How to Find and Replace String with Special Characters Using Sed

Looking for literal strings with a regular expression, when the search-string contains special characters, is sometimes not as simple as looking for patterns, but you can do it with a bit of juggling.

Note: The echo command must cater for CMD-special-characters, so it needs ^^ to escape a single ^ and ^| to escape | ... You don't need CMD's escape-character ^ if you type directly into the file.

Step 1: Create a file, named literal-srch-strings.txt, which containing the exact (unaltered) string to be replaced. There are 2 ways to create this file:

1. As a command issued at CMD's commandline, or as a command in a .cmd/.bat command-script.

   echo sed -i^^/\\*$/$[{" ;"> literal-srch-strings.txt

2. Make literal-srch-strings.txt yourself, in your text editor.
   In this case, you should not use the CMD-escape-character ^, so the line is has just one ^, not ^^ -- This is because you are bypassing the CMD-shell.
   Here is what is needed in the .txt file (just as the filename says :)

   sed -i^/\\*$/$[{" ;"

Step 2: Make a sed script, named str-to-regex.sed , to convert the string(s) into sed regex(s).
Note that the same issue of the CMD-escape-character ^ applies to this step, so again, there are 2 ways you can create the .sed file:

1. As a command:

   echo s/[]\/$*.^^^|[]/\\^&/g; s/.*/s\/^&\/Ciao!\/g/> str-to-regex.sed

2. Using your text editor, make a file named str-to-regex.sed, containing:

   s/[]\/$*.^|[]/\\&/g; s/.*/s\/&\/Ciao!\/g/

Step 3: Run the sed-script which converts the string into a sed regeular expression, and
send its output to another sed-script, replace-text.sed, which will make the actual replacement.

sed -f str-to-regex.sed  literal-srch-strings.txt > replace-text.sed

Step 4: Run replace-text.sed -- For the test we can use literal-srch-strings.txt as the input file, but you can, of course, use any input file.

sed -f replace-text.sed  literal-srch-strings.txt

Here is the output:

Ciao! 

Three things wrong:-

• The first quote in the sed command should be before the s/ option, not before sed itself - I presume this is a typing error.
• The > character has no special meaning in regular expressions, and must not be escaped - the sequence \> has special significance: it means end of word, and because .* is "greedy" it matches the end of the last word on the line, hence the retention of the #".
• If you match the source >, this will be included in the string to be replaced, so it must also appear in the replacement string.

So your edit command should be:

sed 's/PackingList.*>/PackingList>/g' xmlfile

This is similar to jherran's solution, but takes account of your original attempt at matching. It might be neater to match up to the trailing double-quote:

sed 's/PackingList.*"/PackingList/g' xmlfile

If you don't want to rely on greediness (and make it more readable), use:

sed 's/PackingList.*".*"/PackingList/g' xmlfile

Note that any subsequent XML tags on the same line may be deleted by any of the above: to avoid this, use:

sed 's/PackingList[^>]*"[^>]*"/PackingList/g' xmlfile