<-- Previous week

sed

2021-12-09

Steven Whitaker

sed

This is a scriptable text editor. You can use sed to run ed commands on the input stream.

There are three points to sed:

• sed reads line-by-line.

• sed has a pattern buffer.

• sed has a hold buffer.

How sed works:

1. "Reads" the first line; it stores the first line into the pattern buffer.

2. Executes any commands that should be run on this line.

3. "Reads" the next line.

If you use the command h, this will push the pattern buffer into the hold buffer.

g does the opposite: it pushes the hold buffer into the pattern buffer.

Usage

sed [options] [script] [file]

options found with sed -h or sed --help. There are not that many and you can read them all.

script is where all the magic or actual benefit to this program actually comes from.

file is any input stream. Can be a file name, or you can pipe data into it.

Examples

Commands

a will append to a specific line, so: seq 10 | sed '5ahello' will append hello to the 5th line.

seq 10 | sed -n '!5p'

Let's step through this command:

seq 10 | sed '3{h;z;g}' and what will happen when I execute this command? Absolutely nothing!

• 3 looks at the 3rd line.

• h holds (stores) the data from the 3rd line. (the { and } mean all commands use 3).

• z zaps (deletes) the data from the 3rd line. (the { and } mean all commands use 3).

• g replaces the 3rd line with what is in h; which was the 3rd line initially.

h is a single register and g writes this single register.

Let's say we want to swap lines 5 and 6. Let's build up this command.

seq 10 | sed '5{h;z;n;G}'

We hold the 5th line in the buffer, zap the 5th line, then go to the next line and print the hold buffer with G.

We've created a gap on the 5th line, though. We need to use some special flags to get this proper.

seq 10 | sed -n '1,4p;5{h;z;n;G;p};7,$p'

The -n flag removes automatic printing of the stream. Now, we need to print manually ourselves.

• 1,4p just prints the beginning by ourselves.

• 5{h;z;n;G;p} is the same as before, but now we are printing manually.

• 7,$p prints the rest of the file. We use 7 because G prints a new line and so the 6th line is new.

At this point, you should start writing a file to save this for later:

seq 10 | sed -n -f swap.sed

Let's reverse a list, which is actually much easier:

seq 10 | sed -n 'G;h;$p'

Let's step through what happens in sed here.

sed goes to the first line, and sees that G and h will be executed.

• G will append whatever is on the hold buffer into the pattern buffer. There is nothing in the hold buffer right now, so it prints just a new line.

• h will store the pattern space

Regex substitutions

Say you only want lines that contain a word:

sed -n '/text/p' file -> Does not automatically print anything and uses regex to print only text. This is grep.

echo sup | sed 's/p/d/' -> sed will (s)ubstitute the letter p into d, so it will output sud.

cat data.txt | sed 's/p/d/' -> Same as the previous example, but using a different function to print the data.

sed 's/p/d/' data.txt -> the data.txt file is used as the [file] input into the sed function.