grep or AWK help

[syntax]

So, a little help if any one can.

I have a file which contains for example

testA
cat
testB
1
testC
cat
testD
dog

i need to search through this file, and display testX only if it DOESNT have the word cat beneath it.

so, the logic would be cat file | grep -v -B1 cat >>output.txt

now i know this doesnt work, but i need something similar.
I cant search the other way round, because the words under testX change or arent there at all. The only constant is the word cat.

So i need to search for the word cat, and then exclude the line above it.

Any help would be massively appreciated!

 

[crayons610]

pcregrep -M '^.+\ncat$' test.txt | grep -v cat

pcregrep uses Perl compatible regular expressions.
The regex looks for a line with a carriage return and then another line with just 'cat'.
The double line results are piped to grep -v which excludes matches.
You might need to change \r to \r\n depending of the platform ( windows / unix ).

HTH

 

[syntax]

thanks, that kind of works, but does the opposite of what I want.

That returns
testA and testC

I want it to return testB and testD

So anywhere I find cat, i want to exclude cat and the word above.

Also note, that the string might not be a single word, so below. I still need a way to search for the word cat, and if i find it, exclude that line and the line above.

testA
this is a cat
testB
1
testC
this is a cat
testD
this is a dog

 

[internetless]

Hi,

Try this:

pcregrep -M '(^test.+\n)(?!^cat$)' file

 

[syntax]

Hi,

Try this:

pcregrep -M '(^test.+\n)(?!^cat$)' file

hmm, that just outputs everything that has test in it and the line below unfortunately.
It doesnt output test when there isnt the word cat below

 

[internetless]

Hi again,

Sorry, that worked on my Mac, but I've now tested on Linux & got different results :-(

Let's try the long way:

perl -ne 'chomp; if (defined($line1) && ($_ ne 'cat')) {printf "%s\n",$line1;} $line1 = (($_ =~ m/^test/o) ? $_ : undef);' file

 

[Nephew_]

Did you come right here?
This is kinda driving me mad , trying to solve the "puzzle", as I try to do it with a one-liner in sed.

 

[porchrat]

Why not just grep for dog and use that option to grab an extra peceeding line?... If I recall correctly the option is -B

EDIT: oh wait I see you have more than just dog as a possibility. Carry on.

 

[Nephew_]

And just as I ask, I find a solution on LQ.

sed -n 'h;n;H;g;/cat/!p' file

here:
http://www.linuxquestions.org/quest...ove-or-below-as-well-as-matching-line-622123/
Logic of the command:
".. it reads a line, puts it in the hold buffer, appends the next line, gets the combo back to pattern space and prints it if it does NOT contain 'cat'"

I use sed everyday, but I need to go advanced

 

[internetless]

The pattern matching works in this solution, but the output includes the lines containing "1" and "dog". The OP wants just "testB" and "testD" if I understand correctly.