Bash - find a keyword in a file and delete its line

Question

I'd like to give a keyword, find the line where this keyword aṕpears in a file and erase the entire line.

This is what I got but it is not working as it should:

KEYWORD='domain.com'
cat /etc/hosts | grep -v "$KEYWORD" > /etc/hosts

UPDATE

This is what worked for me:

sed -i_bak -e '/domain\.com/d' /etc/hosts

However, as I had two lines with "domain.com", is there a way to tell sed to erase only the line where the exact keyword "domain.com" appears

This is the original content of /etc/hosts:

127.0.0.1       localhost       localhost.localdomain
222.111.22.222      hvn.domain.com
222.111.22.222      domain.com

Here's how it end up after the command sed -i_bak -e '/domain\.com/d' /etc/hosts:

127.0.0.1       localhost       localhost.localdomain

I tried sed -i_bak -e '/\<namourphoto\.com\.br\>/d' ~/Desktop/hosts but it didn't work.

CONCLUSION

This is the code I came up with (based on the help the fine folks have given me):

D=domain.com
DOMAIN=`echo "$D" | sed 's/\./\\\\./g'`
sed -i_bak -e "/[\t]$DOMAIN/d" /etc/hosts

Note that:

I am counting that there is always a tab before the domain to be erased
I am automatically escaping the dots of the domain name.

Please also note that dot has special meaning in regexps so `domain.com` will also match `domainxcom.net`. You should escape the dot. — duri, Aug 13 '11 at 13:41
Yes, you do. Try `echo abc | grep 'a.c'`, the output will be `abc`. — duri, Aug 13 '11 at 15:25
add a space to the `[\t]` part, since /etc/hosts is whitespace separated. — chad, Aug 13 '11 at 20:46
What if I let the option of a space or a tab? (In my case, it's a tab). Would this be right? /[:space:\t]$DOMAIN/d — Roger, Aug 15 '11 at 13:46

miku · Accepted Answer · 2011-08-13T18:51:50.397

25

Use the stream editor, sed:

sed -i ".bak" '/culpa/d' test.txt

The above will delete lines containing culpa in test.txt. It will create a backup of the original (named test.txt.bak) and will modify the original file in-place.

edited Aug 13 '11 at 18:51

answered Aug 13 '11 at 13:42

miku

181,842
47
306
310

3

My tests didn't work with the space among -i and .bak. But when I pu -i.bak it did work. – Roger Aug 13 '11 at 17:05
how do you do this while discarding the original without making backup? UPDATE: just made the ".bak" quote blank to answer my own question. – Adam Johns Feb 05 '14 at 20:09
in my case keyword is ipaddress with `.` and port `:` IS this right? `sed -i "/$proxy[^0-9]/d" proxy.txt` – Luka Jan 19 '17 at 10:14

sidyll · Answer 2 · 2011-08-13T16:12:18.317

7

Pipe it to another file, not the same one that you're reading from, and be careful with the useless use of cat.

grep -v "$KEYWORD" /etc/hosts > newfile

Apart from the fine answer given regarding sed, you can also use Perl and a little improved regex to solve this:

perl -pi.old -e 's/.*\sdomain\.com\s*\n//' file

Notice I'm considering domain.com will be isolated by space characters (tabs or spaces, and so on), and that nothing but zero or more spaces will appear after it until the newline. Similarly to -i in sed,
-i.old in perl sets the $^I variable, and your original file will receive the changes while a copy will be kept under the old name and a .old appended.

edited Aug 13 '11 at 16:12

answered Aug 13 '11 at 13:37

sidyll

57,726
14
108
151

Is there a way to edit in place with this approach or I always need to make another file in order to edit it? – Roger Aug 13 '11 at 15:18
Very elegant your code, thank you. – Roger Aug 13 '11 at 16:54
@Roger: De nada, e bonito cachorro. Será que essa área verde aí atrás é aqui em SP? :P – sidyll Aug 14 '11 at 23:25
origado, é um Akita. Sim, a paisagem é no Morro do Ouro, em Apiaí; conhece? – Roger Aug 15 '11 at 13:44

score 1 · Answer 3 · answered Aug 13 '11 at 16:11

1

If you want to only delete last line of your example file

sed -i '/[[:space:]]\+domain\.com/d' test.txt

answered Aug 13 '11 at 16:11

Dotti

791
4
4

Bash - find a keyword in a file and delete its line

3 Answers3

Linked