Question about a query

Hi all... I'm trying to make some sense out of how Informix does grouping. It looks like there are times that I phrase the same query 2 different ways and get 2 different results. Here's an example: Given this table (named "dp"): ***** Column name Type Nulls a integer yes b integer yes c integer yes ***** with this data: ***** insert into dp values (1,2,1); insert into dp values (1,2,2); insert into dp values (2,1,1);***** these queries: ***** select a,b,sum(a),(a-b) from dp group by a,b; select a,b,sum(a),(a-b) from dp group by 1,2; select a,b,sum(a),(a-b) from dp group by 1,2,4;***** return these results: ***** a b (sum) (expression) 1 1 2 0 2 1 2 1 a b (sum) (expression) 1 2 2 0 2 1 2 1 a b (sum) (expression) 1 2 2 -1 2 1 2 1 2 row(s) retrieved. ***** All the above stuff was cut from an isql session. There are several things about these results that don't make sense to me: 1) Why do queries 1 & 2 give different results? I thought that using the position numbers was just a shorthand way of referring to the columns. 2) Where does the first row in the output of query #1 come from? There are no rows in the table with a=1 and b=1! 3) Why is the math in #2 inconsistent with the rest of the data? (1-2!=0) 4) The third query is the only one that gives the results I expected. Why does grouping by a computed column (which is changing with every group of a & b in this case) change the results? FWIW, tbstat -V reports 5.02.UC6; isql -V says 4.11.UC1. The results are the same with dbaccess and 4gl. Thanks in advance for any answers/thoughts/hints... -- --- - Derrek (dirk@xmission.com)