Sed multiline problem

mvairavan · 07-14-2012, 11:59 AM

I have to find pattern which is followed by pattern2 on the 3 rd line.

so if pattern1 comes in line10 there should be pattern2 in line 13.

so i got
sed ' /PATTERN1/{
N
/.*/{
N
/.*/{
N
/PATTERN2/ p
}
}
} file

It kinna works but the problem is when pattern2 isn't matched i wanted it start searching for pattern1 from wr it left while searching for pattern1.

i.e., if pattern 1 found on line 10 and pattern 2 isn't found on line 13 it should search for pattern1 again from line 11 rather its not starting from line 14 i guess.

Thanks in advance.

Kenhelm · 07-14-2012, 02:55 PM

This keeps a sliding window of four lines in the pattern space as it goes through the file.
The '=' command prints out the line number of the last line of each matched group of four lines.
It finds overlapping matches, as shown in this example.

Code:

echo 'PATTERN1
text
PATTERN1
abc
PATTERN1
PATTERN2
def
PATTERN2
text' | sed -n '
1{N;N}
N
/^PATTERN1.*PATTERN2$/{p;=}
D'

PATTERN1
abc
PATTERN1
PATTERN2             # Line 6
6
PATTERN1
PATTERN2             # This is also line 6
def
PATTERN2             # Line 8
8

David the H. · 07-15-2012, 08:39 AM

Please use ***[code][/code] tags*** around your code and data, to preserve formatting and to improve readability. Please do not use quote tags, bolding, colors, or other fancy formatting.

Vague descriptions mean people have to make guesses and assumptions, which leads to vague or unhelpful answers. If you'd provide an actual example of input text, and show us what needs to be matched/extracted/changed/whatever, maybe we could provide better solutions. sed isn't a particularly good multi-line editor, and perhaps awk or another tool would end up being more suitable.