UNIX for Dummies Questions & Answers

Hi

I have a shell script that looks for running processes in order to kill them. The script (ksh) gets the PID of these processes using the following:

Code:

PROCS=`ps -fu ${USERID} | egrep "MONITOR" | grep -v grep | cut -d" " -f4`

However, I spotted an issue where PID's of different lengths would cause the logic to fail (since the ps -fu command aligns columns.. therefore there is sometimes a double space between the UID and PID columns (issue a ps -fu $USER to see this)

So, I thought using sed to globally replace the double space with a single would be the way to go....

Code:

PROCS=`ps -fu ${USERID} | egrep "MONITOR" | grep -v grep | sed -e "s/  / /g" | cut -d" " -f3`

This returns the PID fine for instances where there is a double or single space between the UID and PID coulmns of the ps -fu command but fails if there are 3 or more spaces. I guess the sed pattern matching doesn't cope with instances like this (where, say 4 spaces should be replaced with a single one).

Any ideas on how to get around this problem using either the above as a starting point, or maybe an alternative approach?

Thanks in advance.

Mike

You could do something like:

Code:

ps -fu $USER | tr -s " " | cut -d" " -f3

the tr -s " "
command eliminates extra " " space characters

great.. that's what I wanted. I looked at the tr command but not with the -s switch. That now works fine, thanks.

Mike

Solaris ps command can do this....

ps -fo 'pid= args=' -u $procUser

Making the PID the first field in the output obviating cut, tr, etc. Note that you can specify the order of any field in the full output. In the above case I was only interested in the pid and args to the command, but all the others are available too...