find/replace with sed yields unexpected results

Question

I'm attempting to find/replace a string found in a number of files spread throughout a directory tree.

I thought I'd use sed for this task.

find /some/dir -name \*.php -exec sed -i "s/cybernetnews/cybernet/g" {} \;

This works as expected with the sample string "cybernetnews", but things get weird when I try to find/replace against this string:

preg_match($_SESSION['infos']['num'], $numeric_variable)

Where I want to replace the string $_SESSION['infos']['num'] with the constant INFOS_NUM.

So I thought this would work:

find /some/dir -name \*.php -exec sed -i "s/$_SESSION['infos']['num']/INFOS_NUM/g" {} \;

No luck, here is what shows in the test php file after running the above command:

preg_match($_SESSION['INFOS_NUMfoINFOS_NUM][INFOS_NUMum'], 
           $INFOS_NUMmeric_variable)

I tried escaping the single quote but get the exact same result. If I escape the bracket, no changes are made.

What character(s) is sed getting hung up on, and how do I properly escape it?

Answer 1

$, [, ] are regex special characters , you must need to escape that in the regex in-order to match those characters literally. To do a case-insensitive match, you need to add i modifier at the last.

find /some/dir -name \*.php -exec sed -i "s/\$_SESSION\['infos'\]\['num'\]/INFOS_NUM/g" {} \;