myfilecontains the following lines:
pink blue red Blue orange blue purple blue blue
If I want to replace all occurrences of the word "blue" with "green", I could issue the following sed command at a Bash shell prompt.
$ sed -i -e 's/blue/green/g' myfile
-i option indicates the file should be edited in
place. I.e., the changes are made within the file itself rather than a new
output file being created.
-e indicates that what follows will be the
expression/command for sed to run.
Within the single quotes is the expression for sed to evaluate. The
s tells sed I want to perform a subsitution operation.
The parts of the substitution are enlosed within the
forward slashes (
/). The string that follows the first
blue, is the string for which sed should search.
The string that follows the second slash is the replacement string, i.e.
green, which is terminated by another slash. I don't
necessarily have to use the slash character as the separator. E.g., I could
use a colon (
:), instead, i.e.,
sed -i -e 's:blue:green:g'
myfile and achieve the same effect.
g after the last slash tells sed that I want to replace all
instances of the first string with the second string, not just the first
instance it finds. If I used
sed -i -e 's/blue/green/' myfile
g after the last slash, sed would only replace the
first instance of "blue" that it found with "green".
After I execute the command
sed -i -e 's/blue/green/g' myfile
the file would then contain the lines below:
pink green red Blue orange green purple green green
All occurrences of the word "blue" were replaced with "green". "Blue" was not replaced because the search is case sensitive, so a capital "B" is not the same as a lower-case "b". The tool supports regular expressions, so, if I wanted instances of "Blue" replaced with "green", I could use the following command, instead.
sed -i -e 's/[bB]lue/green/g' myfile
If I enclose letters within brackets, that is a regular expression indicating to sed that I want it to look for any instances of either a "b" or a "B" followed by "lue" with "green". If I ran that command on the original file, I would then have the following lines in it.
pink green red green orange green purple green green
I could enclose more than just two characters within the brackets. E.g., if I also wanted sed to replace any instances of "clue" or "Clue" with "green" as well, I could use the command below.
$ sed -i -e 's/[bBcC]lue/green/g' myfile
Alternatively, if I wanted sed to ignore the case of any characters in
the search string, e.g., if I wanted it to replace "blue", "Blue", "BLUE",
etc., I could just put a
I after the last separator. E.g.,
I could use
sed -i -e 's/blue/green/Ig' myfile. The search
would then be case-insensitive and global replacing any variations in
capitalization of the word "blue" with "green" even if the word occurred
multiple times on the line. If I only wanted the first occurrence of the
word to be replaced, I would omit the "g" from
If I wanted to create a Bash
script that any user could run that would
prompt the user for text to be replaced in a file, the replacement text, and
file name, I could put the following lines in the script, which I will name
#!/usr/bin/bash read -p "File name: " filename read -p "Text to replace: " replace read -p "Replacement text: " replacement sed -i -e 's/$replace/$replacement/g $filename
read command is explained in section
Using the read built-in command" of the Bash Guide for Beginners.
To make the script executable, I could change the
file permissions to give
the owner of the file, any accounts in the same group as the owner, and all
accounts on the system as well execute permission with the command
$ ls -l replace_text -rw-rw-r--. 1 ann ann 171 Jan 8 12:49 replace_text $ chmod 775 replace_text $ ls -l replace_text -rwxrwxr-x. 1 ann ann 171 Jan 8 12:49 replace_text $
Then when the script is executed, it will prompt the user running it for the search and replacement strings and the file on which the command should operate.
$ ./replace_text File name: myfile Text to replace: blue Replacement text: green $
Rather than having sed take its input from a file, you can also pipe input to it from another command. E.g.:
$ echo "My favorite color is blue." | sed -e 's/blue/green/' My favorite color is green. $
That isn't particularly useful in the simple example above, but that can be useful in Bash scripts. E.g., I have some HTML files that I need to alter to pass the W3C Markup Validation check for compliance with HTML standards, so I include sed commands to do so in the following script.
#!/usr/bin/bash if [ -z "$1" ]; then read -p "File name: " filename else filename=$1 fi if [ -z "$2" ]; then read -p "Link Name: " linkname else linkname=$2 fi sed -i -e 's/^/<p style="margin: 0px; text-align: center;">\n/' $filename sed -i -e 's/&/&/g' $filename sed -i -e 's/border=0/style="border:0;"/g' $filename sed -i -e 's/subid=0" >/subid=0" alt="1x1 px">\n<\/p>/' $filename sed -i -e "s/<IMG /\n<IMG alt=\"$linkname\" /" $filename sed -i -e "s/<\/a>/<br>\n$linkname<\/a>\n/" $filename newfilename=$(echo "$filename" | sed -e 's/480x270/240x135/') cp $filename $newfilename sed -i -e 's/<IMG /<IMG width="240" height="135" /' $newfilename
The script, named
updatefile can accept the parameters it
uses, a file name (filename) and a link name (linkname), as arguments given on
the command line, e.g.,
updatefile myfile "Linux Tips". The
script checks to determine if the parameters are given on the command line and,
if not, prompts for them - the first parameter,
be the file name and the second parameter,
$2, should be the
descriptive text for the link, which will be assigned to the
In the first sed line,
sed -i -e 's/^/<p style="margin: 0px; text-align: center;">\n/'
$filename, the string I want to replace is represented by the
^) symbol. In regular expressions, a
the beginning of the line while a dollar sign (
$) represents the
end of the line. So you can use a caret as the search string for
sed and then put in text you want to insert at the beginning of a line.
\n signifies that a
should be inserted after that text, which is like hitting the Enter
key when you are editing a file to start a new line. So
style="margin: 0px; text-align: center;"> will be inserted as the
first line in the file before any existing text in the file.
The next three sed lines perform some text substitution in the
file. The first replaces any occurences of an ampersand character with the
HTML code for an ampersand, i.e.,
&. The second sed
command makes the code HTML5
compliant by replacing instances of
style="border:0;" because the former code, which was acceptable
with HTML 4, was deprecated in HTML5. The next line inserts
an "alt" tag for one image in the file that is only 1 pixel wide by 1 pixel
high, since "img" tags should have a text description that a browser can show
instead of the image, if the image can't be displayed for any reason. E.g.,
there are still text browsers, such as
Lynx. The link for that particular image has
the URL followed by a double quote, so that is what I have
sed search on. I place the closing
</p> paragraph tag
after the alt tag, since that is the last element in the file.
The next sed line inserts the linkname variable between the double
alt="". Either single quotes or double quotes can be
used to enclose the expression that the sed command will evaluate and process.
In the first 3 sed lines, I used single quotes. But in the fourth one that
inserts the link name text between the double quotes, I need to use double
quotes. If single quotes are used, sed will assume that
is to be inserted just as it appears, i.e., it won't evaluate the variable
and place its value between the double quotes. But, if I enlose the
expression that sed will evaluate, i.e., the part that appears after the
-e, I need to use double quotes for that expression rather than
single quotes to get sed to put the text for
I provided on the command line or when prompted for
by the script. But when I surround the expression in double quotes, but
have double quotes within the expression itself, I have another problem.
I need to have sed ignore the meaning it might otherwise give to those
double quotes, i.e., the first one it sees would appear to it to match
the prior double quote before the expression. I can do so by placing an
escape character before any double quote that I want it to regard as
just another character like "a", "b", "c", etc. The escape character is
\) character, which indicates that the character
immediately following the backslash is to be treated normally rather than to be
viewed as a special character. So I use
a backslash before each double quote. I also needed a backslash character before
the slash character in the prior line for the closing paragraph tag, i.e.,
<\/p> for the same reason.
If you want sed to insert a newline character, i.e., to start another line at a
point in the text, you can place a
\n at the point where you
want the new line to start which is what I do on the following line:
sed -i -e "s/<\/a>/<br>\n$linkname<\/a>\n/" $filename
That line inserts the HTML <br> break tag in the text and then breaks the line at that point as well, by inserting a newline character, which makes the text easier for me to read.
The file names I'm working with all terminate with "480x270" which are
the dimensions of an image linked to by the file. I want to edit the original
file, but also create a new file that will display the image at 1/2 the
original dimensions, i.e., 240
pixels wide by
135 pixels high. So I pipe the value of the filename variable to the
sed command using the
echo command and, using
sed, replace the "480x270" at the end of the file name with
"240x135". I use command substituion to assign the output of those
commands to the variable
newfilename by placing
$() around the echo and sed commands.
newfilename=$(echo "$filename" | sed -e 's/480x270/240x135/')
I can then copy the 480x270 file to one with the same name except with the
"480x270" at the end of the filename replaced with "240x"135". I can then
edit the new file and add
width="240" height="135" to the
IMG tag for the image, so though the image is still
480 pixels wide by 240 pixels high, browsers will display the image as
one 240 pixels wide and 135 pixels high.
sed -i -e 's/<IMG /<IMG width="240" height="135" /' $newfilename
So then I can run the script from a shell prompt with a command similar to the one below and have sed perform all of the substitutions I need in the input file and then copy that file to another similarly named file, but with the "480x240" part of the file name replaced with "240x135" and then have sed make the appropriate alteration to the contents of the new file.
./updatefile Learn_GREP_SED_on_Linux_for_Beginners_480x270 "Learn GREP and SED
on Linux for Beginners"