UNIX for Dummies Questions & Answers

The file is repeating blocks of text separated by a blank line. Some of the lines are unique.

e.g.:

This is unique line
This is line 2
This is line 3
This is line 4

This is unique line
This is line 2
This is line 3
This is line 4

...


The goal is to change/replace line 4 only on the last block, in a single pass.

Available tools are /bin/sh and some of the POSIX standard utilities. But no awk. No Perl, etc.

It's possible in sed. But how to do it in one pass?

Here's a two-pass.

Code:

sed '$d' < file |sed '$d'
printf "This is the new line 4\n"

And here's a script that changes only line 4 in the _first_ block.

Code:

# sed -nf script < file

/This is line 4/{
s//This is the new line 4/
:loop
p
n
bloop
}
p

Cheers.

Here's a single pass sed that should work. I've put each command on a separate line to make it more clear.

Code:

sed '
$ b last
/^$/ b blank
H
d
b
:blank
x
p
d
x
b
:last
x
s/This is line 4/New Line Four/
p
x
' tdata

I tested this with GNU sed version 4.1.5.

I've not added an explanation of the logic on the off chance that you'd like to figure it out; I'd be happy to if you'd like.

There is a big assumption that you know the line 4 text to put into a substitute command. If it's not known, this won't work, but from your description I believe you do know it.

The one liner is below:

Code:

' $ b last; /^$/ b blank; H; d; b; :blank x; p; d; x; b; : last x; s/This is line 4/new line 4/; p; x' tdata

Found via the infamous sed one-liners file.

Code:

# delete last 2 lines of file

sed 'N;$!P;$!D;$d' < file
printf "This is the new line 4\n"
printf "\n"

---------- Post updated at 08:36 PM ---------- Previous update was at 08:24 PM ----------

Thanks agama.

Indeed line 4 is known.

This looks clear and I will study it as I really want to start using loops and the full range of sed commands. I use sed constantly but I'm just not there yet.