Combining -v flag and -A flag in grep

Question

I need to search a file for a string, remove any line that contains the string, and also remove the two lines following any line that contains the string. I was hoping I could accomplish this using something like this...

$ grep -v -A 2 two temp.txt
one
five
$

...but unfortunately this did not work. Is there a simple I can do this with grep or another shell command?

score 7 · Accepted Answer · edited Apr 17 '13 at 17:36

7

The following works both with GNU sed and with OS X.

$ sed '/two/{N;N;d;}' temp.txt
one
five

find line matching two
read in two more lines
delete them

edited Apr 17 '13 at 17:36

Daniel Standage

8,136
19
69
116

answered Apr 17 '13 at 01:54

Zombo

1
62
391
407

Both your answer and @JohnKugelman's are concise, they both work on GNU/Linux, and neither work on OS X (not a huge deal for me for this particular task). For providing a brief explanation of the syntax, the accepted answer goes to you. Thanks. – Daniel Standage Apr 17 '13 at 02:48
2

Add a semicolon after the d and. Enforce the brace and it should work on MacOS X too. – Jonathan Leffler Apr 17 '13 at 05:03

John Kugelman · Answer 2 · 2013-04-17T03:00:27.133

3

With GNU sed:

sed '/two/,+2d' temp.txt

This uses two-address syntax (addr1,addr2) to match lines with the word two (/two/) plus the two lines after (+2). The d command deletes those lines.

edited Apr 17 '13 at 03:00

answered Apr 17 '13 at 01:55

John Kugelman

349,597
67
533
578

The only thing I can think of is that maybe they're using a older, non-GNU sed, without the `+number` option. But since it works (and I dislike drive-by downvoters), here's a vote to reverse it. – paxdiablo Apr 17 '13 at 02:31

paxdiablo · Answer 3 · 2013-04-17T02:02:04.157

3

You can do this with awk, as per the following transcript:

pax> echo 'one
two
three
four
five' | awk '/two/ {skip=3} skip>0 {skip--;next} {print}'

one
five

It basically starts a counter of lines to throw away (3) whenever it finds the two string on a line. It then throws those lines away until the skip counter reaches zero. Any line that isn't marked for skipping is printed.

edited Apr 17 '13 at 02:02

answered Apr 17 '13 at 01:56

paxdiablo

854,327
234
1,573
1,953

This has the advantage of handling `one two three two three four five` → `one five` correctly. – John Kugelman Apr 17 '13 at 02:40

score 1 · Answer 4 · answered Apr 17 '13 at 02:59

Here's a way to do it with Perl:

$ perl -ne'if (/two/){$x=<>;$x=<>;}else{print}' temp.txt 
one
five

The -n is an implicit loop over the input. If you match /two/, then read the next two lines, otherwise print the line you're on.

The problem is, however, that if you had the third or fourth lines matched /two/, then you would still get the same output. @paxdiablo's solution is more complete. But mine's more Q&D.

Combining -v flag and -A flag in grep

4 Answers4

Linked