[SOLVED] Problems with sort

BarataPT · 03-31-2011, 05:11 AM

Hi,

I have a file with the following contents:

Quote:

20100228062520 10.134.53.113
20100228062520 10.134.53.114
20100228062520 10.134.53.114
20100228062520 99.247.216.111
20100228062521 10.134.53.111
20100228062521 10.134.53.114
20100228062522 201.230.227.197
20100228062524 99.247.216.111
20100228062525 10.134.53.114
20100228062525 99.247.216.110

The columns are separated by a TAB.

I want to sort by the IP address and i do this:

Code:

cat file2.txt | sort -n -k2,2 -t "    " | less

But the output isn't quite correct. It seems that it sorts only by the 1st IP number (before the first point).

Quote:

20100228062520 10.134.53.113
20100228062520 10.134.53.114
20100228062520 10.134.53.114
20100228062521 10.134.53.111
20100228062521 10.134.53.114
20100228062525 10.134.53.114
20100228062520 99.247.216.111
20100228062524 99.247.216.111
20100228062525 99.247.216.110
20100228062522 201.230.227.197

If i cut only the IPs i get the correct result

Code:

cat file2.txt | cut -d "    " -f2,2 | sort -n

Quote:

10.134.53.111
10.134.53.113
10.134.53.114
10.134.53.114
10.134.53.114
10.134.53.114
99.247.216.110
99.247.216.111
99.247.216.111
201.230.227.197

Can someone explain to me why this happen?

jschiwal · 03-31-2011, 06:04 AM

An IP address isn't a number.

You have two types of delimiters on the same line. tabs and dots.

You could use sed or tr to convert the tab to a dot. Sort the results, and convert the dot back again.

Code:

tr '\t' '\.' <test | sort -t'.' -k2 -k3 -k4 -k5 | sed 's/\./\t/'

kurumi · 03-31-2011, 06:13 AM

Code:

$ ruby -ane 'BEGIN{a=[];b=[]} ;b<<[$F[0].to_i,$F[1]];END{b.sort_by{|x| [ x[1].split(".")[0].to_i , x[1].split(".")[-1].to_i] }.each{|x|puts x.join(" ")}}'  file
20100228062521 10.134.53.111
20100228062520 10.134.53.113
20100228062520 10.134.53.114
20100228062525 10.134.53.114
20100228062520 10.134.53.114
20100228062521 10.134.53.114
20100228062525 99.247.216.110
20100228062524 99.247.216.111
20100228062520 99.247.216.111
20100228062522 201.230.227.197

---------- Post added 03-31-11 at 06:13 AM ----------

Quote:

Originally Posted by jschiwal

An IP address isn't a number.

You have two types of delimiters on the same line. tabs and dots.

You could use sed or tr to convert the tab to a dot. Sort the results, and convert the dot back again.

Code:

tr '\t' '\.' <test | sort -t'.' -k2 -k3 -k4 -k5 | sed 's/\./\t/'

this does not give correct output.

grail · 03-31-2011, 08:19 AM

How about using the version sort instead ... not quite what it was intended for, but it seems to work:

Code:

sort -V -k2,2 file

jschiwal · 03-31-2011, 09:05 AM

Correction. I left out the -n option for numeric sort.

Code:

tr '\t' '\.' <test | sort -n -t'.' -k2 -k3 -k4 -k5 | sed 's/\./\t/'
20100228062521  10.134.53.111
20100228062520  10.134.53.113
20100228062520  10.134.53.114
20100228062520  10.134.53.114
20100228062521  10.134.53.114
20100228062525  10.134.53.114
20100228062525  99.247.216.110
20100228062520  99.247.216.111
20100228062524  99.247.216.111
20100228062522  201.230.227.197

Nice catch about using a Version sort. I didn't think about that. The format of an IP address matches version numbers, so it would work.

BarataPT · 03-31-2011, 02:19 PM

Thanks guys