Is there a way to specify multiple spaces as a field delimiter with the cut
command (something like a " "+
regex)? For example, what field delimiter I should specify for the following string to reach value 3744
?
$ps axu | grep jboss
jboss 2574 0.0 0.0 3744 1092 ? S Aug17 0:00 /bin/sh /usr/java/jboss/bin/run.sh -c example.com -b 0.0.0.0
cut -d' '
is not what I want, because it's only for a single space. awk
is not what I am looking for either, so how to do this with cut
?
Actually awk
is exactly the tool you should be looking into:
ps axu | grep '[j]boss' | awk '{print $5}'
or you can ditch the grep
altogether since awk
knows about regular expressions:
ps axu | awk '/[j]boss/ {print $5}'
But if, for some bizarre reason, you really can't use awk
, there are other simpler things you can do, like collapse all whitespace to a single space first:
ps axu | grep '[j]boss' | sed 's/\s\s*/ /g' | cut -d' ' -f5
That grep
trick, by the way, is a neat way to only get the jboss
processes and not the grep jboss
one (ditto for the awk
variant as well).
The grep
process will have a literal grep [j]boss
in its process command so will not be caught by the grep
itself, which is looking for the character class [j]
followed by boss
.
This is a nifty way to avoid the | grep xyz | grep -v grep
paradigm that some people use.