regx in shell script

[tanuj]

#!/bin/bash
cmd=`ps -ef | grep web | grep -v grep | awk '{print $2}'`
top -b -p $cmd -d 60 >> $1

#---End---#
In the above script first i will grep the process ids of all the processes running with the name 'web'.
and after that i will start a top process to capture the mem and other resources utilized by the process 'web'
-------------------
in my case cmd will return output as:
123
345
456
but as you can see in my top command , we need to pass process ids as:
top -b -p <pid1,pid2,pid3>
which means i need to pass something like: 123,345,456
so can someone please help me to write a regx to get the pids in above from.

Thanks
tanuj

[tanuj]

Got the solution, i was trying the same thing with before but was trying to replace a whitesapce with comma (,). after some attempt i tried replacing non-numeric char with comma (,) and that worked..

bash-3.00#

abc=`ps -ef | grep web | grep -v grep | awk '{print $2}'`

bash-3.00#

echo $abc
4954 4966 5023

bash-3.00#

echo $abc | sed s/[^0-9]/,/g
4954,4966,5023

bash-3.00#

[ddoxey]

That's a pretty neat idea.

How about a one-liner:

echo `ps -ef | grep chrom[e] | awk '{print $2}'` | sed 's/ /,/g' | awk '{print "top -b -p "$1}' | sh

(Finds instances of 'chrome' rather than 'web'.)

I just added it as a function in my .bashrc:

function top-for() {

    echo `ps -ef | grep $1 | grep -v grep | awk '{print $2}'` | sed 's/ /,/g' | awk '{print "top -b -p "$1}' | sh
}

Then you can just call it like:

top-for web

[nplusplus]

You know, if the command is available, you could use pgrep and get rid of that "ps -ef" and "grep -v grep" stuff.

[tanuj]

@ddoxey : really nice idea tried and it worked

[achaplot]

Hi,
pgrep is really a good option. You can also use "pgrep -d ," to return a "," separated list.

[tanuj]

@achaplot: cool this also worked. Thanks Man!

bash-3.00# ps -ef | pgrep -d "," web
230,231