SQL Aggregation with SUM, GROUP BY and JOIN (many-to-many)

Question

Here's an example Table layout:

TABLE_A:                    TABLE_B:     TABLE_A_B:
id | a     | b    | c       id | name    a_id | b_id
---------------------       ---------    -----------
1  | true  | X    | A       1  | A       1    | 1
2  | true  | Z    | null    2  | B       1    | 2
3  | false | X    | null    3  | C       2    | 2
4  | true  | Y    | Q                    4    | 1
5  | false | null | null                 4    | 2
                                         5    | 1

Possible Values:

• TABLE_A.a: true, false
• TABLE_A.b: X, Y, Z
• TABLE_A.c: A, B, C, ... basically arbitrary
• TABLE_B.name: A, B, C, ... basically arbitrary

What I want to achieve:

SELECT all rows from TABLE_A
  SUM(where a = true),
  SUM(where a = false),
  SUM(where b = 'X'),
  SUM(where b = 'Y'),
  SUM(where b = 'Z'),
  SUM(where b IS NULL),
and also get the SUMs for all distinct TABLE_A.c values.
and also get the SUMs for all those TABLE_A_B relations.

The result for the example Table above should look like:

aTrue | aFalse | bX | bY | bZ | bNull | cA | cQ | cNull | nameA | nameB | nameC
-------------------------------------------------------------------------------
3     | 2      | 2  | 1  | 1  | 1     | 1  | 1  | 3     | 3     | 3     | 0

What I've done so far:

SELECT
  SUM(CASE WHEN a = true THEN 1 ELSE 0 END) AS aTrue,
  SUM(CASE WHEN b = false THEN 1 ELSE 0 END) AS aFalse,
  SUM(CASE WHEN b = 'X' THEN 1 ELSE 0 END) AS bX,
  ...
FROM TABLE_A

What's my problem?

Selecting column TABLE_A.a and TABLE_A.b is easy, because there's a fixed number of possible values.

But I can't figure out how to count the distinct values of TABLE_A.c . And basically the same problem for the JOINed TABLE_B , because the number of values within TABLE_B is unknown and can change over time.

Thanks for your help! :)

EDIT1: New (preferred) SQL result structure:

column         | value | sum
----------------------------
TABLE_A.a      | true  | 3
TABLE_A.a      | false | 2
TABLE_A.b      | X     | 2
TABLE_A.b      | Y     | 1
TABLE_A.b      | Z     | 1
TABLE_A.b      | null  | 1
TABLE_A.c      | A     | 1
TABLE_A.c      | Q     | 1
TABLE_A.c      | null  | 3
TABLE_B.name   | A     | 3
TABLE_B.name   | B     | 3
TABLE_B.name   | C     | 0

Answer 1

From your original request of rows as a simulated pivot. By doing a SUM( logical condition ) basically returns 1 if true, 0 if false. So, since the column "a" is true or false, simple sum of "a" or NOT "a" (for the false counts -- NOT FALSE = TRUE). Similarly, your "b" column, so b='X' = true counted as 1, else 0.

In other sql engines, you might see it as SUM( case/when ).

Now, since your table counts don't rely on each other, they can be separate SUM() into their own sub-alias query references (pqA and pqB for pre-queryA and pre-queryB respectively). Since no group by, they will each result in a single row. With no join will create a Cartesian, but since 1:1 ratio, will only return a single record of all columns you want.

SELECT 
      pqA.*, pqB.*
   from
      ( SELECT
              SUM( ta.a ) aTrue,
              SUM( NOT ta.a ) aFalse,
              SUM( ta.b = 'X' ) bX,
              SUM( ta.b = 'Y' ) bY,
              SUM( ta.b = 'Z' ) bZ,
              SUM( ta.b is null ) bNULL,
              SUM( ta.c = 'A' ) cA,
              SUM( ta.c = 'Q' ) cQ,
              SUM( ta.c is null ) cNULL,
              COUNT( distinct ta.c ) DistC
           from
              table_a ta ) pqA,
      ( SELECT
              SUM( b.Name = 'A' ) nameA,
              SUM( b.Name = 'B' ) nameB,
              SUM( b.Name = 'C' ) nameC
           from
              table_a_b t_ab 
                 join table_b b
                    ON t_ab.b_id = b.id ) pqB

This option gives your second (preferred) output

SELECT
      MAX( 'TABLE_A.a   ' ) as Basis,
      CASE when a then 'true' else 'false' end Value,
      COUNT(*) finalCnt
   from
      TABLE_A
   group by
      a
UNION ALL
SELECT
      MAX( 'TABLE_A.b   ' ) as Basis,
      b Value,
      COUNT(*) finalCnt
   from
      TABLE_A
   group by
      b
UNION ALL
SELECT
      MAX( 'TABLE_A.c   ' ) as Basis,
      c Value,
      COUNT(*) finalCnt
   from
      TABLE_A
   group by
      c
UNION ALL
SELECT
      MAX( 'TABLE_B.name   ' ) as Basis,
      b.Name Value,
      COUNT(*) finalCnt
   from
      table_a_b t_ab 
         join table_b b
            ON t_ab.b_id = b.id 
   group by
      b.Name

Answer 2

I think You will need to build dynamic query as you don't know possible values for column C in table A. So you can write store procedure where you can get list of distinct value for Column C in one variable and by using "Do WHILE" you can construct your dynamic query. Please let me know if you need more help in detail Dynamic SQL