You must be heard the word, on-the-fly, when read some sed guide books. It’s really a very powerful editor tool (we all know). So in this post, I wanna write down something of my journey in learning it.

The Key of Sed Editor

Please remember that the sed editor is a myriad of commands and formats. And here I described some of the basic commands and features, that you can make life easier. Oh! Another thing you should keep in mind is that sed won’t modify the original file, but output to the STDOUT.

Substitution Command

First, let’s see how to replace matching patterns in the test string.

$ cat a
this line is line 0
this line is line 1
this line is line 2
this line is line 3
$
$
$ sed 's/line/row/' a
this row is line 0
this row is line 1
this row is line 2
this row is line 3

What result you got? By default, sed only replaces the first occurrence in each line. To make it work on different occurrences of the test, you must use a substitution flag. The syntax like below:

s/pattern/replacement/flags

There are four options of subtitution flags available (cite from a book :>):

A number, indicating the pattern occurrence for which new text should be substituted.
g - Indicates that new text should be substituted for all occurrences of existing text.
p - Indicated that the contents of the original line should be printed.
w file - Write the results of the substitution to a file.

Ok! See some example and you will see it.

$ sed 's/line/row/2' a
this line is row 0
this line is row 1
this line is row 2
this line is row 3

The above we specify a 2 as the substitution flag. Next is for flag g:

$ sed 's/line/row/g' a
this row is row 0
this row is row 1
this row is row 2
this row is row 3

Ha, the left two flags I won’t give example, but you can test them if you like.

Mention: Substituting pathnames (include forward slash) in a file can get awkward. When you deal with forward slash, you’d have to use backslash to escape it. Like this:

$ sed 's/\/bin\/sh/\/bin\/csh/' /etc/passwd

Using Address

By default, sed works on all lines of the text data. If you only want to apply to a specify line, or a range of lines, you must ust line address, like [address]command.

There are two form of line addressing in the sed editor:

A numeric range of lines
A text pattern that filters out a line

Examples will tell anything.

$ sed -n '2s/line/row/gp' a
this row is row 1
$
$ sed -n '2,$s/line/row/gp' a
this row is row 1
this row is row 2
this row is row 3

Here I use -n option with flag p to suppresses output any line that has been modified. And the line number starts from 1, I worry you stay in C/C++. The ‘$’ indicates the end line of file, similar to that in vim editor.

$ sed -n '/1/s/line/row/p' a
this row is line 1

The above is for specifying a text pattern that it uses to filter lines for the command.

Also, if you need use more than one command on an individual line, you can do like this:

$ sed '[address]{
	command 1 
	command 2
	...
	command n}' file

If you’re interested, try it. Here I won’t give any examples.

Deleting Lines

The delete command is obviously most useful when used in conjunction with a specified address. But be careful, if you forget to include an addressing scheme, all of the lines will be deleted from the stream(don’t worry, no original file)

$ sed 'd' a     	--- nothing will output
$ sed '1,3d' a
this line is line 	--- only line 4 left
$
$ sed '/3/d; /2/d' 	--- the pattern-matching feature also applies the d command
this line is line 0
this line is line 1
$

please think of this syntax: sed '/2/,/5/d' a. What will happen?

Inserting and Appending Text

The insert command (i) adds a new line before the specified line.
The append command (a) adds a new line after the specified line.

The format for doing this is:

sed '[address]command\new line' file	---be careful it is the backslash '\'

Here we see two simple examples.

$ sed '1i\this line is new line' a	---if you don't specify a address number, another case.
this line is new line
this line is line 0
...(ignore)
$
$ sed '1a\this line is new line' a
this line is line 0
this line is new line
...(ignore)
$

You may try to specify a range of address, and see what’s the result!

Changing Lines

The change command © allows you to change the content of an entire line of text in the data stream.

$ sed '2c\this is changed line 1' a	---Careful! backslash
this line is line 0
this is changed line 3
this line is line 2
this line is line 3
$
$ sed '/line 1/c\this is changed line 1' a
(same output as above)

Mention: If you specified a range of address, it will use specified line to replace all line in the range.

The Transform Command

The transform command (y) is the only sed editor command that operates on a single character. The format is:

[address]y/inchars/outchars/

This command performs a one-to-one mapping of the inchars and the outchars values. A simple example:

$ sed -n '1y/ /\t/' a
this	line	is	line	0
this line is line 1
...(ignore)

The command is a global command; that is, it performs the transformation on any character found in the text line automatically, without regard to the occurrence.

Printing Revisited

There are three commands that also can be used to print information from the data stream:

The lowercase p command to print a text line
The equel sign (=) command to print line numbers
The l (lowercase L) command to list a line

As before, a example:

$ sed -n '/line 1/{
p
s/line/row/p
}' a
this line is line 1
this row is line 1
$
$ sed -n '/line 1/{=; p}' a
2
this line is line 1

The l command may be useful to transform text type.

Using File with sed

The w command is used to write lines to a file. The format is:

[address]w filename

An simple example:

$ sed '1,2w temp' a &
$ cat temp
this line is line 0
this line is line 1

The r command allows you to insert data contained in a separate file. The format is:

[address]r filename

Also, an simple example:

$ echo "from another file" > b
$ sed '/line 0/r b' a
this line is line 0
from another file
...(ignore)

Summary

Nothing to summarize. Keep hungry, keep foolish.

In Shanghai Vortex