Yrurtj

I can't find this case in the board, so I'm asking the question.

This is input file:

module  

x(a,b,c)  

module  

y(d,e,f,  

g,h,i)  

module  

z(j,k,l)

And output file should be:

module x(a,b,c)  

module y(d,e,f,  

g,h,i)  

module z(j,k,l)

What you want to do is to join the module lines with the next line.

Using sed:

$ sed '/^module/N;s/n//' file

module  x(a,b,c)

module  y(d,e,f,

g,h,i)

module  z(j,k,l)

This is with your data copied and pasted as is, with spaces at the end of each line.

The sed command will print each line as it is read, but when it encounters a line that starts with the string module, it appends the next line with an embedded newline character in-between (this is what N does). We remove that newline character with a substitution before the result is printed.

If your data has no spaces at the end of the lines, use

$ sed '/^module/N;s/n/ /' file

module x(a,b,c)

module y(d,e,f,

g,h,i)

module z(j,k,l)

Just in case you'd want this (assuming no spaces at end of input lines):

$ sed -e '/^module/bpp' -e 'H;$bpp' -e 'd' 

      -e ':pp' -e 'x;/^$/d;s/n/ /g' file

module x(a,b,c)

module y(d,e,f, g,h,i)

module z(j,k,l)

Annotated sed script:

/^module/ b print_previous; # print previous record

H;                          # append this line to hold space

$         b print_previous; # print previous (last) record

d;                          # end processing this line



:print_previous;            # prints a record accumulated in the hold space

x;                          # swap in the hold space

/^$/ d;                     # if line is empty, delete it

s/n/ /g;                   # replace embedded newlines by spaces

                            # (implicit print)

Using awk:

~ awk '/^module/ {l = $0; getline; printf "%s", l} 1' input-file

module  x(a,b,c)

module  y(d,e,f,

g,h,i)

module  z(j,k,l)

For each line that starts with module, save the line in l, move to the next line (getline), and print the saved line without a newline. Then print every line.

Another option: create an ed script!

This starts by pre-counting the number of joins that are required; it then generates that number of ed search & join commands and pipes them, along with a save & quit at the end, into ed:

#!/bin/bash

n=$(grep -c '^module *$' input)

{

  for((i=1; i <= n; i++))

  do

    printf '/^module *$/n.,+1jn'

  done

  echo w

  echo q

} | ed -s input >/dev/null