SQL：根据条件选择唯一值

Question

I have a table containing 5 columns. The first column contains an ID, two columns contain parameters for those IDs with the values 0 or 1, a third column contains a parameter which I need as output, the last column contains a date. The same ID can appear in several rows with different parameters:

ID        parameter1      parameter2        parameter3       date

001       0               1                 A                01.01.2010
001       0               1                 B                02.01.2010
001       1               0                 C                01.01.2010
001       1               1                 D                01.01.2010
002       0               1                 A                01.01.2010

For each unique ID I want to return the value in parameter3 , the decision from which row to return this value is based on the values in parameter1 and parameter2 and the date:

• If there is a row with both parameters being 0 , I want the value in this row.
• If there is no such row, I want the value from the row where parameter1 is 0 and parameter2 is 1 ,
• If there is no such row, I want the row where parameter1 is 1 and parameter2 is 0 .
• Finally, if there is no such row, I want the value from the row with both parameters being 1 .

If there is more than one row matching the required conditions, I want the row with the most recent date.

eg, for the table above, for the ID 001 I would want the second row with the value B in parameter3.

What would be the most effective / fastest way to accomplish this? I tried two approaches so far:

the first one was to select all distinct IDs and then loop through the distinct IDs, using a select statement with the ID in the where clause and then loop through all the rows matching the ID while storing the necessary values in variables.:

foreach
    select distinct ID into i_ID from table1
        foreach
            let o_case = 5
            select case
                when parameter1 = 0 and parameter2 = 0 then 1
                when parameter1 = 0 and parameter2 = 1 then 2
                when parameter1 = 1 and parameter2 = 0 then 3
                when parameter1 = 1 and parameter2 = 1 then 4
                end, parameter3, date
                into i_case, i_p3, i_date
                from table3
                where table3.ID = i_ID

                if i_case < o_case 
                    then let o_p3, o_case, o_date = i_p3, i_case, i_date;
                    else ( if i_case = o_case and i_date > o_date
                        then let o_p3, o_date = i_p3, i_date;
                    end if;
                end if;
        end foreach;
        insert into table_output values(i_ID; o_p3);
end foreach;

The second approach was to left join the table four times with itself on the ID and apply the different combinations of the parameter1 & parameter2 as described above in the left joins, then selecting the output via nested nvl clauses:

select ID, 
    nvl(t1.parameter3, 
        nvl(t2.parameter3,
            nvl(t3.parameter3,
                nvl(t4.parameter3)))) parameter3
from table1 t0
    left join table1 t1
        on t0.ID = t1.ID and t1.parameter1 = 0 and t1.parameter2 = 0
        and t1.date = (select max(date) from table1 t1a where t1a.ID = t1.ID)        
    left join table1 t2
        on t0.ID = t2.ID and t2.parameter1 = 0 and t2.parameter2 = 1
        and t2.date = (select max(date) from table1 t2a where t2a.ID = t1.ID)
    left join table1 t3
        on t0.ID = t3.ID and t3.parameter1 = 1 and t3.parameter2 = 0
        and t3.date = (select max(date) from table1 t3a where t3a.ID = t3.ID)
    left join table1 t4
        on t0.ID = t4.ID and t4.parameter1 = 1 and t4.parameter2 = 1
        and t4.date = (select max(date) from table1 t4a where t4a.ID = t4.ID)

Both approaches basically worked, however, as the table is really long, they were much too slow. What would you recommend?

PS: DBMS is IBM Informix 10, this unfortunately restricts the range of available functions a lot.

Answer 1

I'm not sure if this is what you wanted, but this could work:

SELECT id, parameter3
FROM (
    SELECT id, parameter3, RANK() OVER (
            PARTITION BY id, parameter3
            ORDER BY parameter1 ASC, parameter2 ASC, date DESC
        )
    FROM tab
) AS x
WHERE x.rank = 1;

Answer 2

ID        parameter1      parameter2        parameter3       date

001       0               1                 A                01.01.2010
001       0               1                 B                02.01.2010

both of the above rows having same ID, paramaeter1, parameter2 but different paraameter3, it can create trouble for you.