sed multiline match combined with delete between two lines? - awk

Question

Can I use sed's multiline match logic to delete all the lines between between a range of lines?
I have been trying all sorts of combinations of expressions with no luck.
Here is an example file ('creatures.txt'):
START TAG
species: swallowtail butterfly flying
legs
wings
head
END TAG
START TAG
species: common lizard running
legs
tail
head
END TAG
START TAG
species: peacock butterfly resting
legs
wings
head
END TAG
START TAG
species: blackbird flying
legs
wings
head
END TAG
I want do perform the following:
Whenever i encounter the following multiline match: (i) 'START' contained somewhere in the first line and (ii) 'butterfly' contained somewhere in the 2nd line
then delete all the lines between the 'START TAG' and 'END TAG' lines.
So using the example file above, the resulting output would be:
START TAG
END TAG
START TAG
species: common lizard
legs
tail
head
END TAG
START TAG
END TAG
START TAG
species: blackbird
legs
wings
head
END TAG
Thanks,
James.

Answer 1

Like this:
sed '/START TAG/{N;/butterfly/{:a;/END TAG/!{N;ba};d}}' file
Explanation:
# Enter block when 'START TAG' is found
/START TAG/ {
N # Append next line to the pattern buffer
# Enter block when 'butterfly' is found
/butterfly/ {
:a # Create a label (could be also :foo)
# Enter block unless 'END TAG' is found
/END TAG/! {
N # Append next line to pattern buffer
ba # branch back to label :a
}
# Once 'END TAG' is found
d # Delete the pattern buffer
}
}
PS: A slightly modified version can be used to keep the START / END tags:
sed '/START/{p;N;/butterfly/{:a;/END/!{N;ba};s/.*\n//}}' file

Answer 2

Any time you find yourself saying sed multiline you are looking at the wrong tool. sed is for doing s/old/new on individual lines, that is all. For anything else you should be using awk.
$ cat tst.awk
{ rec = (rec=="" ? "" : rec ORS) $0 }
/END/ {
numLines = split(rec,lines,ORS)
print (lines[2] ~ /butterfly/ ? lines[1] ORS lines[numLines] : rec)
rec = ""
}
$ awk -f tst.awk file
START TAG
END TAG
START TAG
species: common lizard running
legs
tail
head
END TAG
START TAG
END TAG
START TAG
species: blackbird flying
legs
wings
head
END TAG
The above will work using any awk in any shell on any UNIX box, is clear, simple, robust and easily modified if/when your requirements change later to look at or print any of the other lines in each record or any combinations of values in each line or across the whole record. For example, to generate a CSV of just records where the creature has wings:
$ cat tst.awk
{ rec = (rec=="" ? "" : rec ORS) $0 }
/END/ {
numLines = split(rec,lines,ORS)
if ( lines[4] == "wings" ) {
sub(/species: /,"",lines[2])
for (i=2; i<numLines; i++) {
printf "%s%s", lines[i], (i<(numLines-1) ? "," : ORS)
}
}
rec = ""
}
$ awk -f tst.awk file
swallowtail butterfly flying,legs,wings,head
peacock butterfly resting,legs,wings,head
blackbird flying,legs,wings,head
and anything else you might want to do is equally trivial and consistent.