PSPP-BUG: [bug #18572] Warn if grouping variables are out of order

[Ben Pfaff]

URL:
  <http://savannah.gnu.org/bugs/?18572>

                 Summary: Warn if grouping variables are out of order
                 Project: PSPP
            Submitted by: blp
            Submitted on: Thursday 12/21/06 at 15:22
                Category: Other
                Severity: 1 - Wishlist
                  Status: None
             Assigned to: blp
             Open/Closed: Open
         Discussion Lock: Any
                 Release: None
                  Effort: 0.00

    _______________________________________________________

Details:

It is a fairly common mistake, in my experience, to forget to sort data on
the split variables when SPLIT FILE is in use, or to similarly forget to sort
data on the group variables for RANK, and so on for other forms of grouping. 
At least, when I'm trying to use PSPP for tasks, I sometimes forget.

It would be nice if we could warn about this, so that users don't get too
surprised, but it's somewhat difficult because we don't know the intended
sort order.  The actual sort order could be any permutation of the grouping
variables, and each column could be in ascending or descending order.  For N
grouping variables there are 2**N * N! sort orders, so a brute-force
algorithm is going to be too slow.  Also, we have to do it "online": we only
get one pass through the data.

I have, however, figured out a decent algorithm that is both efficient and
usually figures out when data is not sorted.  At some point, I should
implement this in PSPP.  See
http://groups.google.com/group/comp.programming/msg/c7ebefe24af2f930  for
details.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?18572>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/