Bash Shell Scripting/Conditional Expressions

Very often, we want to run a certain command only if a certain condition is met. For example, we might want to run the command cp source.txt destination.txt ("copy the file source.txt to location destination.txt") if, and only if, source.txt exists. We can do that like this:

#!/bin/bash

if [[ -e source.txt ]] ; then
  cp source.txt destination.txt
fi

The above uses two built-in commands:

• The construction [[ condition ]] returns an exit status of zero (success) if condition is true, and a nonzero exit status (failure) if condition is false. In our case, condition is -e source.txt, which is true if and only if there exists a file named source.txt.
• The construction

  if command1 ; then
  command2
fi

  < first runs command1; if that completes successfully (that is, if its exit status is zero), then it goes on to run command2.

In other words, the above is equivalent to this:

#!/bin/bash

[[ -e source.txt ]] && cp source.txt destination.txt

except that it is more clear (and more flexible, in ways that we will see shortly).

In general, Bash treats a successful exit status (zero) as meaning "true" and a failed exit status (nonzero) as meaning "false", and vice versa. For example, the built-in command true always "succeeds" (returns zero), and the built-in command false always "fails" (returns one).

if statements are more flexible than what we saw above; we can actually specify multiple commands to run if the test-command succeeds, and in addition, we can use an else clause to specify one or more commands to run instead if the test-command fails:

#!/bin/bash

if [[ -e source.txt ]] ; then
  echo 'source.txt exists; copying to destination.txt.'
  cp source.txt destination.txt
else
  echo 'source.txt does not exist; exiting.'
  exit 1 # terminate the script with a nonzero exit status (failure)
fi

The commands can even include other if statements; that is, one if statement can be "nested" inside another. In this example, an if statement is nested inside another if statement's else clause:

#!/bin/bash

if [[ -e source1.txt ]] ; then
  echo 'source1.txt exists; copying to destination.txt.'
  cp source1.txt destination.txt
else
  if [[ -e source2.txt ]] ; then
    echo 'source1.txt does not exist, but source2.txt does.'
    echo 'Copying source2.txt to destination.txt.'
    cp source2.txt destination.txt
  else
    echo 'Neither source1.txt nor source2.txt exists; exiting.'
    exit 1 # terminate the script with a nonzero exit status (failure)
  fi
fi

This particular pattern — an else clause that contains exactly one if statement, representing a fallback-test — is so common that Bash provides a convenient shorthand notation for it, using elif ("else-if") clauses. The above example can be written this way:

#!/bin/bash

if [[ -e source1.txt ]] ; then
  echo 'source1.txt exists; copying to destination.txt.'
  cp source1.txt destination.txt
elif [[ -e source2.txt ]] ; then
  echo 'source1.txt does not exist, but source2.txt does.'
  echo 'Copying source2.txt to destination.txt.'
  cp source2.txt destination.txt
else
  echo 'Neither source1.txt nor source2.txt exists; exiting.'
  exit 1 # terminate the script with a nonzero exit status (failure)
fi

A single if statement can have any number of elif clauses, representing any number of fallback conditions.

Lastly, sometimes we want to run a command if a condition is false, without there being any corresponding command to run if the condition is true. For this we can use the built-in ! operator, which precedes a command; when the command returns zero (success or "true"), the ! operator changes returns a nonzero value (failure or "false"), and vice versa. For example, the following statement will copy source.txt to destination.txt unless destination.txt already exists:

#!/bin/bash

if ! [[ -e destination.txt ]] ; then
  cp source.txt destination.txt
fi

All those examples above are examples using the test expressions. Actually if just runs everything in then when the command in the statement returns 0:

# First build a function that simply returns the code given
returns() { return $*; }
# Then use read to prompt user to try it out, read `help read' if you have forgotten this.
read -p "Exit code:" exit
if (returns $exit)
  then echo "true, $?"
  else echo "false, $?"
fi

So the behavior of if is quite like the logical 'and' && and 'or' || in some ways:

# Let's reuse the returns function.
returns() { return $*; }
read -p "Exit code:" exit

# if (        and                 ) else            fi
returns $exit && echo "true, $?" || echo "false, $?"

# The REAL equivalent, false is like `returns 1'
# Of course you can use the returns $exit instead of false.
# (returns $exit ||(echo "false, $?"; false)) && echo "true, $?"

Always notice that misuse of those logical operands may lead to errors. In the case above, everything was fine because plain echo is almost always successful.

In addition to the -e file condition used above, which is true if file exists, there are quite a few kinds of conditions supported by Bash's [[ … ]] notation. Five of the most commonly used are:

-d file

True if file exists and is a directory.

-f file

True if file exists and is a regular file.

-e file

True if file exists, whatever it is.

string == pattern

True if string matches pattern. (pattern has the same form as a pattern in filename expansion; for example, unquoted * means "zero or more characters".)

string != pattern

True if string does not match pattern.

string =~ regexp

True if string contains Posix extended regular expression regexp. See Regular_Expressions/POSIX-Extended_Regular_Expressions for more information.

In the last three types of tests, the value on the left is usually a variable expansion; for example, [[ "$var" = 'value' ]] returns a successful exit status if the variable named var contains the value value.

The above conditions just scratch the surface; there are many more conditions that examine files, a few more conditions that examine strings, several conditions for examining integer values, and a few other conditions that don't belong to any of these groups.

One common use for equality tests is to see if the first argument to a script ($1) is a special option. For example, consider our if statement above that tries to copy source1.txt or source2.txt to destination.txt. The above version is very "verbose": it generates a lot of output. Usually we don't want a script to generate quite so much output; but we may want users to be able to request the output, for example by passing in --verbose as the first argument. The following script is equivalent to the above if statements, but it only prints output if the first argument is --verbose:

#!/bin/bash

if [[ "$1" == --verbose ]] ; then
  verbose_mode=TRUE
  shift # remove the option from $@
else
  verbose_mode=FALSE
fi

if [[ -e source1.txt ]] ; then
  if [[ "$verbose_mode" == TRUE ]] ; then
    echo 'source1.txt exists; copying to destination.txt.'
  fi
  cp source1.txt destination.txt
elif [[ -e source2.txt ]] ; then
  if [[ "$verbose_mode" == TRUE ]] ; then
    echo 'source1.txt does not exist, but source2.txt does.'
    echo 'Copying source2.txt to destination.txt.'
  fi
  cp source2.txt destination.txt
else
  if [[ "$verbose_mode" == TRUE ]] ; then
    echo 'Neither source1.txt nor source2.txt exists; exiting.'
  fi
  exit 1 # terminate the script with a nonzero exit status (failure)
fi

Later, when we learn about shell functions, we will find a more compact way to express this. (In fact, even with what we already know, there is a more compact way to express this: rather than setting $verbose_mode to TRUE or FALSE, we can set $echo_if_verbose_mode to echo or :, where the colon : is a Bash built-in command that does nothing. We can then replace all uses of echo with "$echo_if_verbose_mode". A command such as "$echo_if_verbose_mode" message would then become echo message, printing message, if verbose-mode is turned on, but would become : message, doing nothing, if verbose-mode is turned off. However, that approach might be more confusing than is really worthwhile for such a simple purpose.)

To combine multiple conditions with "and" or "or", or to invert a condition with "not", we can use the general Bash notations we've already seen. Consider this example:

#!/bin/bash

if [[ -e source.txt ]] && ! [[ -e destination.txt ]] ; then
  # source.txt exists, destination.txt does not exist; perform the copy:
  cp source.txt destination.txt
fi

The test-command [[ -e source.txt ]] && ! [[ -e destination.txt ]] uses the && and ! operators that we saw above that work based on exit status. [[ condition ]] is "successful" if condition is true, which means that [[ -e source.txt ]] && ! [[ -e destination.txt ]] will only run ! [[ -e destination.txt ]] if source.txt exists. Furthermore, ! inverts the exit status of [[ -e destination.txt ]], so that ! [[ -e destination.txt ]] is "successful" if and only if destination.txt doesn't exist. The end result is that [[ -e source.txt ]] && ! [[ -e destination.txt ]] is "successful" — "true" — if and only if source.txt does exist and destination.txt does not exist.

The construction [[ … ]] actually has built-in internal support for these operators, such that we can also write the above this way:

#!/bin/bash

if [[ -e source.txt && ! -e destination.txt ]] ; then
  # source.txt exists, destination.txt does not exist; perform the copy:
  cp source.txt destination.txt
fi

but the general-purpose notations are often more clear; and of course, they can be used with any test-command, not just the [[ … ]] construction.

The if statements in the above examples are formatted to make them easy for humans to read and understand. This is important, not only for examples in a book, but also for scripts in the real world. Specifically, the above examples follow these conventions:

• The commands within an if statement are indented by a consistent amount (by two spaces, as it happens). This indentation is irrelevant to Bash — it ignores whitespace at the beginning of a line — but is very important to human programmers. Without it, it is hard to see where an if statement begins and ends, or even to see that there is an if statement. Consistent indentation becomes even more important when there are if statements nested within if statements (or other control structures, of various kinds that we will see).
• The semicolon character ; is used before then. This is a special operator for separating commands; it is mostly equivalent to a line-break, though there are some differences (for example, a comment always runs from # to the end of a line, never from # to ;). We could write then at the beginning of a new line, and that is perfectly fine, but it's good for a single script to be consistent one way or the other; using a single, consistent appearance for ordinary constructs makes it easier to notice unusual constructs. In the real world, programmers usually put ; then at the end of the if or elif line, so we have followed that convention here.
• A newline is used after then and after else. These newlines are optional — they need not be (and cannot be) replaced with semicolons — but they promote readability by visually accentuating the structure of the if statement.
• Regular commands are separated by newlines, never semicolons. This is a general convention, not specific to if statements. Putting each command on its own line makes it easier for someone to "skim" the script and see roughly what it is doing.

These exact conventions are not particularly important, but it is good to follow consistent and readable conventions for formatting your code. When a fellow programmer looks at your code — or when you look at your code two months after writing it — inconsistent or illogical formatting can make it very difficult to understand what is going on.