help on sed command

pavan27 · 02-03-2011, 04:26 PM

hello experts,
i am looking for help on shell scripting. here is my requirement.
I have a file called blackout.txt and content of the file is

Quote:

Oracle Enterprise Manager 10g Release 5 Grid Control 10.2.0.5.0.
Copyright (c) 1996, 2009 Oracle Corporation. All rights reserved.
[orclstg:3872, oracle_emd]
[LISTENER_ORA_ORCLSTG_orclstg, oracle_listener]
[orclstg, host]
[orclstg, oracle_database]

now i need to remove the first three lines and first character and last character in each line. and output should come like this(in one line with one space between each word)
LISTENER_ORA_ORCLSTG_orclstg orclstg orclstg
i was failed to get all three lines in one line. could you please help me.

i have done upto now

Quote:

blackhoney:{oracle}269% sed '1,3d' blackout_txt | awk '{print $1}' | sed 's/.$.*$/\1/' | sed 's/$.*$./\1/'
LISTENER_BMC_EURSTG_bmeurstg
bmeurstg
BMEURSTG.WORLD

please help me.

colucix · 02-03-2011, 04:42 PM

You can do all the work with awk:

Code:

awk 'NR > 3 { gsub(/^.|.$/,"",$1); printf "%s ", $1 } END { printf "\n" }' blackout.txt

z1p · 02-03-2011, 04:47 PM

Good start, but a couple of things.

First combining the seds together will make it more efficient.
Next by default awk will use a newline between output records. You can either strip them out later. Setup your awk to concatenate then print the output, or tell awk to use a different record delimiter. I went with the latter.

Code:

sed -e '1,3d' -e's/.\(.*\)/\1/' -e 's/\(.*\)./\1/' blackout.txt | awk -F, -v ORS=' ' '{print $1}'

oops. looks like colucix beat me to it... and it looks like I need to expand my awk skills.

colucix · 02-03-2011, 04:58 PM

Quote:

Originally Posted by z1p

oops. looks like colucix beat me to it... and it looks like I need to expand my awk skills.

Actually you introduced another way by changing the ORS value. In any case, we have to add a newline at the end - if required.

pavan27 · 02-03-2011, 04:59 PM

Thanks for Quick reply.
i am getting following error. please help me where i am doing wrong. i am bit poor in sed and awk command

Quote:

blackhoney:{oracle}: sed -e '1,3d' -e's/.$.*$/\1/' -e 's/$.*$./\1/' blackout_txt | awk -F, -v ORS=' ' '{print $1}'
awk: syntax error near line 1
awk: bailing out near line 1

z1p · 02-03-2011, 05:06 PM

I cut and pasted that line straight into my terminal window and it works.

What OS and shell are you using?

pavan27 · 02-03-2011, 05:08 PM

Quote:

Originally Posted by z1p

I cut and pasted that line straight into my terminal window and it works.

What OS and shell are you using?

shell is :
echo $shell
/usr/bin/tcsh

OS is Solaris 10

colucix · 02-03-2011, 05:31 PM

What about this?

Code:

awk 'NR > 3 { sub(/^./,"",$1); sub (/.$/,"",$1); printf "%s ", $1 } END { printf "\n" }' blackout.txt

This is a slightly modified version of my previous one, since I'm not sure the Solaris' awk can manage the alternative regular expression (see the pipe in the regexp of the gsub statement). Eventually you can try nawk which is more compatible with GNU awk (that is the linux version).

pavan27 · 02-03-2011, 05:50 PM

getting errors.

Quote:

blackhoney:{oracle}317% awk 'NR > 3 { sub(/^./,"",$1); sub (/.$/,"",$1); printf "%s ", $1 } END { printf "\n" }' blackout_txt
awk: syntax error near line 1
awk: illegal statement near line 1
awk: syntax error near line 1
awk: illegal statement near line 1

i have tired different way, but i am not getting what i am looking for.

Quote:

{oracle}316% sed -e '1,3d' -e's/.$.*$/\1/' -e 's/$.*$./\1/' -e 's/^ *[^ ]* //' blackout_txt
oracle_listener
host
oracle_database

jschiwal · 02-03-2011, 06:13 PM

Your version of sed may not have the -e option. Try using a semicolon to separate sed commands.
sed '1,3d;s/.$.*$.$/\1/' blackout_txt

Also use the 'od' or hexdump commands to print out a character by character output of a couple lines of the input file. A non-printable character or different line endings may be tripping you up.

sed -n '4p' blackout_txt | od -c

crts · 02-03-2011, 06:51 PM

Here's another sed:

Code:

sed -nr '4,$ {s/\[([^,]+).*$/\1 /;H};$ {g;s/\n//gp}' blackout_txt

z1p · 02-03-2011, 07:03 PM

Its the assignment using the -v that awk doesn't like. try:

Code:

sed -e '1,3d' -e's/.\(.*\)/\1/' -e 's/\(.*\)./\1/' blackout_txt | awk -F, '{ORS=" "} {print $1} END {print "\n"} '

grail · 02-03-2011, 08:34 PM

Assuming the file is comma separated for all the data we are looking at, getting rid of character at the end of the line is a waste of time (with awk) as
it is never called. So just:

Code:

awk -F"[][,]" 'NR > 3{printf $2" "}END{print ""}' file

kurumi · 02-03-2011, 09:07 PM

Quote:

Originally Posted by crts

Here's another sed:

Code:

sed -nr '4,$ {s/\[([^,]+).*$/\1 /;H};$ {g;s/\n//gp}' blackout_txt

from memory, don't think Solaris's sed has the -r option. i may be wrong

kurumi · 02-03-2011, 09:08 PM

Quote:

Originally Posted by grail

Assuming the file is comma separated for all the data we are looking at, getting rid of character at the end of the line is a waste of time (with awk) as
it is never called. So just:

Code:

awk -F"[][,]" 'NR > 3{printf $2" "}END{print ""}' file

does Solaris's awk or nawk allow multiple field delimiters? I don't have nawk at hand to check.