Hi team,


sed -ne 's/regex/{h;p}' <file.txt>

The above prints the whole file(and not just the searched regex).How does the 'h' and 'p' gets resolved in this case?

You are missing the RHS with the replacement expression. If you want to hold & print on a match, use:
/regex/{h;p}. I assume you need to do more as well, since just to print, you don't need to hold the line as well.

Ah, thats right(I actually am using ,/regex/{h;p}. My text file is below:

Linux
AIX
Windows
Mac

So when I use sed -ne '/Linux/{h;p}', it prints the whole file and not just the "Linux".why?

No, it doesn't It works as you expected.

Markus

The 'h' command copies the matching lines into the hold buffer, then the 'p' command prints the line. But you never do anything with the hold buffer after you copy to it, so it's as if the 'h' command doesn't exist.

You have to include another expression that moves the contents of the hold buffer back into the pattern buffer for it to have any use.

If you haven't seen it yet, the grymoire tutorial has a good lineup on multi-line use:
http://www.grymoire.com/Unix/Sed.html#uh-47

(Frankly, I've never been able to fully understand the use of the hold buffer though. I get the concept, but getting the flow control to come out correctly always seems to hang me up.)

1 members found this post helpful.