Multiple column alphabetical and numerical sort

Question

I want to sort

1. based on the file name.
2. For the case where the file name prefix matches and the file ends in a number, I would like to have those numerically sorted based on the number at the end of the file name.

The following

cat /tmp/foo.txt | sort -t/ -k3,3 -k3,3n

accomplishes 1, but not 2.

Input /tmp/foo.txt

dirA/catA/apple.txt
dirA/catA/addition.txt
dirA/catA/difference
dirA/catB/binary.txt
dirA/catB/carry.txt
dirA/catB/digit
dirA/catC/test-10.txt
dirA/catC/test-100.txt
dirA/catC/test-1000.txt
dirA/catC/test-11.txt
dirA/catC/test-2.txt
dirA/catC/test-20.txt
dirA/catC/test-25.txt
dirA/catC/test-5.txt
dirA/catC/test-50.txt
dirA/catC/test-500.txt
dirA/catC/test-7.txt
dirA/catC/test-75.txt
dirA/catC/test-8.txt
dirA/catC/abc-test-9.txt
dirA/catC/abc-test-999.txt
dirA/catC/abc-test-75.txt
dirA/catC/abc-test-8.txt

Desired Output

dirA/catC/abc-test-8.txt
dirA/catC/abc-test-9.txt
dirA/catC/abc-test-75.txt
dirA/catC/abc-test-999.txt
dirA/catA/addition.txt
dirA/catA/apple.txt
dirA/catB/binary.txt
dirA/catB/carry.txt
dirA/catA/difference
dirA/catB/digit
dirA/catC/test-2.txt
dirA/catC/test-5.txt
dirA/catC/test-7.txt
dirA/catC/test-8.txt
dirA/catC/test-10.txt
dirA/catC/test-11.txt
dirA/catC/test-20.txt
dirA/catC/test-25.txt
dirA/catC/test-50.txt
dirA/catC/test-75.txt
dirA/catC/test-100.txt
dirA/catC/test-500.txt
dirA/catC/test-1000.txt

Answer 1

Perl to the rescue!

perl -e '
    print for sort { (($a =~ m{.*/([^0-9]*)})[0] cmp ($b =~ m{.*/([^0-9]*)})[0])
                     ||
                     (($a =~ /-([0-9]+)/)[0] <=> ($b =~ /-([0-9]+)/)[0]) } <>
' -- /tmp/foo.txt

• <> reads the lines of input
• sort sorts the list based on the given code
• m{.*/([^0-9]*)} extracts the basename up to a digit (if present)
• cmp does the string comparison
• If they are equal, the || "or" applies the second comparison, where:
• /-([0-9]+)/ extracts the number
• <=> does the numeric comparison
• The (...)[0] construct is needed as matching returns a list of matches (corresponding to $1, $2, etc). List context is needed to get the matches. We're interested in the first match only (as there's no other).

Answer 2

awk '
    BEGIN {FS = "[-/.]"; OFS = "\t"}
    {n = 0}
    $(NF-1) ~ /^[0-9]+$/ {n = $(NF-1)}
    {print $3, n, $0}
' foo.txt \
| sort -k1,1 -k2,2n \
| cut -f3-

This is a Schwarzian transform:

• the awk program puts the first word of the filename and the file's number as columns before the file's path
• the data is sorted, first by name then by number
• then the new columns are removed.

Outputs

dirA/catC/abc-test-8.txt
dirA/catC/abc-test-9.txt
dirA/catC/abc-test-75.txt
dirA/catC/abc-test-999.txt
dirA/catA/addition.txt
dirA/catA/apple.txt
dirA/catB/binary.txt
dirA/catB/carry.txt
dirA/catA/difference
dirA/catB/digit
dirA/catC/test-2.txt
dirA/catC/test-5.txt
dirA/catC/test-7.txt
dirA/catC/test-8.txt
dirA/catC/test-10.txt
dirA/catC/test-11.txt
dirA/catC/test-20.txt
dirA/catC/test-25.txt
dirA/catC/test-50.txt
dirA/catC/test-75.txt
dirA/catC/test-100.txt
dirA/catC/test-500.txt
dirA/catC/test-1000.txt

---

Same process with a perl one-liner (except that you read perl statements "bottom-up")

perl -e '
  print join "",
        map  { $_->[2] }
        sort { $a->[0] cmp $b->[0] || $a->[1] <=> $b->[1] }
        map  { [m{.*/(\D+)(\d*)}, $_] }
        <>;
' foo.txt

Answer 3

With sed:

cat /tmp/foo.txt | sed "s/[[:alnum:]-]*\/[[:alnum:]-]*\/\([[:alpha:]-]*\)\([[:digit:]]*\).*/\0|\1|\2 /"|sort -t"|" -k2,2 -k3n|sed "s/\([^|]*\).*/\1/"

The trick is to temporarily put the desired fields to the end of the lines.

Oops: it's better:

cat source | sed "s/[^/]*\/[^/]*\/\([^[:digit:]]*\)\([[:digit:]]*\).*/\0|\1|\2 /"|sort -t"|" -k2,2 -k3n|sed "s/\([^|]*\).*/\1/"

I changed a little the original question. Sort order the last name, without numbers.

dirA/catC/abc-test-8.txt
dirA/catC/abc-test-9.txt
dirA/catC/abc-test-75.txt
dirA/catC/abc-test-999.txt
dirA/catA/addition.txt
dirA/catA/apple.txt
dirA/catB/binary.txt
dirA/catB/carry.txt
dirA/catA/difference
dirA/catB/digit
dirA/catC/test-2.txt
dirA/catC/test-5.txt
dirA/catC/test-7.txt
dirA/catC/test-8.txt
dirA/catC/test-10.txt
dirA/catC/subdir/test-11.txt
dirA/catC/test-11.txt
dirA/cat C/subdir/test-12.txt
dirA/catC/test-20.txt
dirA/catC/test-25.txt
dirA/catC/test-50.txt
dirA/catC/test-75.txt
dirA/catC/test-100.txt
dirA/catC/test-500.txt
dirA/catC/test-1000.txt

cat /tmp/foo.txt | sed "s/\([^/]*\/\)\+\([^[:digit:]]*\)\([[:digit:]]*\)\(.*\)/\0|\2\4|\3 /"|sort -t"|" -k2,2 -k3n|sed "s/\([^|]*\).*/\1/"

Output:

dirA/catC/abc-test-8.txt
dirA/catC/abc-test-9.txt
dirA/catC/abc-test-75.txt
dirA/catC/abc-test-999.txt
dirA/catA/addition.txt
dirA/catA/apple.txt
dirA/catB/binary.txt
dirA/catB/carry.txt
dirA/catA/difference
dirA/catB/digit
dirA/catC/test-2.txt
dirA/catC/test-5.txt
dirA/catC/test-7.txt
dirA/catC/test-8.txt
dirA/catC/test-10.txt
dirA/catC/subdir/test-11.txt
dirA/catC/test-11.txt
dirA/cat C/subdir/test-12.txt
dirA/catC/test-20.txt
dirA/catC/test-25.txt
dirA/catC/test-50.txt
dirA/catC/test-75.txt
dirA/catC/test-100.txt
dirA/catC/test-500.txt
dirA/catC/test-1000.txt

Explain: \([^/]*\/\)\+ Cut the whole path. =>\1

\([^[:digit:]]*\) file name part without digits =>\2

\([[:digit:]]*\) digits =>3 \(.*\) extension =>4

\0|\2\4|\3 Print the whole line | file name 1st part and the extension | digits

sort -t"|" -k2,2 -k3n|sed "s/\([^|]*\).*/\1/ sort, and cut the unnecessary parts.

Instead of last sed cut -d "|" -f1 also works