Delete multiline pattern containing another pattern - sed

Question

I would to like to delete all instances of a pattern which contain another specified pattern (conveniently on a new line):
MID:
Sample Input:
header
BEGIN:
1abc
7wurw
END:
BEGIN:
22xyz
MID:
34utov
END:
Desired Output:
header
BEGIN:
1abc
7wurw
END:
I'm looking for possible one liners. Any help would be appreciated.

Answer 1

Using GNU sed:
sed -e :a -e '/^BEGIN:/,/^END:/ { /END:/!{$!{N;ba};};/MID:/d;}' inputfile
For your input, it'd return:
header
BEGIN:
1abc
7wurw
END:

Answer 2

I would use the RS, ORS variable. here is one-liner:
awk -v RS="BEGIN:\n" -v ORS="" '/MID/{next}NR>1{printf RS}7' file
test with your file:
kent$ cat f
header
BEGIN:
1abc
7wurw
END:
BEGIN:
22xyz
MID:
34utov
END:
kent$ awk -v RS="BEGIN:\n" -v ORS="" '/MID/{next}NR>1{printf RS}7' f
header
BEGIN:
1abc
7wurw
END:
note, printf RS is not very nice, I used it because I know it is BEGIN: good practice would be printf "%s", RS

Answer 3

This worked for me on sample:
sed '/BEGIN:/,/END:/{/BEGIN/{h;d};H;/END:/!d;x;/MID:/d}' input.txt
I am pretty sure it can be simplified a lot.

Answer 4

BEGIN { in_block = 0; }
/BEGIN:/ { in_block = 1; lineno = 0; arr[lineno] = $0; must_write = 1; next; }
/END:/ { in_block = 0;
if (must_write == 1) {
for (i = 0; i <= lineno; ++i) print arr[i]; print;
}
next;
}
/MID:/ { must_write = 0; next; }
in_block == 1 && must_write == 1 { lineno++; arr[lineno] = $0; next; }
in_block == 0 { print }
This should work (worked with the supplied test case). Some awk-wizards will find a denser solution, probably. But you can use this kind of processing also for other tasks.

Answer 5

This might work for you (GNU sed):
sed '/^BEGIN:/{:a;$!{N;/END:/!ba};/MID:/d}' file

Answer 6

This gnu awk divides section by the END: so its not true to BEGIN:-END:, but works:
awk '!/MID:/{printf "%s%s\n",$0,RT}' RS="END:" t
header
BEGIN:
1abc
7wurw
END:

Answer 7

sed -n '/BEGIN:/,/END:/ {
H
/END:/ {
s/.*//
x
/\nMID:/ !p
}
}' YourFile
For OneLiner
sed -n '/BEGIN:/,/END:/{H;/END/{s/.*//;x;/\nMID:/ !p;};}' YourFile
would not work with the last ; (i have to keep a "\n" on my AIX) (should work on GNU sed)
#for AIX
sed -n '/BEGIN:/,/END:/{H;/END/{s/.*//;x;/\nMID:/ !p;}
}' YourFile