What you have above sorts the accumulated input twice each time you read a line from your input file. Clearly that isn't what you want. Even if you were doing the sorts in an END clause instead of in a clause that is executed every time you read an input line, performing a sort on the entire line and then performing a sort on the second sort key is not the same as performing a single sort with a primary and secondary sort key.

I don't have access to a system running gawk, but just using standard interfaces, I get the output:
which I think is what you're trying to get, when I use the script:

Code:

#!/bin/ksh
awk 'BEGIN{FS=OFS="/"
    tmpfile = "asorti.out"
    sortcommand = "sort -t/ -o " tmpfile
    cleanup = "rm " tmpfile
}
{   split ($2, sub, " ")
    array[$1 "/" sub[2] "/" $0] = $0
}
END{for (i in array) print i | sortcommand
    close(sortcommand)
    while(getline i < tmpfile) print array[i]
    close(tmpfile)
    system(cleanup)
}' in

where in contains the data listed in your first posting on this thread. If I read the gawk man page correctly, this should be roughly equivalent to:

Code:

#!/bin/ksh
gawk 'BEGIN{FS = OFS = "/" }
{   split ($2, sub, " ")
    array[$1 "/" sub[2] "/" $0] = $0
}   
END{n = asorti(array)
    for(i = 1; i <= n; print array[i++]);
}' in