Bash - numerical sort sorts 10s before 1s

Question

I'm in a directory. There are two subdirectories and I want to list the files in them, in numerical order. In the first directory, the files are named 01, 02, 03... and they sort fine. But in the second directory, the files are named 1, 2, 3 and thus the 10 sorts before the 1. Keying does not seem to work as I'm a directory back from the files, so a script such as this:

find "$PWD" -type f|sort -n -t _ -k 1

Does not work. How do I list in true numerical order?

Please show a bit sample data (in) and desired result. Just 3 to 5 lines. — user unknown, Mar 31 '12 at 12:39
Agree with userunknown. This works perfectly fine: `echo -e '10\n1\n11\n2' | sort -k 1 -n` — phemmer, Mar 31 '12 at 15:00
The key field you specified probably is not what you think it is. Analyze the output of find "$PWD" -type f and see what you end up with is really the begining of a numeric key. — pizza, Mar 31 '12 at 20:19

score 2 · Accepted Answer · answered Mar 31 '12 at 14:18

sort manual : Sort numerically: the number begins each line ... I doubt that the output from find is going to be as such. Therefore you should write a script that strips leading characters from the filenames and then call sort.

For example find "$WORD" -type f | cut -d / -f 2- will work on sorting the filenames in first level directory numerically , but the rest will not be sorted as numbers.

You have to write the script according your needs, but if what you are looking for is in a single directory, the above line will do.

score 0 · Answer 2 · answered Mar 31 '12 at 13:01

You pose an interesting challenge. Try this Perl script:

#!/usr/bin/perl
use warnings;
use strict;
use integer;

# This script sorts filenames while respecting their ordinal component,
# outputting 'dir/stem9.ext' before 'dir/stem10.ext'.  As input, it
# expects one filename per line on stdin.

our $n_digits = 6;

sub compare_filenames_ordinally ($$) {
    my ($a, $b) = @_;
    my $pattern = qr|^(.*?)(\d*)((?:\.[^/\d]*)?)$|;
    my ($stem_a, $ordinal_a, $extension_a) = $a =~ /$pattern/;
    my ($stem_b, $ordinal_b, $extension_b) = $b =~ /$pattern/;
    length($ordinal_a) <= $n_digits && length($ordinal_b) <= $n_digits
      or die "$0: the ordinal part of the filename "
      . "is longer than $n_digits digits\n";
    my $ordinal_a1 = length($ordinal_a)
      ? sprintf "%0${n_digits}d", $ordinal_a : '';
    my $ordinal_b1 = length($ordinal_b)
      ? sprintf "%0${n_digits}d", $ordinal_b : '';
    my $a1 = "${stem_a}${ordinal_a1}${extension_a}";
    my $b1 = "${stem_b}${ordinal_b1}${extension_b}";
    return $a1 cmp $b1;
}

my @file = <>;
chomp for @file;
@file = sort { compare_filenames_ordinally $a, $b } @file;
print "$_\n" for @file;

The script undoubtedly could be improved, not least because it repeats some lines of code with respect to $a and $b, but its idea is that it extends the likes of dir/stem10.txt to dir/stem000010.txt before comparing it against other filenames.

potong · Answer 3 · 2012-03-31T16:20:16.477

This might work for you:

mkdir test
cd test
mkdir t1 t2
cd t1
seq -f 'test_%02g.txt 20 | xargs touch
cd ../t2
touch test_{1..20}.txt
cd ..
find . -type f | sort -t_ -k1,1 -k2,2g
./t1/test_01.txt
./t1/test_02.txt
./t1/test_03.txt
./t1/test_04.txt
./t1/test_05.txt
./t1/test_06.txt
./t1/test_07.txt
./t1/test_08.txt
./t1/test_09.txt
./t1/test_10.txt
./t1/test_11.txt
./t1/test_12.txt
./t1/test_13.txt
./t1/test_14.txt
./t1/test_15.txt
./t1/test_16.txt
./t1/test_17.txt
./t1/test_19.txt
./t1/test_20.txt
./t2/test_1.txt
./t2/test_2.txt
./t2/test_3.txt
./t2/test_4.txt
./t2/test_5.txt
./t2/test_6.txt
./t2/test_7.txt
./t2/test_8.txt
./t2/test_9.txt
./t2/test_10.txt
./t2/test_11.txt
./t2/test_12.txt
./t2/test_13.txt
./t2/test_14.txt
./t2/test_15.txt
./t2/test_16.txt
./t2/test_17.txt
./t2/test_18.txt
./t2/test_19.txt
./t2/test_20.txt

or this:

find . -type f | 
sed 'h;s/.*\//& /;H;s/.* //;s/[^0-9]//g;x;G;s/\n/ /g' |
sort -k2,2 -k4,4g |
cut -d' ' -f1
./t1/test_01.txt
./t1/test_02.txt
./t1/test_03.txt
./t1/test_04.txt
./t1/test_05.txt
./t1/test_06.txt
./t1/test_07.txt
./t1/test_08.txt
./t1/test_09.txt
./t1/test_10.txt
./t1/test_11.txt
./t1/test_12.txt
./t1/test_13.txt
./t1/test_14.txt
./t1/test_15.txt
./t1/test_16.txt
./t1/test_17.txt
./t1/test_19.txt
./t1/test_20.txt
./t2/test_1.txt
./t2/test_2.txt
./t2/test_3.txt
./t2/test_4.txt
./t2/test_5.txt
./t2/test_6.txt
./t2/test_7.txt
./t2/test_8.txt
./t2/test_9.txt
./t2/test_10.txt
./t2/test_11.txt
./t2/test_12.txt
./t2/test_13.txt
./t2/test_14.txt
./t2/test_15.txt
./t2/test_16.txt
./t2/test_17.txt
./t2/test_18.txt
./t2/test_19.txt
./t2/test_20.txt

Bash - numerical sort sorts 10s before 1s

3 Answers3