How to replace many intermediate blank lines with only 1 blank line each..

Question

Hi all,

I am learning to use sed and have a hypothetical requirement:

bash$ cat states
California


Massachusetts


Arizona

Wisconsin

I would like the output: (Basically I want only one blank line in between)

California

Massachusetts

Arizona

Wisconsin

Please suggest ways to do it with sed. Should it use concept of multiline pattern space etc.

Thanks,

Jagrati

---

In response to one of the comments, Ok, I ran some test scripts, and I was wrong with hypopthesis that address space selection criteria is applied only once. It seems to get activated again and again.

bash$ cat file 
a 
b 
c 
a 
b 
c 

bash$ sed -e '/a/,/b/s/$/x/g' file 
ax 
bx 
c 
ax 
bx 
c 

I had the impression that it would yield:

ax 
bx 
c 
a 
b 
c

So could someone please explain how the sed proceeds in case of comma separated address spaces.

Answer 1

Try this it will delete all consecutive blank lines

cat states | sed '/./,/^$/!d'

outputs:

California

Massachusetts

Arizona

Wisconsin

Answer 2

you can use awk as well

$ cat file
California


Massachusetts


Arizona

Wisconsin

$ awk 'NF{print;print ""}'  file
California

Massachusetts

Arizona

Wisconsin

Answer 3

I know the OP explicitly asked for a solution using sed and/or awk, but the problem could also be solved with just cat:

cat -s file

Otherwise, here is yet another way to solve the problem with sed:

sed '/./!d; G' file

which exploits G to append an empty line to all non-empty lines found in files.

Answer 4

This can be accomplished without sed, but also without meeting your requirement of learning sed:

cat -s states

It's interesting to note that the three examples for doing this given in info sed for GNU sed are each much more complex than dalton's and marco's answers.

Here are those versions condensed into one-liners:

sed ':x;/^\n*$/{N;bx};s/\n*/\n/' states

sed '1,/^./{/./!d};:x;/./!{N;s/^\n$//;tx}' states

sed -ne '/./!d;:x;p;n;/./bx;:z;n;/./!bz;i\ ' -e 'bx' states

Answer 5

grep . file > file_op