restriced key sort in unix

Question

Hi , I have a file file1.txt with data:

2010/09/04,21:53:42.048,a
2010/09/04,21:53:40.923,b
2010/09/04,21:53:40.923,a
2010/09/04,21:50:42.048,a.

I want to sort the file based on time stamp. I am currently using

 sort  -t% -k1.1,1.23 file1.txt > file2.txt 

My expected output is

2010/09/04,21:50:42.048,a
2010/09/04,21:53:40.923,b
2010/09/04,21:53:40.923,a
2010/09/04,21:53:42.048,a

But, I am getting the following output

2010/09/04,21:50:42.048,a
2010/09/04,21:53:40.923,a
2010/09/04,21:53:40.923,b
2010/09/04,21:53:42.048,a

I am using SFU3.5 for windows. My sort usage is

usage: sort [-o output] [-cmubdfinr] [-t char] [-T char] [-k keydef] ... [files]

Please provide possible soultion.

Answer 1

Looks like a stability problem, so try to use the --stable option:

-s, --stable
      stabilize sort by disabling last-resort comparison

Update:
It seems that SFU does not support the -s option and is unstable all the time. You can try to use an alternative sort, like the one shipped with unxutils.

Answer 2

If you can't get a sort with a "stable" option then you can make an extra key with "cat -n", or awk, use that in the sort when you need the original order, and strip it off later. Why not tell sort that comma is the field separator too, then you don't need to count 23 characters.

awk '{print NR "," $0}' dat | sort -t, -k2,2 -k1n,1 | sed 's/^[^,]*,//'

A stable sort is one that preserves the original order if the sort keys are equal.