Sql Union All *with* “distinct”

Question

UNION joins two results and remove duplicates, while UNION ALL does not remove duplicates. UNION also sort the final output.

What I want is the UNION ALL without duplicates and without the sort. Is that possible?

The reason for this is that I want the result of the first query to be on top of the final result, and the second query at the bottom. (And each sorted as if they where run individually)

Answer 1

Consider these tables (Standard SQL code, runs on SQL Server 2008):

WITH A 
     AS 
     (
      SELECT * 
        FROM (
              VALUES (1), 
                     (2), 
                     (3), 
                     (4), 
                     (5), 
                     (6) 
             ) AS T (col)
     ),
     B 
     AS 
     (
      SELECT * 
        FROM (
              VALUES (9), 
                     (8), 
                     (7), 
                     (6), 
                     (5), 
                     (4) 
             ) AS T (col)
     ), ...

The desired effect is this to sort table A by col ascending, sort table B by col descending then unioning the two, removing duplicates, retaining order before the union and leaving table A results on the "top" with table B on the "bottom" e.g. (pesudo code)

(
 SELECT *
   FROM A
  ORDER 
     BY col
)
UNION
(
 SELECT *
   FROM B
  ORDER 
     BY col DESC
);

Of course, this won't work in SQL because there can only be one ORDER BY clause and it can only be applied to the top level table expression (or whatever the output of a SELECT query is known as; I call it the "resultset").

The first thing to address is the intersection between the two tables, in this case the values 4, 5 and 6. How the intersection should be sorted needs to be specified in SQL code, therefore it is desirable that the designer specifies this too! (i.e. the person asking the question, in this case).

The implication in this case would seem to be that the intersection ("duplicates") should be sorted within the results for table A. Therefore, the sorted resultset should look like this:

      VALUES (1), -- A including intersection, ascending
             (2), -- A including intersection, ascending
             (3), -- A including intersection, ascending
             (4), -- A including intersection, ascending
             (5), -- A including intersection, ascending
             (6), -- A including intersection, ascending
             (9), -- B only, descending 
             (8), -- B only, descending  
             (7), -- B only, descending 

Note in SQL "top" and "bottom" has no inferent meaning and a table (other than a resultset) has no inherent ordering. Also (to cut a long story short) consider that UNION removes duplicate rows by implication and must be applied before ORDER BY. The conclusion has to be that each table's sort order must be explicitly defined by exposing a sort order column(s) before being unioned. For this we can use the ROW_NUMBER() windowed function e.g.

     ...
     A_ranked
     AS
     (
      SELECT col, 
             ROW_NUMBER() OVER (ORDER BY col) AS sort_order_1
        FROM A                      -- include the intersection
     ),
     B_ranked
     AS
     (
      SELECT *, 
             ROW_NUMBER() OVER (ORDER BY col DESC) AS sort_order_1
        FROM B
       WHERE NOT EXISTS (           -- exclude the intersection
                         SELECT * 
                           FROM A
                          WHERE A.col = B.col 
                        )
     )
SELECT *, 1 AS sort_order_0 
  FROM A_ranked
UNION
SELECT *, 2 AS sort_order_0 
  FROM B_ranked
ORDER BY sort_order_0, sort_order_1;

Answer 2

select T.Col1, T.Col2, T.Sort
from 
    (
      select T.Col1,
             T.Col2,
             T.Sort,
             rank() over(partition by T.Col1, T.Col2 order by T.Sort) as rn
      from
          (
            select Col1, Col2, 1 as Sort
            from Table1
            union all
            select Col1, Col2, 2
            from Table2
          ) as T
    ) as T
where T.rn = 1    
order by T.Sort

Answer 3

1,1: select 1 from dual union all select 1 from dual 1: select 1 from dual union select 1 from dual