56

I know that the -k option for the Unix sort allow us to sort by a specific column and all of the following. For instance, given the input file:

2 3
2 2
1 2
2 1
1 1

Using sort -n -k 1, I get an output sorted by the 1st column and then by the 2nd:

1 1
1 2
2 1
2 2
2 3

However, I want to keep the 2nd column ordering, like this:

1 2
1 1
2 3
2 2
2 1

Is this possible with the sort command?

3 Answers 3

85

Give this a try:

sort -s -n -k 1,1

The -s disables 'last-resort' sorting, which sorts on everything that wasn't part of a specified key.

The -k 1 doesn't actually mean "this field and all of the following" in the context of numeric sort, as you can see if you try to sort on the second column. You're merely seeing ties broken by going to the rest of the line. In general, however, you need to specify -k 1,1 to sort only on field one.

7
  • You are right. This is exactly what I needed. Thanks!
    – ssn
    Aug 31, 2009 at 16:31
  • is it possible to use join on the output of this sort?
    – N. F.
    Sep 27, 2012 at 20:13
  • @MiNdFrEaK: The requirement of join is that the input be sorted on the fields you're joining on. So sure, this output is sorted on the first field, and you can join on it.
    – Cascabel
    Sep 27, 2012 at 21:49
  • I have 2 files, one having 2 columns, another having 1 column. The second file is sorted using sort -u. Now the task is I need to join this column with the first column of the first file, which is not sorted, so what will be the syntax? will this work? join -j 1 file2.txt sort -s -n -k 1 file1.txt?
    – N. F.
    Sep 28, 2012 at 4:10
  • 1
    The -k 1,1 (the ",1" part) doesn't work any better for me. What works is -s -k 1, with -n if you need it.
    – Totor
    Aug 8, 2013 at 15:12
10

To only sort on the first column you should do:

sort -n -s -k1,1

From Unix and Linux System Administration Handbook

sort accepts the key specification -k3 (rather than -k3,3), but it probably doesn’t do what you expect. Without the terminating field number, the sort key continues to the end of the line

2
  • Not working for me, I need to add the -s option as Cascabel pointed out.
    – Jean Paul
    Jan 7, 2019 at 17:08
  • @JeanPaul you are right, the documentation for -s says "This option maintains the original record order of records that have an equal key."
    – tidbeck
    Jan 9, 2019 at 10:09
2

None of the provided answers work generally for me.

Both sort -s -k 2 file1 and sort -n -k1,1 do additional sorting with this file:

# cat file1
 3 3 5
 3 2 3
 1 4 7
 0 1 2
 3 2 1

I just had to do this exact thing and ended up using a shell loop. This solution might not work well on a very large file because the entire file needs to be read for each unique value in the sorted column.

Here the file is sorted on column 2 only.

# awk '{print $2}' file1 | sort | uniq | while read index
do  
    awk -v var=$index '$2 == var { print $0}' file1 
done
 0 1 2
 3 2 3
 3 2 1
 3 3 5
 1 4 7
2
  • sort -s -k2,2 file1
    – plhn
    Feb 24, 2017 at 10:22
  • The answer proposed by Cascabel is working but I think you missread it.
    – Jean Paul
    Jan 7, 2019 at 17:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.