“If a query has an ‘order by’ or ‘group by’ clause that includes columns not in the select list, Adaptive Server adds those columns as hidden columns in the columns being processed. The columns listed in the ‘order by’ or ‘group by’ clause are included in the test for distinct rows. To comply with ANSI standards, include the ‘order by’ or ‘group by’ column in the select list.”

Interestingly, MySQL 4.1.11 handles this “correctly” as I would have expected:

mysql> SELECT DISTINCT x FROM test ORDER BY y DESC;
+------+
| x    |
+------+
|    3 |
|    1 |
|    2 |
+------+
3 rows in set (0.00 sec)

havard=# SELECT distinct a from test order by b desc;
ERROR:  for SELECT DISTINCT, ORDER BY expressions must appear in select list
jhavard=# SELECT distinct on (a) a from test order by b desc;
ERROR:  SELECT DISTINCT ON expressions must match initial ORDER BY expressions

Take a look at the query optimizer and query planner output if possible. I’m willing to bet the query is executed as if it were SELECT DISTINCT ON (b) a FROM test ORDER BY b. After all, you are supplying an ordered field.

ORDER BY items must appear in the select list if SELECT DISTINCT is specified.

Clearly, there is something with this type of mix that Sybase didn’t handle properly if Microsoft disabled ordering by columns that did not appear in a distinct resultset.