bash du with byte size in decimals to the nearest thousandth

Is it possible in BASH to do a "du" command with the byte size shown in decimals.

For example, say I have the following files (numbers are in bytes):

12345 file1
2345 file2
6491 file3

I would like to do a "du" command in linux that would output the following instead:

12.3 file1
2.3 file2
6.5 file3

Note: The "-h" f开发者_JAVA技巧lag does not work.

You should be able to use awk for this, something like:

du -b * | awk '{printf "%10.1f %s\n", $1/1000, $2}'

As in the following transcript:

pax> ls -l
total 3092
-rwxrwxrwx  1 pax   pax       807 2008-09-14 08:26 combo.pl*
-rwxrwxrwx  1 pax   pax       236 2008-09-14 08:26 match.pl*
-rwxrwxrwx  1 pax   pax       754 2008-09-14 08:26 mkdb.pl*
-rwxrwxrwx  1 pax   pax       689 2008-09-14 08:26 nine.pl*
-rwxrwxrwx  1 pax   pax   2089522 2008-05-25 21:06 words.db*
-rwxrwxrwx  1 pax   pax   1044761 2008-05-25 21:06 words.txt*

pax> du -b *
807       combo.pl
236       match.pl
754       mkdb.pl
689       nine.pl
2089522   words.db
1044761   words.txt

pax> du -b * | awk '{printf "%10.1f %s\n", $1/1000, $2}'
       0.8 combo.pl
       0.2 match.pl
       0.8 mkdb.pl
       0.7 nine.pl
    2089.5 words.db
    1044.8 words.txt

You could use the du -k flag to report in Kilobytes?

On Linux, an alternative would be to not use du, but instead use find to report sizes in bytes, and then do the math with perl, awk, whatever.

du reports disk usage, so it includes sizes of directory entries. If you just want the sizes of actual files, find will give you that.

find . -type f -printf "%s %p\n" | perl -pe 's|^(\d+)(.*)|sprintf("%10.1f", $1/1000).$2|e;'

If you only want the total of the file sizes:

find . -type f -printf "%s\n" | perl -ne '$s+=$_; END{printf "%-10.1f\n", $s/1000}'

No, not with du alone. You need to use the -b parameter to get the finest precision, and then do the math.

You could try this:

du --apparent --block-size=100 | sed -r 's/^([0-9]*)([0-9])/\1.\2/'

This tells du to report sizes in 100-byte blocks, then I add a decimal place before the last digit using sed.

In GNU du if -k won't help (block size = 1024 bytes), you can use --si which is similar to -h (1024), but reports to the nearest thousandth (1000), e.g.

du -k --si some/folder
8.2k    foo
4.1k    bar

Alternatively calculate manually per each line:

du -b * | awk '{"echo "$1"/1000 | bc" | getline size; print size" "$2}'

To customize scale for bc, prefix with scale=1;, e.g.

$ bc <<<"scale=2;1234/1000"
1.23