Useful one-line scripts for sed (2005)

http://www.pement.org/sed/sed1line.txt

by kawera • 9 years ago

52 7 9 years ago

pement.org

baldfat• 9 years ago

I don't know what wins for most under used tool, sed or awk?

Both of these do wonders for clearing up messy data and for a wealth of other things.

gwaefawfewaef• 9 years ago

This guy is very interesting. pement.org

http://pement.org/perl/biblink126_pl.txt

bpchaps• 9 years ago

Definitely. I heard about him (10 years ago) from one of his old bosses who mentioned this page. This sed one liner page took about a month to go through from basically zero Unix experience. Probably the most important learning piece I've ever gone through. Good stuff.

zurn• 9 years ago

Widely useful pick of the bunch:

  # print section of file between two regular expressions (inclusive)
  sed -n '/Iowa/,/Montana/p'             # case sensitive

predakanga• 9 years ago

I've found that pattern to be very useful when altering/sanitising MySQL backups.

  # Set the engine of all tables in schema `foo` to TokuDB
  sed -re '/^USE `foo`/,/^USE / s/^\) ENGINE=\S+/) ENGINE=TokuDB/'

majewsky• 9 years ago

I use stuff like that all the time with config files where sections span multiple lines. For example, if ~/.ssh/config has a section like

  Host example
    HostName example.com
    IdentityFile ~/.ssh/id_example
    Port 1234

  Host next
    ...

Then the following selects the identity file:

  sed -n '/^Host example/,/^Host\>/{/^\s*IdentityFile/{s/^\s*IdentityFile\s*//;p}}' ~/.ssh/config

Although I would probably use awk after sed because it handled space-separated fields nicer:

  sed -n '/^Host example/,/^Host\>/p' ~/.ssh/config | awk '$1=="IdentityFile"{print$2}'

801699• 9 years ago

Here is something to use as a replacement for "grep -q".

In the past I found grep -q was not portable.

There are times when I do not have grep.

But I always have sed.

   cat grepq

   test $# -ge 1||
   exec echo usage: $0 PATTERN \[FILE\]

   # count lines until PATTERN 
   __=$(exec sed -n '/'"$1"'/!d;=;q' $2);
   # no of lines 
   exec test ${#__} -gt 0;

For example,

   grepq '93.184.216.34 example.com' /etc/hosts||
   echo 93.184.216.34 example.com >> /etc/hosts

One of the very early OReilly books has a chapter or two about sed that tries to describe it using an analogy to a scrivener in a monastary. Quite amusing. This is not the "sed and awk" book. It was an earlier book on text editing with UNIX. It may have been co-authored by OReilly himself; I cannot remember.

Useful one-line scripts for sed (2005)

Related Stories