Shell Programming and Scripting

Not tested with more data, but something like:

Code:

sort -t';' -r -k5.12,5.15 -k5.8,5.10M -k5.5,5.7 -k5.17,5

---------- Post updated 09-15-09 at 06:03 PM ---------- Previous update was 09-14-09 at 10:12 PM ----------

To sort properly, time has to be in 24 hour format. Otherwise it won't work because 12:59pm is one minute earlier than 1:00pm (similar problem for midnight).

Did a little more testing, the following:

Code:

sort -t';' -r -k5.12,5.15 -k5.8,5.10rM -k5.5,5.6 -k5.17,5.24

seemed to work as verified by this test:

Code:

gawk '
BEGIN {
  srand()
  for (i = 0; i < 1e4; ++i) {
    x = 1234567890 + int(rand()*1e5)
    printf("%010d;;;;%s\n", x, strftime("%a %d %b %Y %T %p MYT", x))
  }
}' |sort -t';' -r -k5.12,5.15 -k5.8,5.10rM -k5.5,5.6 -k5.17,5.24 |sort -cr

Interesting that "-M" option to sort on months. I learned something today. And of course, the OP - if (s)he is still interested - should prefer the sort command as it is the right tool for the job. However, I was expecting sort to be much faster than my awk snippet. So I ran a comparative test on a sample file of 10000 records (nice file generation snippet, binlib) and the results are:

time sort -t';' -r -k5.12,5.15 -k5.8,5.10rM -k5.5,5.6 -k5.17,5.24 /tmp/time > /dev/null
real 0m1.434s
user 0m1.404s
sys 0m0.028s

time awk -f sortdate.awk /tmp/time > /dev/null
real 0m1.769s
user 0m1.732s
sys 0m0.036s

sort is faster but with a narrower margin that I would have expected.