du question: sort by size & show human-readable sizes

[svk]

Hello,

I'm trying to print out a listing of all the files (and subdirectories) within a directory along with their sizes. I would like the output to be sorted by size (greatest to least), and in human-readable format.

-----------------

EDIT on Jun 20, 2011: As posted by meithan, with GNU coreutils >= 7.5, which is available with Maverick and later, it's possible to do:

du -hs | sort -h

Read on for the old discussion.

-----------------

Without that last restriction, I am able to generate a listing by running this command:

serg@bijou:~/Pictures$ du -s ./* | sort -nr
2172    ./Backgrounds
1540    ./100_1065.jpg
1448    ./100_1024.jpg
1192    ./100_1033.jpg
1136    ./100_1028.jpg
1128    ./100_1068.jpg
1088    ./100_1022.jpg
1032    ./100_1027.jpg
1016    ./100_1011.jpg
1008    ./100_1025.jpg
952    ./100_1021.jpg
944    ./100_1061.jpg
104    ./conference.jpg

I can get the output in human-readable format by adding the -h switch to the du command; however, then the sort -nr command does not work properly (for example, it sorts 4.0K before 2.2M, because 4.0 is a bigger number than 2.2 -- it ignores the suffix).

Here's what I want the output to look like:

 2.2M    ./Backgrounds
1.6M    ./100_1065.jpg
 1.5M    ./100_1024.jpg
 1.2M    ./100_1033.jpg
 1.2M    ./100_1028.jpg
 1.2M    ./100_1068.jpg
 1.1M    ./100_1022.jpg
 1.1M    ./100_1027.jpg
 1016K    ./100_1011.jpg
 1008K    ./100_1025.jpg
 952K    ./100_1021.jpg
 944K    ./100_1061.jpg
 104K    ./conference.jpg

Unfortunately, this one had to be hand crafted -- I don't know how I could rewrite the command to get it to show this.

Can sed or awk be used to accomplish this? Or would I have to write a script? I'm only marginally familiar with any of these tools, so I would like some help.

Thank you in advance!

[ghostdog74]

you can write you own routine, or if you insist on using du

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

[svk]

Thank you very much, very clever! I understood each individual sub-command, but I wouldn't have thought of putting them together like this.

The only change I made to make it do exactly what I wanted:
du -ks * | sort -nr | cut -f2 | xargs -d '\n' du -sh
Because I didn't want it to descend into subdirectories.

Thank you very, very much!

[jsaunders]

Thanks for this! Works great. A slight twist if you will from my own personal usage (aliases stem from my slackware days, old habits die hard):

alias v='ls -lh'
alias vv="du --max-depth=1 -k | sort -nr | cut -f2 | xargs -d '\n' du -sh"

Cheers,

jsaunders

[lavinog]

Thanks for this! Works great. A slight twist if you will from my own personal usage (aliases stem from my slackware days, old habits die hard):

alias v='ls -lh'
alias vv="du --max-depth=1 -k | sort -nr | cut -f2 | xargs -d '\n' du -sh"

Cheers,

jsaunders

using vv|head gives a terminated by signal 13 message.
This seems to take care of that:

alias vv="du --max-depth=1 -k | sort -nr | cut -f2 | xargs -d '\n' du -sh 2>/dev/null"

I wrote a script before finding this thread to do a similar thing:

#!/bin/bash

# du-sort : Provides a sorted du in human readable format

# MAXDEPTH :  set to zero to show all
MAXDIR=1

# SORT ORDER  (true/false)
ASCENDING=true

# DEFAULT ARGUMENTS
ARGS='-xh'

# mktmp file
DATA=$(mktemp)

# ADD MAX-DEPTH ARG
if [ "$MAXDIR" -gt "0" ]
	then
	ARGS="$ARGS --max-depth=$MAXDIR"
fi

# CHECK FOR EXTRA ARGUMENT
if ! [ -z $1 ]
then
	ARGS="$ARGS $1"
fi

# RUN DU
du $ARGS >$DATA 2>/dev/null

# SORT DATA AND OUTPUT
if $ASCENDING
then
	for P in K M G
	do
		grep -e "[0-9]$P" $DATA|sort -n
	done

else
	for P in G M K
	do
		grep -e "[0-9]$P" $DATA|sort -nr
	done
fi

#remove datafile
rm $DATA

[fusiondog77]

These all work but either require running du twice or using a temp file. Here is a way to do it with neither using perl and a Schwartzian Transform http://en.wikipedia.org/wiki/Schwartzian_transform

du -shx | perl -e '%byte_order = ( G => 0, M => 1, K => 2 ); print map { $_->[0] } sort { $byte_order{$a->[1]} <=> $byte_order{$b->[1]} || $b->[2] <=> $a->[2] } map { [ $_, /([MGK])/, /(\d+)/ ] } <>'

[jackashe]

fusiondog77 I was certainly looking for something that doesn't run du twice because I am dealing with > 5 minute times to run du. Wanted to comment to say thanks! Also, I use du --si, since the newest ubuntu versions use the si units. in this case, kilobytes come from du as lowercase k, not uppercase K, so I made the following version of the perl:

perl -e '%byte_order = ( G => 0, M => 1, K => 2, k => 2 ); print map { $_->[0] } sort { $byte_order{$a->[1]} <=> $byte_order{$b->[1]} || $b->[2] <=> $a->[2] } map { [ $_, /([MGKk])/, /(\d+)/ ] } <>'

Another problem I had was zero size files don't have any suffix, and they got put between G and M. Didn't try to fix this yet though.

[lavinog]

fusiondog77 I was certainly looking for something that doesn't run du twice because I am dealing with > 5 minute times to run du.

That shouldn't be an issue because the kernel will buffer the file stats so subsequent du commands shouldn't take very long.

[jackashe]

oh it just dawned on my what to do in the final case! When du is saying that the size is 0, there is no suffix, so perl is just using 0 as the byte order character. So I fixed this by changing that part to '%byte_order = ( G => -3, M => -2, K => -1, k => -1);

[jackashe]

That shouldn't be an issue because the kernel will buffer the file stats so subsequent du commands shouldn't take very long.

Even for Network attached storage? I would imagine that by the time the first du finishes, enough time has passed that the first files logically checked might have changed and they would need to be rechecked. In any case, seems the safest to do the du once -- otherwise you could technically get the result out of order since you are sorting the 2nd du based on the 1st. not guaranteed to be the same.