Shell Programming and Scripting

Thanks that is very useful to know, that this cannot be relied upon to work on any system. So it is appears to be an extension and not standard to perform a stable sort. I guess a possible indicator might then be if a particular sort supports a "stable sort" option in the first place. For example if this option is called "-s" then if this works:

Code:

sort -st, -k1,3 infile | awk -F, '{n=$1 FS $2 FS $3}p!=n;{p=n}'

then probably this works too:

Code:

sort -ut, -k1,3 infile

I checked some man pages and systems that provide a stable "-u", it says:

Code:

-u      with -c, check for strict ordering; without -c, output only the first of an equal run

instead of something like this

Code:

-u      [..] If used with the -c option, check that there are no lines with duplicate keys, in addition to checking that the input file is sorted.

..

Just tested it on a Solaris 10 box with same results...looks like the legacy unix offerings dont have a stable sort whereas linux has it.
None of the legacy unix flavors HPUX Solaris or AIX have the "-s" switch so looks like linux may have optimized it to spit out the first record of a group.
That must only be on linux system...
Yes that is pretty much the case on all of AIX Solaris and HPUX...so looks like sort will break if the code is to be deployed on different platforms...unless it is different flavors of linux.

Not just Linux systems, but on systems that use GNU sort, so also freeBSD and OSX and any system that has GNU sort installed. But it does not work with POSIX sort, so in conclusion it is not a solution for UNIX systems in general... Thanks again Shamrock, for your effort...

So then I guess a portable solution would be to force a stable sort, for example like this:

Code:

nl -s, infile | sort -t, -k2,4 -k1 | cut -f2- -d, | awk -F, '{n=$1 F2 $2 FS $3}p!=n;{p=n}'