Options
# -n (default printing of pattern buffer)
sed 'p' # print twice
sed -n 'p' # cat
# -e (chain commands)
sed -e '1d' -e '3d'
# -f (script file)
echo -e '1d\n3d' > cmd.txt
sed -f cmd.txt file.txt
|
Useful recipes
# Unix commands
sed '' file # cat file
sed '10q' # head
sed -n '$p' # tail
sed 'w file.bak' file # cp file file.bak
sed -n '$=' file # wc -l file
sed 'y/a/A/' # tr "a" "A"
# useful
sed -n '/^$/p' # show non-empty lines
sed 's/^#.*//g' # remove comments
sed '3,6 s/^/#/' # add comments
|
substitue -[a1[,a2]]s/p1/p2/[flags]
# occurrences
sed 's/foo/bar/' # only first
sed 's/foo/bar/2' # second
sed 's/foo/bar/g # global flag, all occurrences
sed -n 's/foo/bar/p' # print only changed lines
sed -n 's/foo/bar/w file' # writes to file
sed -n 's/FOO/bar/pi' # case insensitive
# uppercase - lowercase
sed -n 's@foo@\ufoo@p' # one letter uppercase
sed -n 's@foo@\Ufoo@p' # uppercase
sed -n 's@foo@\LFOO@p' # lowercase
sed -n 's@foo@\Ufoo\Ebar@p' # back to normal
# catching char
# with & (no regex, only one char)
sed 's/[[:digit:]]/Number &/'
# with catching group \( \)
sed 's@\(\w\+\) \(\w\+\)@\2 \1@' # re-order
|
regex - need escape
\?
\+
# catching group
\( \)
# occurrence
\{ \}
# logical or
\|
|
^, $ and \` (beginning of pattern space)
|
|
print - [address1[,address2]]p
# line number
sed -n '3p'
sed -n '3,$p' # range ($ is end)
# pattern
sed -n '/foo/p'
sed -n '/foo/,/bar/p' # between matching lines
# + and ~
sed -n '3,+2p' # 3 and next 2 lines
sed -n '1~2p' # every two lines
|
same for delete - [address1[,address2]]d and write - [address1[,address2]]w file
insert - [address]i line
sed '4i text' # insert line above 4th
sed '/foo/i text'
|
same for append - [address]a line (below)
change -[address1[,address2]]c line
sed '3,6c line' # replace by single line (note)
|
consider range as single block
translate -[addr1[,addr2]]y/list1/list2/
echo "a b c" | sed 'y/abc/123/' # outputs 1 2 3
|
replace char i of list1 by char i of list2
line numbering -[address1[,address2]]=
sed -n '/foo/='
sed -n '$=' # wc -l
|
line number above every match
read - [address]r file
sed '3r file' # content of file after 3rd line
|
hidden char -[addr1[,addr2]]l [wrap]
sed -n 'l' # displays \t and \n as $
sed -n 'l len' # wraps line (\) after len char
|
|
|
quit - [address]q [exit code]
sed '3q' # stops execution after 3rd line
|
execute -[address1[,address2]]e cmd
sed '3e ls' # execute ls before 3rd line
|
loops
# b label
sed -n ':label;s/^/-/;/---/!b label;p'
# t label
sed -n ':label;s/^/-/;/---/!t label;p'
|
N and P
# N adds two lines to buffer, separated by '\n'
sed -n 'N;s/\n/,/p' # concatenate lines 2 by 2
# P print first part of N (before '\n')
sed -n 'N;P' # sed -n '1~2p'
|
n and x - load / exchange buffers
sed 'n' # cat
# exchange, load next, print
sed 'x;n;p' # sed -n '2~2p'
# exchange, load next, exchange, print
sed 'x;n;x;p' # sed -n '1~2p'
|
h and H - pattern to hold
# [address1[,address2]]h
# copies pattern to hold - overwrite hold
# exchange, print previous
sed -n '/foo/!h;/foo/{x;p}'
# [address1[,address2]]H
# append pattern to hold
# append current, exchange, print
sed -n '/foo/!h;/foo/{H;x;p}' # order (n-1, n)
|
g and G - hold to pattern
# [address1[,address2]]g
# copies hold to pattern - overwrite pattern
# print, copy previous, print
sed -n '/foo/!h;/foo/{p;g;p}' # order (n, n-1)
# [address1[,address2]]G
# append hold to pattern
# print, copy previous, print
sed -n '/foo/!h;/foo/{G;p}' # same as above
|
|
