[SOLVED] Simple bash + sed s command to join six lines

Issue

Given the following file:

temp.txt

line1
line2
line3
line4
line5
line6
line7

And the following bash/sed commands:

for (( i=0; i<6; i++ ))
do
sed "N;1s/\\(.\*\\)\n\\(.\*\\)/\1 \2/" temp.txt
done

doesn't produce the desired stout:

line1 line2 line3 line4 line5 line6
line7

Thanks in advance for any insight you can provide.

Solution

Your code is:

for (( i=0; i<6; i++ ))
do
sed "N;1s/\(.*\)\n\(.*\)/\1 \2/" temp.txt
done

• N tells sed to read the next line. By doing so the current line number is incremented. So testing for line 1 can never succeed. (You should test for line 2.)
• You have not used -i and so the file will not be changed.
• It is rather inefficient to call sed 6 times instead of just doing everything in a single invocation. (And you should only call it 5 times to join 6 lines.)

A simpler alternative:

sed -i '1{N;N;N;N;N;s/\n/ /g;}' temp.txt

---

In the case that the input could have fewer than 6 lines and up to 6 lines should be joined, an alternative script is:

sed -i ':a;$q;N;1,6s/\n/ /;ta' temp.txt

Note that BSD-ish sed requires a literal newline after any label.

Behaviour of N when there is no next line varies between implementations. Some output the existing contents of the pattern space, others follow posix and discard it. (GNU sed will do either: seq 3 | sed N vs seq 3 | sed --posix N.) Prefixing $q sidesteps this ambiguity and forces output.

Answered By - jhnc

Answer Checked By - David Goodson (WPSolving Volunteer)