[SOLVED] In need of some SED assistance

EricTRA · 07-22-2010, 02:25 AM

Hi all,

I'm not that familiar with sed and awk in order to be able to solve this problem on my own, so I'm calling on you for a bit of assistance.

I'm writing a Nagios plugin to check our Oracle tablespaces and the output is given in one line like this:

Code:

TABLESPACE_NAME USED_PERCENT ------------------------------ ------------ AXIONAL_INDEX 62.2645833 DATA 71.926875 SYSAUX .885200923 SYSTEM 1.04007771 TEMP 0 UNDOTBS1 .005340579 USERS 0 7 rows selected.

I've been playing around with sed like below to delete the obsolete info and change every second space into a newline:

Code:

cat myfile | sed -e 's/TABLESPACE_NAME USED_PERCENT ------------------------------ ------------ //' -e 's/ /\n/2' -e 's/ /\n/3' -e 's/ /\n/4' -e 's/ /\n/5' -e 's/ /\n/6' -e 's/ /\n/7' -e 's/ /\n/8'

which gives me an expected result:

Code:

AXIONAL_INDEX 62.2645833
DATA 71.926875
SYSAUX .885200923
SYSTEM 1.04007771
TEMP 0
UNDOTBS1 .008392338
USERS 0
7 rows selected.

My 'problem' is that I don't know up front how many tablespaces there are so I'd have to check all databases and 'hardcode' the tablespaces in my script. Is there any way to 'automate' this knowing that 'rows selected' preceded by a number is always the last line and using a sort of counter to auto-adjust the number to put in the -e 's/ /\n/2' part?

Also if you know of a 'cleaner' solution (I'm sure some of you can come up with a nifty one-liner) I'd appreciate the help.

Kind regards,

Eric

druuna · 07-22-2010, 03:01 AM

Hi,

Is this what you are looking for:

cut -c74- infile | sed 's/$[A-Z][_A-Z0-9]* [0-9\.][0-9\.]*$ /\1\n/g'

I assumed that the first 74 characters are the header and always the same length (the cut command).

The sed command uses back-referencing (that what is found in the search part, all between $ and $, is used as the replace part, the \1 part).

The sed command looks for tablenames that contain these possible characters: Starts with A-Z, followed by A-Z, 0-9 or a _ (zero or more times)
This must be followed by a space and 0-9 or a dot followed by 0-9 and/or a dot (zero or more times). followed by a space.

What is found is printed and folowed by a return (the \n).

Example run:

Code:

$ cat infile
TABLESPACE_NAME USED_PERCENT ------------------------------ ------------ AXIONAL_INDEX 62.2645833 DATA 71.926875 SYSAUX .885200923 SYSTEM 1.04007771 TEMP 0 UNDOTBS1 .005340579 USERS 0 7 rows selected

TABLESPACE_NAME USED_PERCENT ------------------------------ ------------ AXIONAL_INDEX 62.2645833 DATA 71.926875 SYSAUX .885200923 SYSTEM 1.04007771 TEMP 0 UNDOTBS1 .005340579 USERS 0 DUMMY_A 1 DUMMY_B 2.2 9 rows selected

$ cut -c74- infile | sed 's/\([A-Z][_A-Z0-9]* [0-9\.][0-9\.]*\) /\1\n/g'
AXIONAL_INDEX 62.2645833
DATA 71.926875
SYSAUX .885200923
SYSTEM 1.04007771
TEMP 0
UNDOTBS1 .005340579
USERS 0
7 rows selected

AXIONAL_INDEX 62.2645833
DATA 71.926875
SYSAUX .885200923
SYSTEM 1.04007771
TEMP 0
UNDOTBS1 .005340579
USERS 0
DUMMY_A 1
DUMMY_B 2.2
9 rows selected

Hope this helps.

EricTRA · 07-22-2010, 03:12 AM

Hi druuna,

Thank you so much, that's exactly what I need. And I realy appreciate the detailed explanation. Yet learned a lot more for future use. Thanks again.

Kind regards,

Eric

druuna · 07-22-2010, 03:18 AM

You're welcome

grail · 07-22-2010, 03:50 AM

Just as an alternative to have a gander at:

Code:

awk '{for(x=0; x < (NF - 5);x+=2)print $(x + 5),$(x + 6)}' in_file

It doesn't show the word 'selected' on the last line, but not sure if you needed that anyway

EricTRA · 07-22-2010, 04:35 AM

Hi grail,

Thanks for the awk alternative. Could you elaborate some more on that construction so that it makes sence to me too?

Kind regards,

Eric

grail · 07-22-2010, 06:44 AM

Sure:

1. for - most similar to a C for loop construct

2. NF - this is awk for the number of fields. In awk the standard delimeter for a field is whitespace, so in your example the simple space separates each field

3. x+=2 - step through loop in units of 2 as we wish to print pairs of information

4. $(x + 5) - the $ is similar to the one in bash that it evaluates and in this case it will evaluate the fields we wish to look at - first being 5 and 6

5. , - seems funny to comment on a comma but in awk this will enable print to output the OFS (output field separator), the standard being a space

So overall it says - loop through each line of the file and starting at positions 5 and 6 print the two fields separated by a space

EricTRA · 07-22-2010, 01:46 PM

Hi grail,

Thanks a lot for the explanation!

Reading up on the Grymoire manuals in regards to sed and awk at this moment. Thanks again guys for the help and sharing your knowledge.

Kind regards,

Eric