Shell Programming and Scripting

Here is an interesting problem that I am seeing when using 'sed'. The goal
is to print out the names of the Oracle SID's on a system and here is the
command I am using:

Code:

ps -ef|grep ora|grep pmon|awk '{print $NF}'|sed 's/ora_pmon_//'

The output of the command after the awk is this:

Code:

$ ps -ef|grep ora|grep pmon|awk '{print $NF}'
ora_pmon_test1
ora_pmon_devel1
ora_pmon_devel2

After this the 'sed' command is supposed to replace 'ora_pmon_' with
empty string so that we get:

test1
devel1
devel2

Unfortunately, in addition to doing this, it is also adding one extra line in
the output:

test1
devel1
devel2
s///

I am running this on an AIX system. Anyone has any idea why the extra 's///'
is showing up?

Thanks.

The sed sub process was picked up by the ps also. Try:

Code:

ps -ef|awk '{if(sub(/[o]ra_pmon_/,"",$NF))print $NF}'

Works perfectly. Thanks for the explanation!!

---------- Post updated at 06:46 PM ---------- Previous update was at 06:11 PM ----------

One quick question - why do we have to use '[o]ra_pmon_' in the regular expression part? Why not just 'ora_pmon_'?

Does 'o' has some special meaning in 'sed'?

If you use ora_pmon, then the filtering expression itself will appear in the ps -ef listing and match itself. With [o]ra_pmon this is will not happen.

For fun, one alternative using sed

Code:

ps -ef|sed -n 's/.*[o]ra_pmon_//p'