As with sort, sed takes pathname operands. Using:

Code:

cat file | sed [options] [OtherOperands]

instead of:

Code:

sed [options] [OtherOperands] file

wastes system resources, slows down your system for other users who may be sharing your system, and makes your command line run slower.

If you have an input file named list containing:

Code:

nothing from this line
KR1201
abcKR3012
KR1111def
defKR3333extra
12345KR678910

Previous line was empty; neither of these should appear in output

and you run the command:

Code:

sed -n 's/\(.*\)\(KR[0-9][0-9][0-9][0-9]\).*/\2/p' list

using a standards conforming version of sed, you get the output:

Code:

KR1201
KR3012
KR1111
KR3333
KR6789

The Linux sed man page says:
and in the description of the sed s(ubstitute) command says: But, there is no other mention of sub-expressions on that man page.

Note that this sed command makes no effort to find more than one occurrence of your pattern if it appears more than once on a single line