[SOLVED] BASH: Reformatting text output to join lines

inversecow · 01-18-2012, 03:37 PM

Hello all,

I am working on a script to collect I/O metrics and am having trouble reformatting output from the iostat command.
Specifically, some "normally single line" rows of data are being placed as double line results (OUTPUT 1).
I expect this has to do with output formatting rules within the utility itself, but have been unable to "unify" the two lines back into one.

What are some recommended solutions to obtain a desirable result (EG: RESULT 1)?
I would like to use tools suitable to include in a command pipeline if possible (EG: sed, tr, awk, etc).
The spacing is less important as I plan to cut the results up with awk and convert to CSV.

# OUTPUT 1

Code:

[20:42:45:user@hostname:~]
$ /usr/bin/iostat -d -x
...
cciss/c0d0        1.44    22.00  3.19  6.78   273.78   190.25    46.53     0.36   36.42   3.82   3.81
cciss/c0d0p1
0.00     0.00  0.00  0.00     0.03     0.00    18.12     0.00   27.78  26.17   0.01
cciss/c0d0p2
0.55     7.65  1.08  1.54   103.03    69.58    66.01     0.16   61.01   4.80   1.26
cciss/c0d0p3
0.88    14.35  2.11  5.24   170.72   120.67    39.62     0.20   27.68   4.22   3.10
...

# RESULT 1

Code:

[20:42:45:user@hostname:~]
$ /usr/bin/iostat -d -x | $EXTRA_COMMANDS
...
cciss/c0d0        1.44    22.00  3.19  6.78   273.78   190.25    46.53     0.36   36.42   3.82   3.81
cciss/c0d0p1      0.00     0.00  0.00  0.00     0.03     0.00    18.12     0.00   27.78  26.17   0.01
cciss/c0d0p2      0.55     7.65  1.08  1.54   103.03    69.58    66.01     0.16   61.01   4.80   1.26
cciss/c0d0p3      0.88    14.35  2.11  5.24   170.72   120.67    39.62     0.20   27.68   4.22   3.10
...

Thanks for your time.

thesnow · 01-18-2012, 03:59 PM

Does this happen only when you output to screen, or also when you redirect to a file?

inversecow · 01-18-2012, 04:01 PM

Quote:

Originally Posted by thesnow

Does this happen only when you output to screen, or also when you redirect to a file?

Hello thesnow,

Thanks for the above reply.
Indeed this is a problem which occurs in both scenarios (screen && file).
I am writing to a file with a simple STDOUT redirection (

Code:

command > file.txt

).

Nominal Animal · 01-18-2012, 06:04 PM

Quote:

Originally Posted by inversecow

Specifically, some "normally single line" rows of data are being placed as double line results (OUTPUT 1).

It seems iostat is splitting the line because the device name is too long. The splits occur at the start of the line, just after the device name, and not later on in the line.

Quote:

Originally Posted by inversecow

What are some recommended solutions to obtain a desirable result (EG: RESULT 1)?

Pipe through

Code:

awk '{ if (NF>1) printf("%s\n", $0); else printf("%s\t", $0); }'

Quote:

Originally Posted by inversecow

The spacing is less important as I plan to cut the results up with awk and convert to CSV.

In that case, you can use a "pre-rule" to merge those lines, so your awk script will see correct records:

Code:

awk '(NF==1) { getline temp; $0 = $0 "\t" temp; }
     { # your actual rule goes here
       printf("\"%s\",%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n", $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12);
     }'

Note that before running the iostat and the awk script, you should probably set LANG=C and LC_ALL=C in the environment, to make sure the output does not depend on locale. For example, in my default locale, the decimal point is , . If you use Bash, start it with

Code:

#!/bin/bash
export LANG=C LC_ALL=C

To set it for just the iostat call, use

Code:

env LANG=C LC_ALL=C iostat -d -x

inversecow · 01-18-2012, 06:25 PM

Quote:

Originally Posted by Nominal Animal

It seems iostat is splitting the line because the device name is too long. The splits occur at the start of the line, just after the device name, and not later on in the line.

Pipe through

Code:

awk '{ if (NF>1) printf("%s\n", $0); else printf("%s\t", $0); }'

In that case, you can use a "pre-rule" to merge those lines, so your awk script will see correct records:

Code:

awk '(NF==1) { getline temp; $0 = $0 "\t" temp; }
     { # your actual rule goes here
       printf("\"%s\",%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n", $1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12);
     }'

Note that before running the iostat and the awk script, you should probably set LANG=C and LC_ALL=C in the environment, to make sure the output does not depend on locale. For example, in my default locale, the decimal point is , . If you use Bash, start it with

Code:

#!/bin/bash
export LANG=C LC_ALL=C

To set it for just the iostat call, use

Code:

env LANG=C LC_ALL=C iostat -d -x

Hello Nominal Animal,

Thanks for the in-depth response.
This fixed my issue beautifully! :-D