Why is a variable behaving differently in ksh script.

Guys i have strange behaviour with command output being saved in a variable instead of a tmp file.

  1. I suck command output into a variable

Sample command output

# cleanstats

DRIVE INFO:
----------

Drv    Type      Mount Time  Frequency   Last Cleaned         Comment
***    ****      **********  *********   ****************     *******
  0    hcart3*   51.9        0                N/A
  1    hcart3*   55.9        0                N/A
  2    dlt*      0.8         0                N/A
  3    dlt*      0.0         0                N/A
  4    dlt*      0.2         0                N/A
  5    dlt*      0.0         0                N/A

MEDIA INFO:
----------

media   media  robot  robot  robot  side/  optical  # mounts/      last
 ID     type   type     #    slot   face   partner  cleanings    mount time
-------------------------------------------------------------------------------
CLN206  DLT_CL NONE     -      -     -       -           0     12/26/2001 08:22
CLN207  DLT_CL NONE     -      -     -       -           0     03/10/2002 10:00
CLN205  DLT_CL NONE     -      -     -       -           0     08/18/2002 06:40
CLN703  DLT_CL NONE     -      -     -       -           0     03/29/2003 05:11
CLN701  DLT_CL NONE     -      -     -       -          20     00/00/0000 00:00
CLN635  DLT_CL NONE     -      -     -       -           0     11/21/2003 04:32
CLN219  DLT_CL NONE     -      -     -       -           0     07/13/2004 06:25
CLN636  DLT_CL NONE     -      -     -       -          14     01/22/2006 08:15
CLN211  DLT_CL TLD      1     21     -       -          12     10/31/2008 22:44
CLN209  DLT_CL NONE     -      -     -       -           0     12/13/2006 22:13
CLN210  DLT_CL NONE     -      -     -       -           0     10/19/2008 06:31

  1. I then print that variable to screen for debug and its as expected
  2. I then print that variable to an awk statement in an if test and it doesn't work as it should.
  3. If the output is captured in a tempfile and the same awk statement is used but taking input from the tmpfile then it does work.

The below code shows one way which doesn't work and one which does. Can anyone shed any light as to why?
I'm obviously trying to use variables instead of tmpfiles everywhere.
I can't understand it because the print statement for debug shows the output as expected.

#!/bin/ksh

....script contents, variable assignment blah blah....

# Suck clean stats to variable
CSTAT=$(cleanstats)

# Or put in tmpfile
cleanstats > ${TMPFILE}

print "CSTAT contains [${CSTAT}]"  #DBG

# Check tape exists in library with free cleanings
# If tape is in unit but with no cleanings OR
# If tape is NOT in unit (TLD), the string returned is empty thus matching -z test

# (section 1)
if [[ -z "$( print ${CSTAT} | nawk '$3 == "TLD" && $8 != "0" {print}' )" ]];then
      .... do stuff ....
fi

# Above doesnt work, but below does

# (section 2)
#if [[ -z "$( nawk '$3 == "TLD" && $8 != "0" {print}' < ${TMPFILE} )" ]];then
#      .... do stuff ...
#fi

Using the sample output above will cause the first section to match when it shouldn't, yet the second section works and just passes by the if.

Anyone shed any light on it? IFS is set as newline as standard.

Cheers

Oddly enough just setting IFS to nothing seems to have fixed it.

I'm not entirely sure why, but gathered it would around the field separator.

I tried doing this :-

oldIFS=$IFS
IFS="\n\r"

if [ ....blah

IFS=$oldIFS

.and noticed it worked correctly but the print output from awk was missing n's and r's.

Through trial and error i found setting

IFS="" 

made it work.

Anyone know why?

cheers