How to sort strings that contain a common prefix and suffix numerically from Bash?

Question

Here is a list of files:

some.string_100_with_numbers.in-it.txt
some.string_101_with_numbers.in-it.txt
some.string_102_with_numbers.in-it.txt
some.string_23_with_numbers.in-it.txt
some.string_24_with_numbers.in-it.txt
some.string_25_with_numbers.in-it.txt

Now I would like to sort it numerically. Starting with *23* and ending with *102*.

I have tried -n and -g. -t does not help in these messy strings.

Can I ignore leading strings to the number with an option or do I have to be clever and script?

Solution 1

Use ls -lv

From the man page:

-v     natural sort of (version) numbers within text

Solution 2

Try the following:

sort -t '_' -k 2n

• -t '_' (sets the delimiter to the underscore character)
• -k 2n (sorts by the second column using numeric ordering)

DEMO.

Solution 3

If available, simply use sort -V. This is a sort for version numbers, but works well as a "natural sort" option.

$ ff=$( echo some.string_{100,101,102,23,24,25}_with_numbers.in-it.txt )

Without sort:

$ for f in $ff ; do echo $f ; done
some.string_100_with_numbers.in-it.txt
some.string_101_with_numbers.in-it.txt
some.string_102_with_numbers.in-it.txt
some.string_23_with_numbers.in-it.txt
some.string_24_with_numbers.in-it.txt
some.string_25_with_numbers.in-it.txt

With sort -V:

$ for f in $ff ; do echo $f ; done | sort -V
some.string_23_with_numbers.in-it.txt
some.string_24_with_numbers.in-it.txt
some.string_25_with_numbers.in-it.txt
some.string_100_with_numbers.in-it.txt
some.string_101_with_numbers.in-it.txt
some.string_102_with_numbers.in-it.txt

Solution 4

In the general case, try the Schwartzian transform.

Briefly, break out the number into its own field, sort on that, and discard the added field.

# In many shells, use ctrl-v tab to insert a literal tab after the first \2
sed 's/^\([^0-9]*\)\([0-9][0-9]*\)/\2   \1\2/' file |
sort -n |
cut -f2-

This works nicely if the input doesn't have an obvious separator, like for the following input.

abc1
abc10
abc2

where you would like the sort to move the last line up right after the first.

View more solutions

Comments

[Jonathan Leffler]

Note that the -v option to ls is a GNU extension; since you're on Linux, that's OK, but it is not portable to other systems necessarily.

[AWE]

Exactly! it was the k option that I had forgotten and was "hidden" in the man pages.

[becko]

is there an analogous command with find instead of ls (or perhaps piping to sort)?

[interestinglythere]

On Mac, can you can brew install coreutils and use gls.

[tripleee]

Don't use ls in scripts, though.

[tripleee]

@becko find always traverses the tree in a deterministic order and outputs the files in the order it finds them.

[Muhamed Huseinbašić]

@becko sort has -V option which is "natural sort of (version) numbers within text"