45.34 Arrays in the Bourne Shell

The C shell (47.5), awk (33.11), the Korn shell, and some other UNIX command interpreters have built-in array support. The standard Bourne shell doesn't, though its command line is a sort-of array that you can store with the set (44.19) command - and get stored values through $1, $2, etc.

You can store and use Bourne shell variables - with names like array1, array2, and so on - to simulate an array with elements 1, 2, and so on. The eval (8.10) command does the trick. As an example, if the n shell variable stores the array index (1, 2, etc.), you can store an element of the array named part with:

eval part$n="value"

and use its value with:

eval echo "The part is \$part$n."

You need the extra quoting in that last command because eval scans the command line twice. The really important part is \$part$n-on the first pass, the shell interprets $n, strips off the backslash, and leaves a line like:

echo "The part is $part5."

The next pass gives the value of the part5 variable.

To store a line of text with multiple words into these fake array elements, the set command won't work. A for loop (44.16) usually will. For example, to read a line of text into the temp variable and store it in an "array" named part:

expr	echo "Enter the line: \c" read temp n=0 for word in $temp do n=`expr $n + 1` eval part$n="$word" done

The first word from $temp goes into the variable part1, the second into part2, and so on.