How to make sed omit not matching lines

bash-scriptingregexsed

In the following example, sed matches lines starting with an a or a c and prints the first character of that line (a or c):

$ echo "ag
bh
ci
dj
ek
fl" | sed 's/\(a\|c\)./\1/' # Matches lines starting with 'a' or 'c'.

output:
a
bh
c
dj
ek
fl

However, the lines that do not match the pattern are also printed out. How do I tell sed to omit the lines that doesn't match the pattern? I can obtain the desired effect by combining it with grep (as follows) but I would like to know if sed can achieve that "by itself".

$ echo "ag
bh
ci
dj
ek
fl" | grep '[ac]' | sed 's/\(a\|c\)./\1/'

output:
a
c

Best Answer

Use the no-print flag (-n) and explicitly print successful substitute commands (s///p):

 sed -n 's/\(a\|c\)./\1/p'
Related Question