5

I want to remove the first two words that come up in my output string. this string is also within another string.

What I have: 

for servers in `ls /data/field`    
do    
  string=`cat /data/field/$servers/time`

This sends this text: 

00:00 down server

I would like to remove "00:00 down" so that it only displays "server".

I have tried using cut -d ' ' -f2- $string which ends up just removing directories that the command searches.

Any ideas?

midori
  • 4,807
  • 5
  • 34
  • 62
hisere
  • 53
  • 1
  • 1
  • 6

4 Answers4

10

Please, do the things properly :

for servers in /data/field/*; do
    string=$(cut -d" "  -f3- /data/field/$servers/time)
    echo "$string"
done
  • backticks are deprecated in 2014 in favor of the form $( )
  • don't parse ls output, use glob instead like I do with data/field/*

Check http://mywiki.wooledge.org/BashFAQ for various subjects

Gilles Quénot
  • 173,512
  • 41
  • 224
  • 223
  • I must admit, I have the same bad habit, so I went about looking for the reason why one shouldn't use ls in this situation. So thanks for that link +1. Anyways, the FAQ is rather large, so to save everyone else some time, why you shouldn't use ls, see: http://mywiki.wooledge.org/ParsingLs – Bitdiot Dec 15 '14 at 21:19
6

Use -d option to set the delimtier to space

$ echo 00:00 down server | cut -d" "  -f3-
server

Note Use the field number 3 as the count starts from 1 and not 0

From man page

  -d, --delimiter=DELIM
              use DELIM instead of TAB for field delimiter

   N-     from N'th byte, character or field, to end of line

More Tests

$ echo 00:00 down server hello world| cut -d" "  -f3-
server hello world

The for loop is capable of iterating through the files using globbing. So I would write something like

for servers in /data/field*
do    
  string=`cut -d" "  -f3- /data/field/$servers/time`
...
...
nu11p01n73R
  • 26,397
  • 3
  • 39
  • 52
2

You can use sed as well:

sed 's/^.* * //'
midori
  • 4,807
  • 5
  • 34
  • 62
0

For the examples given, I prefer cut. But for the general problem expressed by the question, the answers above have minor short-comings. For instance, when you don't know how many spaces are between the words (cut), or whether they start with a space or not (cut,sed), or cannot be easily used in a pipeline (shell for-loop). Here's a perl example that is fast, efficient, and not too hard to remember:

 | perl -pe 's/^\s*(\S+\s+){2}//'

Perl's -p operates like sed's. That is, it gobbles input one line at a time, like -n, and after dong work, prints the line again. The -e starts the command-line-based script. The script is simply a one-line substitute s/// expression; substitute matching regular expressions on the left hand side with the string on the right-hand side. In this case, the right-hand side is empty, so we're just cutting out the expression found on the left-hand side.

The regular expression, particular to Perl (and all PLRE derivatives, like those in Python and Ruby and Javascript), uses \s to match whitespace, and \S to match non-whitespace. So the combination of \S+\s+ matches a word followed by its whitespace. We group that sub-expression together with (...) and then tell sed to match exactly 2 of those in a row with the {m,n} expression, where n is optional and m is 2. The leading \s* means trim leading whitespace.

Otheus
  • 785
  • 10
  • 18