12

I have a text file on Linux where the contents are like below:

help.helloworld.com:latest.world.com
dev.helloworld.com:latest.world.com

I want to get the contents before the colon like below:

help.helloworld.com
dev.helloworld.com

How can I do that within the terminal?

3
  • 2
    The grep utility is used for looking for lines matching regular expressions. You could possibly use it here, but it would be more appropriate to use a tool that extracts data from fields given some delimiter, such as the cut utility.
    – Kusalananda
    Commented Aug 27, 2019 at 17:23
  • I've submitted an edit to take out the word "grep" and replace it with "find" in the title and "get" in the question body, to avoid the X/Y issue of assuming grep is the right tool to solve the actual problem. Commented Aug 28, 2019 at 18:21
  • 1
    All I can say is that the contents before the colon is much better than the contents after the colon ;-). Commented Aug 30, 2019 at 14:02

7 Answers 7

40

This is what cut is for:

$ cat file
help.helloworld.com:latest.world.com
dev.helloworld.com:latest.world.com
foo:baz:bar
foo

$ cut -d: -f1 file
help.helloworld.com
dev.helloworld.com
foo
foo

You just set the delimiter to : with -d: and tell it to only print the 1st field (-f1).

20

Or an alternative:

$ grep -o '^[^:]*' file
help.helloworld.com
dev.helloworld.com

This returns any characters beginning at the start of each line (^) which are no colons ([^:]*).

19

Would definitely recommend awk:

awk -F ':' '{print $1}' file

Uses : as a field separator and prints the first field.

0
5

updated answer

Considering the following file file.txt:

help.helloworld.com:latest.world.com
dev.helloworld.com:latest.world.com
no.colon.com
colon.at.the.end.com:

You can use sed to remove everything after the colon:

sed -e 's/:.*//' file.txt

This works for all the corner cases pointed out in the comments—if it ends in a colon, or if there is no colon, although these weren't mentioned in the question itself. Thanks to @Rakesh Sharma, @mirabilos, and @Freddy for their comments. Answering questions is a great way to learn.

5
  • 4
    sed -e 's/:.*//' file.txt is another way with Posix sed. Commented Aug 28, 2019 at 4:02
  • 1
    sed -ne 'y/:/\n/;P' file.txt also can be used. Commented Aug 28, 2019 at 4:05
  • Make .+ to .* Commented Aug 28, 2019 at 4:37
  • @Randy Joselyn Since there's an implicit if in the s///p syntax, you need to modify your regex to take care of lines with no colons, something like, sed -nEe 's/([^:]*)(:.*|)/\1/p'. Note this requires GNU sed but since anyway you are on GNU sed so this shouldn't matter. Commented Aug 28, 2019 at 5:05
  • This answer could have been my favourite, but the ERE are unnecessary. sed -n '/:/s/^\([^:]*\):.*$/\1/p (add --posix if you use GNU sed, just to spite the extensionism of theirs)
    – mirabilos
    Commented Aug 28, 2019 at 18:09
4

Requires GNU grep. It would not work with the default grep on e.g. macOS or any of the other BSDs.

Do you mean like this:

grep -oP '.*(?=:)' file

Output:

help.helloworld.com
dev.helloworld.com
2
  • 4
    If there are two or more colons on the line, this will print everything until the last one, so not what the OP needs. Try echo foo:bar:baz | grep -oP '.*(?=:)'. This will work for the OP's example, but not for the general case as described in the question.
    – terdon
    Commented Aug 27, 2019 at 17:19
  • there is only one colon and its working fine , but thanks for the update Commented Aug 27, 2019 at 17:25
-2

You could achieve this with bash string handling, by removing the longest match from the string directly for each line read like so:

for line in $(cat inputfile); do echo "${line%%:*}"; done

This might be a useful alternative if you are parsing the file in a shell script (though I suspect using cut might be more efficient).

1
-2

In pure POSIX shell without using external commands, I'd do:

#/bin/sh
IFS=:
while read -r a _; do
  echo "$a"
  done < file.txt
unset IFS
1

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .