Isn't it a pain to have to get the contents of a variable, edit it, and put it into another variable?

For example, to update a file's name and contents to a new format, such that given three input files, apache.old, mysql.old and postgres.old, the script should transform them as shown, and rename them from .old to .new:

A classic approach would be as script like this, which transforms the filename, uses sed to update the contents, and write the result to the new file:

#!/bin/bash
for oldfile in *.old
do
  # Replace ".old" ending with ".new" for the new file
  newfile=`echo $oldfile | sed s/".old"/".new"/g`
  echo "Updating ${oldfile}..."
  # replace "old thing" with "new thing" for the new version, and write to .new
  sed s/"old thing"/"new thing"/g $oldfile > $newfile
done

For the purposes of just editing a file to a new version, there is nothing wrong with this approach. But there is a better way, which saves the need to create the extra variable for this one-time use, and makes far more easily readable code:

#!/bin/bash
for oldfile in *.old
do
  echo "Updating ${oldfile}..."
  # replace "old thing" with "new thing" for the new version, and write to .new
  sed s/"old thing"/"new thing"/g $oldfile > ${oldfile/.old/.new}
done

The use of ${oldfile/.old/.new} tells Bash to take the value of $oldfile, such as "apache.old", and replace any occurrences of ".old" within it to ".new".

You will probably have noticed that ${oldfile/.old/.new} is also very similar to the syntax in the sed s/"old thing"/"new thing"/g search-and-replace method used to change "old thing" to "new thing" within the file.

Get this tutorial as a PDF for only $5