Where I work, we have tons of static html files that are published using our custom built Content Management System. Sometimes, we have to change a single line on every single page. For example, the latest case involved a change involving the size of ads that were on the pages. Rather than use publishing and database resources just to make this relatively simple change, I use the power of the shell. Specifically, I use find, sed and a shell script that I wrote. It's based on a shell script in the Unix Power Tools book.

First off, we need to find all the files. I usually accomplish this with something simple like

$ find /www -type f -name "*.html" -print

This simply finds that end with .html in the /www directory. You can do more complex things with find like find all files modified in the last 2 days by Frank if you needed to change files like that.

The results of this will be passed off to a file named replace.sh located in my ~/bin directory. replace.sh is reproduced here:

echo -n "editing $1: "
if test "$1" = sedscr; then
    echo -n "Not editing sedscript!"
elif test -s $1; then
         sed -f sedscr $1 > $temp
    if test -s "$temp"; then
        if cmp -s "$1" $temp; then
            echo -n "FILE NOT CHANGED: "
            # save original, just in case
            # mv $1 $1.bak
            cp $temp "$1"
        echo -n "done"
        echo -n "Sed produced an empty file \
- check your sedscript".
rm -f $temp

So, the command we would run would now look like this:

$ find /www -type f \
      -name "*.html" \
      -exec ~/bin/replace.sh {} \;

This does the same as above but passes each file found by the find command above to the ~/bin/replace.sh script.

You'll notice that the replace.sh file calls sed using a file named sedscr. The next step is creating the sedscr file.

The sedscr files simple contains sed commands. It must exist in the same directory that you call the find command above from. Here's a sample sedscr that just does a simple replace.


You can enter in as many complex sed commands as you want. It's sed so the power is there!

This simply replaces all instances of BigHonkingAd with NiceSmallAd in each of your files found by the find command. The nice thing about the replace.sh script is that it will not edit the file if the contents of your sedscr don't produce an altered file. Also, if you want the replace.sh to make a backup of your original file, just uncomment the mv line.

Using this methodology, I'm able to edit about 5000 files a minute. It could probably be faster if I used xargs and the output of replace.sh is a little verbose but this solution has worked for me for years and if it ain't broke, why fix it?