sed: How to Escape Forward Slash with the Right Delimiter
/* Posted November 11th, 2008 at 10:02am *//* Filed under How-To, Linux */
/* */
If you’ve used Linux at all, you may have come across ’sed,’ short for "stream editor," which is a powerful little utility to modify streams like console IO, files or text. I was using it recently to do a simple search and replace operation in a few files. However, a curious problem arose that I had not encountered before. I’ve used ’sed’ many a time to perform a find and replace on text within a file, but I’d never included the slash ‘/’ (also known as "forward slash") character in my search text with sed. As you may know, sed can easily perform a search and replace on all *.txt files in a directory with this command:
% sed s/text_to_replace/replacement_text/g *.txt
Note how forward slash "/" is used as part of the regular expression to separate the command options and search text. Naturally, since I had text that I wanted to replace with forward slash characters, my sed command contained escaped slash characters and looked like this:
% sed s/file:\/\//http:\/\//g *.txt
I wanted to replace "file://" with "http://" text so in order not to confuse the delimiter with the text, I escaped the forward slashes with backslashes like so \/. Unfortunately, sed didn’t like what I gave it and gave and output like:
sed: -e expression #1, char 10: unknown option to `s'
No matter how I escaped the forward slash or how many other characters I escaped with backslash, sed simply would not do what I wanted it to do. Then I discovered quite an exciting little tidbit about sed. In the regular expression, it’s not necessary to delimit the find and replace texts and search options with the slash ‘/’ character. In fact, you can use just about any character to delimit the expression:
% sed s#file:\/\/#http:\/\/#g *.txt Using #, @, and A as the delimiter will all work out to the same valid sed operation to find all strings of "file://" text and replace it with "http://" text.
% sed s@file:\/\/@http:\/\/@g *.txt
% sed sAfile:\/\/Ahttp:\/\/Ag *.txt
Image/Flickr/Biappi
Related Posts
Related Stories
New SQL Truncation Attacks And How To Avoid Them
(Programming) This article reviews some new ideas that can result in either modifying SQL statements or injecting SQL code even if the code has escaped the delimiting characters. Starts with some best practices for constructing delimited identifiers and SQL literals, and then shows ways attackers can inject SQL code in applicationsDugg 15 timesDefining one’s own string de-limiter in Oracle 10g
(Programming) Oracle 10g provided one new enhancement for string literals. We can define our own string delimiter character. Normally, when a string containing a single quote needs to be assigned to some variable, we have to put an extra single quote in front of the existing single quote of the string - that acts as the escape character. Example:Dugg 2 times
Subscribe via email
on 
on 
on 
on 
on 
on 
Incidentally sed works a bit differently in a tclsh shell (incase it is used as part of a tcl script)
The following code works! No need to use use backslash.
exec echo $abc | sed “s,file://,http://,g”
Excellent post! This really helped me out a lot. I would have never thought of changes the delimiters!
I managed to escape slashes and then use my new escaped VAR in another sed expr:
ESCAPED_VAR = echo “$SITE_PATH” | sed ’s/\//\\\//g’
(append new line after some_string)
sed “/^some_string /a\ $ESCAPED_VAR” file_old > file_new
Thx
Please delete my previous post. This one works:
$ escape_slash=/fileserver/sn-cvsroot/batchpltf/
$ escape_slash_b=`echo $escape_slash | sed s,/,\\\\\\\\\\/,g`
$ echo $escape_slash_b
\/fileserver\/sn-cvsroot\/batchpltf\/