formatting output in human readable numbers

Shell Programming and Scripting
1

Hi,

The following command provides the usage in 1024-byte blocks

du -ks * | sort -n | echo "$1"
...
1588820 user10
2463140 user11
2464096 user12
5808484 user13
6387400 user14
.....

I am trying to produce an output of first coulmn by multiplying by 1024 so that the output should display in bytes instead of 1024 kb blocks.

du -ks * | sort -n | awk '{print ($1 * 10242, $2)}'

getting the following output

1.62727e+10 user10
2.52275e+10 user11
2.52373e+10 user12
5.94905e+10 user13
6.54198e+10 user14

I even tried the printf as under:

du -ks * | sort -n | awk '{printf("%-20s%-10s\n", $1 * 1024, $2)}'

3.17764e+06 user10
4.92628e+06 user12
4.92819e+06 user13
1.1617e+07 user14
1.27748e+07 user15

Can some help to fix this output.

Thanks,
Ghazi

2

This may not be the solution that you are looking for, but it comes to mind.... you might consider du with the -b flag. From the man page:

3

I am on HP.

-b For each name operand that is a directory for
which file system swap has been enabled, print the
number of blocks the swap system is currently using.

-k Gives the block count in 1024-byte blocks.

$> ll
total 96
-rwxr-xr-x 1 user test 23677 Jan 8 15:55 file1
-rwxr-xr-x 1 user test 23040 Jan 8 15:56 file2

$> du -bs * | sort -n
48 file1 #comment: result of 23677/512
48 file2 #comment: result of 23040/512
$> du -ks * | sort -n
24 file1 #comment: result of 23677/1024
24 file2 #comment: result of 23040/1024

thanks,
Ghazi

4

Hmm... this is largely untested as I don't really have a heap of large directories to test on - however, something like

#!/bin/sh

for i in `ls /my/dirs/are/here`
do
   if [ -d "$i" ]; then
      result=`ls -lR "$i" | awk 'BEGIN {c=0;} {c+=$5} END{print c}'`
      echo "$result * 1024" | bc | awk -vdir=$i '{print $1, dir }'
   fi
done

will yield *approximate* values (not a true byte count as we're just multiplying the kilobyte count by 1024) on systems without GNU du for directories only....

Cheers
ZB

5

Nooop... it does not work for me...

Is there other command which can produce the disk usage
in bytes instead of blocks.

thanks

6

Why doesn't it work for you? What is wrong with the output? Is there any output at all? What error messages are you getting? etc...

I only tested this under Linux and UNICOS as I don't have access to an HP-UX box until Monday when I'm back at work.

Cheers
ZB

7

No there was no output. I did not get any error. The syntax looks good but does not do any thing.

Thanks,
Ghazi

8

When I used to work on HP-UX systems, the first thing I did was download a gcc development environment and build GNU utilities (like du, for example).

If I were you (of course I am not) I would create a GNU development environment for HP-UX and then you an have access to a very large (and free!) world of wonderful GNU tools that run on HP-UX.

See this link: http://www.gnu.org/ You will need to install GCC:

http://gcc.gnu.org/ ....find your HP-UX platform here:

http://gcc.gnu.org/install/specific.html

As an editorial note, back in the 'good ole days' when I was a UNIX systems consultant/contractor for big firms, I installed a GNU development environment will my clients. In every case, we saved the client time and big $$$ using GNU software.

Neo

9

du -b does not give bytes on HP-UX, BTW:

See:

http://docs.hp.com/en/B2355-90128/du.1.html

PS: See earlier post. I recommend you install GCC build some GNU utlities for HP-UX :slight_smile:

For an example HP-UX 10 build environment see:

http://supportweb.cs.bham.ac.uk/documentation/php/install.hpux.html

For an HP-UX 10 example environmental setup (from above):

10 
du -sk * | sort -n | awk '{print $1*1024 " " $2}'
11

If using awk, either change the value of OFMT (Output ForMaT) or just do something like...

du -ks * | sort -n | awk '{printf("%20d %-10s\n", $1 * 1024, $2)}'

12

Thanks to Ygor and others as well for helping me out.

With best regards,
Ghazi:)