Using sort with numeric columns separated by letters & underscores

Question

I'm trying to sort phrases such as the following:

a12_b7
a12_b11
a5_b3
a5_b30
a12_b10

using the numbers following the letters, lexicographically. For the example above, I expect the result to be:

a5_b3
a5_b10
a12_b7
a12_b11
a12_b30

Reading man sort, I thought I had this figured out:

Fields are delimited by underscores
Two search keys
First search key: first field, beginning with 2nd character in the field
Second search key: second field, beginning with 2nd character in the field.

But - that does not work like I thought it would:

$ cat | sort --debug --key=1.2 --key=2.2 --field-separator=_
sort: text ordering performed using ‘en_IL.UTF-8’ sorting rules
a12_b10
 ______
     __
_______
a12_b11
 ______
     __
_______
a12_b7
 _____
     _
______
a5_b3
 ____
    _
_____
a5_b30
 _____
    __
______

What have I gotten wrong? And what would be the appropriate sort command-line in this case?

Look into the -n option. I don't think "lexicographically" means what you think it means. — Scott Hunter
– Scott Hunter, Commented Aug 7 at 15:58

Ed Morton · Accepted Answer · 2025-11-20 12:50:56Z

Looks like you want to sort numerically when the default sort order is alphabetically. You could do:

$ sort -nt'_' -k1.2 -k2.2 file
a5_b3
a5_b30
a12_b7
a12_b10
a12_b11

but if the input was any more complicated than that (e.g. not always single letter chars before each sort key) then I'd use the Decorate-Sort-Undecorate idiom, e.g.:

$ cat file
phc12_bob7
efg12_bk11
cfad5_xxxx3
df5_chekb30
a12_tg10

$ sed -E 's/([^0-9]+)(.*)(_[^0-9]+)(.*)/\1\t\2\t\3\t\4/' file
phc     12      _bob    7
efg     12      _bk     11
cfad    5       _xxxx   3
df      5       _chekb  30
a       12      _tg     10

$ sed -E 's/([^0-9]+)(.*)(_[^0-9]+)(.*)/\1\t\2\t\3\t\4/' file | sort -k2,2n -k4,4n
cfad    5       _xxxx   3
df      5       _chekb  30
phc     12      _bob    7
a       12      _tg     10
efg     12      _bk     11

$ sed -E 's/([^0-9]+)(.*)(_[^0-9]+)(.*)/\1\t\2\t\3\t\4/' file | sort -k2,2n -k4,4n | tr -d '\t'
cfad5_xxxx3
df5_chekb30
phc12_bob7
a12_tg10
efg12_bk11

The sed modifies (Decorates) the input so the sort command can Sort it, then the tr removes the extra chars that the sed added (Undecorates).

Collectives™ on Stack Overflow

Using sort with numeric columns separated by letters & underscores

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related