE='`

The resulting variable will contain 1 if the variable is defined in the environment or 0 if it is not.

Overriding Environment Variables for Child Processes (Bourne Shell)

Because the BASH Bourne shell variant automatically exports all variables inherited from its environment, any changes you make to preexisting environment variables such as PATH are automatically inherited by any tool or script that your script executes. (This is not true for other Bourne shell variants; see Using the export Builtin (Bourne Shell) for further explanation.)

While automatic export is usually convenient, you may sometimes wish to change a preexisting environment variable without modifying the environment of any script or tool that your script executes. For example, if your script executes a number of tools in /usr/local/bin, it may be convenient to change the value of PATH to include /usr/local/bin. However, you may not want child processes to also look in /usr/local/bin.

This problem is easily solved by overriding the environment variable PATH on a per-execution basis. Consider the following script:

#!/bin/sh

echo $MYVAR

This script prints the value of the variable MYVAR. Normally, this variable is empty, so this script just prints a blank line. Save the script as printmyvar.sh, then type the following commands:

chmod a+x printmyvar.sh      # makes the script executable

MYVAR=7 ./printmyvar.sh      # runs the script

echo "MYVAR IS $MYVAR"       # prints the variable

Notice that the assignment statement MYVAR=7 applies only to the command that follows it. The value of MYVAR is altered in the environment of the command ./printmyvar.sh, so the script prints the number 7. However, the original (empty) value is restored after executing that command, so the echo statement afterwards prints an empty string for the value of MYVAR.

Thus, to modify the PATH variable locally but execute a command with the original PATH value, you can write a script like this:

#!/bin/sh

GLOBAL_PATH="$PATH"

PATH=/usr/local/bin

PATH="$GLOBAL_PATH" /bin/ls

Using the setenv Builtin (C shell)

In the C shell, variables are exported if you set them with setenv, but not if you set them with set. Thus, if you want your shell variable modifications to be seen by any tool or script that you call, you should use the setenv builtin. This builtin is the C shell equivalent to issuing an assignment statement with the export builtin in the Bourne shell.

setenv VALUE "Four"

echo "VALUE is '$VALUE'."

If you want your shell variables to only be available to your script, you should use the set builtin (described in Shell Variables and Printing). The set builtin is equivalent to a simple assignment statement in the Bourne shell.

set VALUE = "Four"

echo "VALUE is '$VALUE'."

Notice that the local variable version requires an equals sign (=), but the exported environment version does not (and produces an error if you put one in).

To remove variables in the C shell, you can use the unsetenv or unset builtin. For example:

setenv VALUE "Four"

unsetenv VALUE

set VALUE = "Four"

unset VALUE

echo "VALUE is '$VALUE'."

This will generate an error message. In the C shell, it is not possible to print the value of an undefined variable, so if you think you may need to print the value later, you should set it to an empty string rather than using unset or unsetenv.

If you need to test an environment variable (not a shell-local variable) that may or may not be part of your environment (a variable set by whatever process called your script), you can use the printenv builtin. This prints the value of a variable if set, but prints nothing if the variable is not set, and thus behaves just like the variable behaves in the Bourne shell.

For example:

set X = `printenv VALUE`

echo "X is "\"$X\"

This prints X is "" if the variable is either empty or undefined. Otherwise, it prints the value of the variable between the quotation marks.

If you need to find out if a variable is simply empty or is actually not set, you can also use printenv to obtain a complete list of defined variables and use grep to see if it is in the list. For example:

set DEFINED = `printenv | grep -c '^VARIABLE='`

The resulting variable will contain 1 if the variable is defined in the environment or 0 if it is not.

Overriding Environment Variables for Child Processes (C Shell)

Unlike the Bourne shell, the C shell does not provide a built-in syntax for overriding environment variables when executing external commands. However, it is possible to simulate this either by using the env command.

The best and simplest way to do this is with the env command. For example:

env PATH="/usr/local/bin" /bin/ls

As an alternative, you can use the set builtin to make a temporary copy of any variable you need to override, change the value, execute the command, and restore the value from the temporary copy.

You should notice, however, that whether you use the env command or manually make a copy, the PATH variable is altered prior to searching for the command. Because the PATH variable controls where the shell looks for programs to execute, you must therefore explicitly provide a complete path to the ls command or it will not be found (unless you have a copy in /usr/local/bin, of course). The PATH environment variable is explained in Special Shell Variables.

As a workaround, you can determine the path of the executable using the which command prior to altering the PATH environment variable.

set GLOBAL_PATH = "$PATH"

set LS = `which ls`

setenv PATH "/usr/local/bin"

$LS

setenv PATH "$GLOBAL_PATH"

unset GLOBAL_PATH

Or, using env:

set LS = `which ls`

env PATH='/usr/local/bin' $LS

The use of the backtick (`) operator in this fashion is described in Inline Execution.