Expansion

Each time we type a command line and press the enter key, bash performs several processes upon the text before it carries out our command. We have seen a couple of cases of how a simple character sequence, for example “*”, can have a lot of meaning to the shell. The process that makes this happen is called expansion. With expansion, we type something and it is expanded into something else before the shell acts upon it. To demonstrate what we mean by this, let's take a look at the echo command. echo is a shell builtin that performs a very simple task. It prints out its text arguments on standard output:

That's pretty straightforward. Any argument passed to echo gets displayed. Let's try another example:

So what just happened? Why didn't echo print “*”? As we recall from our work with wildcards, the “*” character means match any characters in a filename, but what we didn't see in our original discussion was how the shell does that. The simple answer is that the shell expands the “*” into something else (in this instance, the names of the files in the current working directory) before the echo command is executed. When the enter key is pressed, the shell automatically expands any qualifying characters on the command line before the command is carried out, so the echo command never saw the “*”, only its expanded result. Knowing this, we can see that echo behaved as expected.

Pathname Expansion

The mechanism by which wildcards work is called pathname expansion. If we try some of the techniques that we employed in our earlier lessons, we will see that they are really expansions. Given a home directory that looks like this:

we could carry out the following expansions:

and:

or even:

and looking beyond our home directory:

Tilde Expansion

As we recall from our introduction to the cd command, the tilde character (“~”) has a special meaning. When used at the beginning of a word, it expands into the name of the home directory of the named user, or if no user is named, the home directory of the current user:

If user “foo” has an account, then:

Arithmetic Expansion

The shell allows arithmetic to be performed by expansion. This allow us to use the shell prompt as a calculator:

Arithmetic expansion uses the form:

$((expression))

where expression is an arithmetic expression consisting of values and arithmetic operators.

Arithmetic expansion only supports integers (whole numbers, no decimals), but can perform quite a number of different operations.

Spaces are not significant in arithmetic expressions and expressions may be nested. For example, to multiply five squared by three:

Single parentheses may be used to group multiple subexpressions. With this technique, we can rewrite the example above and get the same result using a single expansion instead of two:

Here is an example using the division and remainder operators. Notice the effect of integer division:

Brace Expansion

Perhaps the strangest expansion is called brace expansion. With it, we can create multiple text strings from a pattern containing braces. Here's an example:

Patterns to be brace expanded may contain a leading portion called a preamble and a trailing portion called a postscript. The brace expression itself may contain either a comma-separated list of strings, or a range of integers or single characters. The pattern may not contain embedded whitespace. Here is an example using a range of integers:

A range of letters in reverse order:

Brace expansions may be nested:

So what is this good for? The most common application is to make lists of files or directories to be created. For example, if we were a photographer and had a large collection of images we wanted to organize into years and months, the first thing we might do is create a series of directories named in numeric “Year-Month” format. This way, the directory names will sort in chronological order. we could type out a complete list of directories, but that's a lot of work and it's error-prone too. Instead, we could do this:

Pretty slick!

Parameter Expansion

We're only going to touch briefly on parameter expansion in this lesson, but we'll be covering it more later. It's a feature that is more useful in shell scripts than directly on the command line. Many of its capabilities have to do with the system's ability to store small chunks of data and to give each chunk a name. Many such chunks, more properly called variables, are available for our examination. For example, the variable named “USER” contains our user name. To invoke parameter expansion and reveal the contents of USER we would do this:

To see a list of available variables, try this:

With other types of expansion, if we mistype a pattern, the expansion will not take place and the echo command will simply display the mistyped pattern. With parameter expansion, if we misspell the name of a variable, the expansion will still take place, but will result in an empty string:

Command Substitution

Command substitution allows us to use the output of a command as an expansion:

A clever one goes something like this:

Here we passed the results of which cp as an argument to the ls command, thereby getting the listing of of the cp program without having to know its full pathname. We are not limited to just simple commands. Entire pipelines can be used (only partial output shown):

In this example, the results of the pipeline became the argument list of the file command. There is an alternate syntax for command substitution in older shell programs which is also supported in bash. It uses back-quotes instead of the dollar sign and parentheses:

Quoting

Now that we've seen how many ways the shell can perform expansions, it's time to learn how we can control it. Take for example:

or:

In the first example, word-splitting by the shell removed extra whitespace from the echo command's list of arguments. In the second example, parameter expansion substituted an empty string for the value of “$1” because it was an undefined variable. The shell provides a mechanism called quoting to selectively suppress unwanted expansions.

Double Quotes

The first type of quoting we will look at is double quotes. If we place text inside double quotes, all the special characters used by the shell lose their special meaning and are treated as ordinary characters. The exceptions are “$”, “\” (backslash), and “`” (back- quote). This means that word-splitting, pathname expansion, tilde expansion, and brace expansion are suppressed, but parameter expansion, arithmetic expansion, and command substitution are still carried out. Using double quotes, we can cope with filenames containing embedded spaces. Imagine we were the unfortunate victim of a file called two words.txt. If we tried to use this on the command line, word-splitting would cause this to be treated as two separate arguments rather than the desired single argument:

By using double quotes, we can stop the word-splitting and get the desired result; further, we can even repair the damage:

There! Now we don't have to keep typing those pesky double quotes. Remember, parameter expansion, arithmetic expansion, and command substitution still take place within double quotes:

We should take a moment to look at the effect of double quotes on command substitution. First let's look a little deeper at how word splitting works. In our earlier example, we saw how word-splitting appears to remove extra spaces in our text:

By default, word-splitting looks for the presence of spaces, tabs, and newlines (linefeed characters) and treats them as delimiters between words. This means that unquoted spaces, tabs, and newlines are not considered to be part of the text. They only serve as separators. Since they separate the words into different arguments, our example command line contains a command followed by four distinct arguments. If we add double quotes:

word-splitting is suppressed and the embedded spaces are not treated as delimiters, rather they become part of the argument. Once the double quotes are added, our command line contains a command followed by a single argument. The fact that newlines are considered delimiters by the word-splitting mechanism causes an interesting, albeit subtle, effect on command substitution. Consider the following:

In the first instance, the unquoted command substitution resulted in a command line containing thirty-eight arguments. In the second, a command line with one argument that includes the embedded spaces and newlines.

Single Quotes

When we need to suppress all expansions, we use single quotes. Here is a comparison of unquoted, double quotes, and single quotes:

As we can see, with each succeeding level of quoting, more and more of the expansions are suppressed.

Escaping Characters

Sometimes we only want to quote a single character. To do this, we can precede a character with a backslash, which in this context is called the escape character. Often this is done inside double quotes to selectively prevent an expansion:

It is also common to use escaping to eliminate the special meaning of a character in a filename. For example, it is possible to use characters in filenames that normally have special meaning to the shell. These would include “$”, “!”, “&”, “ “, and others. To include a special character in a filename we can to this:

To allow a backslash character to appear, escape it by typing “\\”. Note that within single quotes, the backslash loses its special meaning and is treated as an ordinary character.

More Backslash Tricks

If we look at the man pages for any program written by the GNU project, we will see that in addition to command line options consisting of a dash and a single letter, there are also long option names that begin with two dashes. For example, the following are equivalent:

Why do they support both? The short form is for lazy typists on the command line and the long form is mostly for scripts though some options may only be available in long form. Sometimes it is better to use a long option when the option is obscure or we want to document more clearly what an option is. This is especially useful when writing scripts where maximum readability is desired, and besides, anytime we can save ourselves a trip to the man page is a good thing.

As we might suspect, using the long form options can make a single command line very long. To combat this problem, we can use a backslash to get the shell to ignore a newline character like this:

Using the backslash in this way allows us to embed newlines in our command. Note that for this trick to work, the newline must be typed immediately after the backslash. If we put a space after the backslash, the space will be ignored, not the newline. Backslashes are also used to insert special characters into our text. These are called backslash escape characters. Here are the common ones:

The use of the backslash escape characters is very common. This idea first appeared in the C programming language. Today, the shell, C++, Perl, python, awk, tcl, and many other programming languages use this concept. Using the echo command with the -e option will allow us to demonstrate: