Question

PROBLEM: Need query to return the MONTH and YIELD for Each Year. For some reason, if the data is not found in a Month a.Month the query will not return the b.Month’s Yield. I need the query to return all monthly data regardless of whether or not a.Month contains a month with data in the same months as “b”.

THE FOLLOWING RESULT: SHOULD RETURN A VALUE FOR “MONTH 1 YIELD_1”. But it doesn’t… because “MONTH 1 YIELD_0” does NOT contain a value for month 1.

**DATA RESULTS WITH: LEFT OUTER JOIN:**
Month   Yield_1    Yield_0
2        11.44      14
3         NULL     3.21
4         NULL     14.24
7         NULL     10.36
8         NULL       0
9         NULL     -9.6
10        NULL     10.35
11        NULL      1.4
12        11.44    -1.18


**DATA RESULTS WITH RIGHT OUTER JOIN:**
Month   Yield_1    Yield_0
NULL     11.44      NULL
2        11.44       14
12       11.44     -1.18

QUERY:

SET @ID_CARTERA = 8;

select     
        a.Month Month,
        b.Monthly_Yield Yield_Year_1,
        a.Monthly_Yield Yield_Year_0

from
    ( select  
          LEFT(A.F_ANOMES, 4) Year,
          RIGHT(A.F_ANOMES, 2) Month,
          ROUND(A.POR_RENTABILIDAD, 2) Monthly_Yield

from      dr_rent_carteras_meses A
where     A.ID_CARTERA = @ID_CARTERA
And       A.IND_RENTABILIDAD = 1

And       LEFT(A.F_ANOMES, 4) = ( select MAX(left(F_ANOMES, 4 ) ) - 0 from dr_rent_carteras_meses where ID_CARTERA = @ID_CARTERA ) ) a


LEFT outer join 
        ( select  
          LEFT(A.F_ANOMES, 4) Year,
          RIGHT(A.F_ANOMES, 2) Month,
          ROUND(A.POR_RENTABILIDAD, 2) Monthly_Yield

from      dr_rent_carteras_meses A
where     A.ID_CARTERA = @ID_CARTERA
And       A.IND_RENTABILIDAD = 1
And       LEFT(A.F_ANOMES, 4) = ( select MAX(left(F_ANOMES, 4 ) ) - 1 from dr_rent_carteras_meses where ID_CARTERA = @ID_CARTERA ) ) b on ( a.Month = b.Month )

order by  month asc

Answer 1

Summary

I don’t think you need either left, right, or full join for this specific query as a GROUP BY/CASEbased solution should work just fine and be at least twice as much faster.

Problem Definition

I found this question rather interesting as it seems to arise from a real life situation and I believe that in real life FULL JOIN is seldom if ever necessary. So to me the real question is not how to combine the data from left and right joins but rather why is a full join needed in the first place?

Besides, simple replacing LEFT with FULL would not be enough as that would lead to

1. Duplicate rows for the same months;
2. NULL months for when the month is not available in the latest year.

So I decided to go ahead and decipher the query.

Table and Data Setup

Unfortunately, @smileyseven didn’t provide table definitions or insert statements, so I had to work them out from the query and sample data. Based on the query syntax I’m assuming that SQL Server was used so that’s what I use as well.

Here’s my guess at how the table looks:

CREATE TABLE dr_rent_carteras_meses 
(
    F_ANOMES varchar(6) NOT NULL PRIMARY KEY,
    POR_RENTABILIDAD float NOT NULL,
    -- These columns are  irrelevant 
    ID_CARTERA INT NOT NULL DEFAULT(8),
    IND_RENTABILIDAD INT DEFAULT(1) NOT NULL
)   

The important point ist hat F_ANOMES is probably the key column as otherwise we would have to expect multiple rows for each month in the query output and that seems unlikely.

The following insert statements should produce the sample data:

INSERT dr_rent_carteras_meses (F_ANOMES, POR_RENTABILIDAD) VALUES 
('201202', 14),
('201203', 3.21),
('201204', 14.24),
('201207', 10.36),
('201208', 0),
('201209', -9.6),
('201210', 10.35),
('201211', 1.4),
('201212', -1.18),
('201101', 11.44),
('201102', 11.44),
('201112', 11.44)

Solution

The first thing to note is that we don’t really need to calculate the maximum year two times, so we’ll start with this:

declare @Max_Year int
select 
    @Max_Year = MAX(left(F_ANOMES, 4))
from dr_rent_carteras_meses 
where ID_CARTERA = @ID_CARTERA

The a and b tables are virtually same, so we could probably reuse them:

;with 
Monthly_Yield_CTE as
( 
    select  
        LEFT(F_ANOMES, 4) Year,
        RIGHT(F_ANOMES, 2) Month,
        ROUND(POR_RENTABILIDAD, 2) Monthly_Yield
    from      dr_rent_carteras_meses
    where     ID_CARTERA = @ID_CARTERA
    and       IND_RENTABILIDAD = 1
    and       LEFT(F_ANOMES, 4) in (@Max_Year, @Max_Year - 1)
)

and use the FULL JOIN but I think a better option is to simply group it by month:

select
    [Month],
    SUM(CASE WHEN [Year] = @Max_Year - 1 THEN Monthly_Yield ELSE 0 END) Yield_Year_1,
    SUM(CASE WHEN [Year] = @Max_Year THEN Monthly_Yield ELSE 0 END) Yield_Year_0
from Monthly_Yield_CTE
group by [Month]
order by [Month]

One could use either the CTE version or re-write it without CTE as the query is simple enough:

SET @ID_CARTERA = 8

declare @Max_Year int
select 
    @Max_Year = MAX(left(F_ANOMES, 4))
from dr_rent_carteras_meses 
where ID_CARTERA = @ID_CARTERA

select  
    RIGHT(F_ANOMES, 2) Month,
    SUM(CASE 
            WHEN LEFT(F_ANOMES, 4) = @Max_Year - 1 
            THEN ROUND(POR_RENTABILIDAD, 2) 
            ELSE 0 
        END) Yield_Year_1,
    SUM(CASE 
            WHEN LEFT(F_ANOMES, 4) = @Max_Year 
            THEN ROUND(POR_RENTABILIDAD, 2) 
            ELSE 0 
        END) Yield_Year_0
from      dr_rent_carteras_meses
where     ID_CARTERA = @ID_CARTERA
and       IND_RENTABILIDAD = 1
and       LEFT(F_ANOMES, 4) in (@Max_Year, @Max_Year - 1)
group by RIGHT(F_ANOMES, 2)
order by 1

The only drawback I’m aware of is that we get 0 instead of NULL for the months where the data is missing and I’m not sure if that’s important.

Performance

The performance of this query seems to be quite a bit better; in my setup the relative combined cost of the GROUP BY/CASE query with a separated MAX is about 30% vs 70% for FULL JOIN solution.