CODEWITHSUNDEEP
sedgrep
jplusc
jplusc

Reputation: 143

Merge lines between two patterns using sed

I have an output file that looks like this:

HEADER 1:
server1 server2 server3
server4 server5 server6
server7 server8 server9
HEADER 2:

HEADER 1:
server10 server11 server12
server13 server14 server15
server16 server17 server18
HEADER 2:

I need to merge everything between the two headers so the updated output file looks like this:

HEADER 1:
server1 server2 server3 server4 server5 server6 server7 server8 server9
HEADER 2:

HEADER 1:
server10 server11 server12 server13 server14 server15 server16 server17 server18
HEADER 2:

The header names always remain constant.

I'm using UnxUtils in Windows, so unfortunately I don't have awk to do the manipulation. How do I go about it using sed?

Upvotes: 10

Views: 11092

Answers (2)

perreal
perreal

Reputation: 97928

Using sed:

sed '/HEADER 1/{n;:l N;/HEADER 2/b; s/\n//; bl}' input
  • n skips/prints the current line (HEADER 1), clears buffer
  • l is a label for looping (goto label)
  • N adds (appends) lines to buffer (preserving newlines)
  • /HEADER 2/b, where b is branch (without the label, it is break), break out when HEADER2 is seen
  • s/\n// removes the newlines in the buffer
  • bl jumps back to label l

Upvotes: 22

Jotne
Jotne

Reputation: 41446

This shows how it could be done using awk
You can download Gawk for Windows

awk '/HEADER/ {printf "\n%s\n",$0;next} {printf "%s ",$0}' file

HEADER 1:
server1 server2 server3 server4 server5 server6 server7 server8 server9
HEADER 2:

HEADER 1:
server10 server11 server12 server13 server14 server15 server16 server17 server18
HEADER 2:

Upvotes: 5

Related Questions