René Nyffenegger's collection of things on the web
René Nyffenegger on Oracle - Most wanted - Feedback -
 

sed, a stream editor

sed sed-programm input-file

sed -f sed-programm-file input-file

sed -f sed-programm-file -e sed-program input-file
The -f flag can be used to specify a file containing the sed programm.

Command line options

-e

-n

Turns off the default behaviour which prints every line. If this switch is used, a line must be explicitly printed for output.

Sed programm

A sed programm consists of 0, 1 or more functions.

function list

The following form applies function to all lines: 
function [arguments] [\]
The following form applies function only to the lines that match addresses: 
addresses function [arguments] [\]
An example for an address is $ which matches the last line. Together with the = function, the lines in a file can be counted
The following form applies function only to the lines between the line that matches first_address and the line that matches last_address 
first_address,last_address function [arguments] [\]
Sed reads the input file line for line. The sed programm is applied to each line.
Here's an example that makes use of this feature.

Grouping commands

Commands can be grouped with curly braces. The closing } must be the only item on a line.
This makes it possible to execute multiple commands for a matching address.
See Printing the first few lines of a file. 
addr_1[,addr2] {
  ..
  ..
}

Functions

s

The substitution function. 
s/regexp/substitution/x
x is nothing or one of the following:
  • N
    N being a number: replaces the Nth occurance of regexp.
  • g
    Replaces all occurences of regexp. The default is to only replace the first.
  • p
    Writes pattern space only if substituion was successful.
  • w file-name
    Writes the pattern space to the file file-name if substituion was successful.

i

The insert function.
See Inserting a line before a pattern

a

The append function.
See Inserting a line after a pattern and Appending text at the end.

d

The delete function. Compare to D
>After a line has been deleted, the next line will be read from the input file and the script re-starts at the beginning of the script.
See Delete empty lines and delete lines with a pattern.

d_

The delete up to '\n' function. Deletes the beginning up to the first new line in the pattern space. Skips the rest of the script and goes to its beginning again. Compare to d.
After a line has been deleted, the next line will be read from the input file and the script re-starts at the beginning of the script.
See Delete empty lines and delete lines with a pattern.

c

The 'change the pattern space' function. Replaces the current pattern space with some text.
See Change a pattern.

n

The (read) next line function.
Compare with N.
The current value of the pattern space is output, the pattern space is filled with the next line of the input file, but the script is not re-started at the beginning.
See printing every other line.

N

The (add) next line function.
adds a '\n' plus the current line to the pattern space.
Compare with n.
See Printing the first few lines of a file.

p

The print command. Prints the pattern space's content. Compare to P.

P

The print command. Prints the pattern space's content up to the first '\n'. Compare to P.

h

Copies (overwriting) the hold area with the pattern space's content.

H

Appends the pattern space's content to the hold area.

g

Gets the hold area's content.

G

Appends the hold area's content to the pattern space.

x

exchanges the hold area's content with the pattern space's content.

t

Branch to a a label if the previous substition was successful.

b

Branch to a a label.

q

Quits sed script.

w

Write pattern space to file.
See writing to a file.

=

Prints the current line number.
See Count lines.

:

Name a label to which can be jumped with t or b

Examples

Inserting a line before a pattern
Inserting a line after a pattern
Delete empty lines.
Count lines.
printing every other line.
writing to a file.
Printing the first few lines of a file.

Writing to files

The following script creates three files (/tmp/f, /tmp/o and /tmp/f) and writes lines according to their first letter into those files: 
echo 'one eins un
two zwei deux
three drei trois
four four quattre
five fuenf cinque' | sed '
 /^f/w /tmp/f
 /^t/w /tmp/t
 /^o/w /tmp/o'

Swapping two lines with sed

Assume, in the following file I want to swap the third with the fifth line: 
one
two
five
four
three
six
seven
Then this could be done with the following sed script: 
sed -ne '
 3!{p;d;}
 h;n;:1
 4!{N;b1}
 G;h;n;p;g;p' numbers
Note, it's a 4, not a 5 in the sed script!

Links

Splitting $PATH so that each component is printed on a single line.