How can I sort du -h output by size

As of GNU coreutils 7.5 released in August 2009, sort allows a -h parameter, which allows numeric suffixes of the kind produced by du -h:

du -hs * | sort -h

If you are using a sort that does not support -h, you can install GNU Coreutils. E.g. on an older Mac OS X:

brew install coreutils
du -hs * | gsort -h

From sort manual:

-h, --human-numeric-sort compare human readable numbers (e.g., 2K 1G)

du | sort -nr | cut -f2- | xargs du -hs

There is an immensely useful tool I use called ncdu that is designed for finding those pesky high disk-usage folders and files, and removing them. It's console based, fast and light, and has packages on all the major distributions.

@Douglas Leeder, one more answer: Sort the human-readable output from du -h using another tool. Like Perl!

du -h | perl -e 'sub h{%h=(K=>10,M=>20,G=>30);($n,$u)=shift=~/([0-9.]+)(\D)/;
return $n*2**$h{$u}}print sort{h($b)<=>h($a)}<>;'

Split onto two lines to fit the display. You can use it this way or make it a one-liner, it'll work either way.

Output:

4.5M    .
3.7M    ./colors
372K    ./plugin
128K    ./autoload
100K    ./doc
100K    ./syntax

EDIT: After a few rounds of golf over at PerlMonks, the final result is the following:

perl -e'%h=map{/.\s/;99**(ord$&&7)-$`,$_}`du -h`;die@h{sort%h}'

du -k * | sort -nr | cut -f2 | xargs -d '\n' du -sh

As far as I can see you have three options:

1. Alter du to sort before display.
2. Alter sort to support human sizes for numerical sort.
3. Post process the output from sort to change the basic output to human readable.

You could also do du -k and live with sizes in KiB.

For option 3 you could use the following script:

#!/usr/bin/env python

import sys
import re

sizeRe = re.compile(r"^(\d+)(.*)$")

for line in sys.stdin.readlines():
    mo = sizeRe.match(line)
    if mo:
        size = int(mo.group(1))
        if size < 1024:
            size = str(size)+"K"
        elif size < 1024 ** 2:
            size = str(size/1024)+"M"
        else:
            size = str(size/(1024 ** 2))+"G"

        print "%s%s"%(size,mo.group(2))
    else:
        print line

I've had that problem as well and I'm currently using a workaround:

du -scBM | sort -n

This will not produce scaled values, but always produce the size in megabytes. That's less then perfect, but for me it's better than nothing (or displaying the size in bytes).

Found this posting elsewhere. Therefore, this shell script will do what you want without calling du on everything twice. It uses awk to convert the raw bytes to a human-readable format. Of course, the formatting is slightly different (everything is printed to one decimal place precision).

#/bin/bash
du -B1 | sort -nr  |awk '{sum=$1;
hum[1024**3]="G";hum[1024**2]="M";hum[1024]="K";
for (x=1024**3; x>=1024; x/=1024){
        if (sum>=x) { printf "%.1f%s\t\t",sum/x,hum[x];print $2;break
}}}'

Running this in my .vim directory yields:

4.4M            .
3.6M            ./colors
372.0K          ./plugin
128.0K          ./autoload
100.0K          ./syntax
100.0K          ./doc

(I hope 3.6M of color schemes isn't excessive.)

Here's an example that shows the directories in a more compact summarized form. It handles spaces in directory/filenames.

% du -s * | sort -rn | cut -f2- | xargs -d "\n" du -sh

53G  projects
21G  Desktop
7.2G VirtualBox VMs
3.7G db
3.3G SparkleShare
2.2G Dropbox
272M apps
47M  incoming
14M  bin
5.7M rpmbuild
68K  vimdir.tgz

This version uses awk to create extra columns for sort keys. It only calls du once. The output should look exactly like du.

I've split it into multiple lines, but it can be recombined into a one-liner.

du -h |
  awk '{printf "%s %08.2f\t%s\n", 
    index("KMG", substr($1, length($1))),
    substr($1, 0, length($1)-1), $0}' |
  sort -r | cut -f2,3

Explanation:

• BEGIN - create a string to index to substitute 1, 2, 3 for K, M, G for grouping by units, if there's no unit (the size is less than 1K), then there's no match and a zero is returned (perfect!)
• print the new fields - unit, value (to make the alpha-sort work properly it's zero-padded, fixed-length) and original line
• index the last character of the size field
• pull out the numeric portion of the size
• sort the results, discard the extra columns

Try it without the cut command to see what it's doing.

Here's a version which does the sorting within the AWK script and doesn't need cut:

du -h |
   awk '{idx = sprintf("%s %08.2f %s", 
         index("KMG", substr($1, length($1))),
         substr($1, 0, length($1)-1), $0);
         lines[idx] = $0}
    END {c = asorti(lines, sorted);
         for (i = c; i >= 1; i--)
           print lines[sorted[i]]}'