Blog

sed is a complex utility to master. This command streams the editor for filtering and transforming text. Entire books have been written dedicated to using sed proficiently. So please keep in mind, this tutorial has the purpose of introducing three basic-common-uses of sed −

• character substitution
• printing operations
• delete operations

The common command syntax for sed is −

sed [options] [file to operate on]

Following are the common sed switches to remember.

-i will apply changes to a file, instead of editing the file’s stream as passed to sed.

sed, when used with the -e option extends the command to process multiple operations on a stream. This can be done instead of piping used recursively.

echo "Windows and IIS run the Internet" | sed -e 's/Windows/Linux/' -e 's/ and           

IIS//'  -e 's/run/runs/' 

Linux runs the Internet

The -n option with sed suppresses default printing to stdout. Using sed’s print command, as we see, each line will be duplicated to stdout.

bash-3.2# sed 'p' ./lines.txt  

line1 

line1 

line2 

line2

This time, we use the -n option with sed −

bash-3.2# sed -n 'p' ./lines.txt  

line1 

line2

sed will send content streams to stdout. When the ‘p‘ or print command is added, a separate stream is sent for each line causing all lines to be duplicated in stdout.

sed substitution command

This command is specified with the ‘s‘ option. We have seen sed used with its substitution command a few times already. Following is a simple example −

[root@centosLocal Documents]# echo "developers eat veggies and fruit" | sed -e

's/veggies/pizza/' -e 's/fruit/coffee/' 

developers eat pizza and coffee 

[root@centosLocal Documents]#

Let’s try this one, on a file named dev.txt −

[root@centosLocal centos]# cat dev.txt  

Developers code all night and sleep all day.

[root@centosLocal centos]#

Now, let’s change the contents of a file instead of just its output stream sent to sed −

[root@centosLocal centos]# sed -ibak 's/sleep/code/' ./dev.txt  

[root@centosLocal centos]# ls dev* 

dev.txt  dev.txtbak 

[root@centosLocal centos]# cat dev* 

Developers code all night and code all day. 

Developers code all night and sleep all day. 

[root@centosLocal centos]#

Note − We used -i option with a unique suffix to create a backup file.

sed print command

This command is specified with the ‘p‘ command.

Let’s use our names.txt file, the output has been edited for brevity sake −

[root@centosLocal Documents]# sed -n "p" ./names.txt  

Ted:Daniel:101 

Jenny:Colon:608 

Dana:Maxwell:602 

Marian:Little:903 

Bobbie:Chapman:403 

Nicolas:Singleton:203 

Dale:Barton:901 

Aaron:Dennis:305

sed allows the use of “addresses”, to more granularly define what is being printed to stdout −

[root@centosLocal Documents]# sed -n "1,10p" ./names.txt

Ted:Daniel:101 

Jenny:Colon:608 

Dana:Maxwell:602 

Marian:Little:903 

Bobbie:Chapman:403 

Nicolas:Singleton:203 

Dale:Barton:901 

Aaron:Dennis:305 

Santos:Andrews:504 

Jacqueline:Neal:102 

[root@centosLocal Documents]#

Just like head, we printed the first 10 lines of our names.txt file.

What if we only wanted to print out those with an office on the 9th floor?

[root@centosLocal Documents]# sed -n "/90/p" ./names.txt 

Marian:Little:903 

Dale:Barton:901 

Kellie:Curtis:903: 

Gina:Carr:902 

Antonia:Lucas:901 

[root@centosLocal Documents]#

Pretty easy. We can also print out everyone, except those with offices on the 9th floor −

[root@centosLocal Documents]# sed -n '/90/ !p' ./names.txt 

Ted:Daniel:101 

Jenny:Colon:608 

Dana:Maxwell:602 

Bobbie:Chapman:403 

Nicolas:Singleton:203 

Aaron:Dennis:305 

Santos:Andrews:504 

Jacqueline:Neal:102 

Billy:Crawford:301 

Rosa:Summers:405 

Matt:Davis:305 

Francisco:Gilbert:101

Sidney:Mac:100 

Heidi:Simmons:204 

Matt:Davis:205 

Cristina:Torres:206 

Sonya:Weaver:403 

Donald:Evans:403

In the above code, we negated ‘p‘ printing operation between / and /with !. This performs similarly to the “d” or deletes command. However, the results can vary with negation in sed. So as a general rule: p to print and negate what you do not want.

sed delete command

As mentioned, the delete command is the opposite of the sed print command. Let’s start with our name.txt file −

[root@centosLocal Documents]# sed 'd' ./names.txt  

[root@centosLocal Documents]#

Nothing printed out. With the above command, we asked sed to delete every line from stdout in the stream. Now, let’s only print the first two lines and “delete” the rest of the stream −

[root@centosLocal Documents]# sed '1,2 !d' ./names.txt  

Ted:Daniel:101 

Jenny:Colon:608

[root@centosLocal Documents]#

See? Similar to the ‘p‘, or print command. Now let’s do something useful with the delete command. Say we want to remove all blank lines in a file −

[root@centosLocal Documents]# cat lines.txt  

line1 

line2 

line3 

line4 

line5 

[root@centosLocal Documents]#

It is not uncommon to receive a file like this containing jumbled text, copying and pasting an email, or formatted with non-standard line breaks. Instead of interactively editing the file in vim, we can use sed to do the work for us.

[root@centosLocal Documents]# sed -i '/^\s*$/ d' ./lines.txt

[root@centosLocal Documents]# cat ./lines.txt  

line1 

line2 

line3 

line4 

line5 

[root@centosLocal Documents]#

The file is now formatted in an easily readable fashion.

Note − When making changes to the important files, use -i switch. Appending a file backup suffix is greatly advised to reserve the file contents (sed can make some extreme changes with the slightest typo).