Issue with passing column name as a parameter to “PREPARE” in Redshift

I am using REDSHIFT for the below question, Here is problem and I am looking for solution.

I have 2 tables, one table contains the column combinations on which the second table should group the results by.

Table 1 containing column combinations (This is output from a stored procedure):

COMBINATIONS

fruit_combinations 
  banana,'ALL'  
  banana, orange

Table 2 containing fruit_baskets:

FRUIT_BASKET

BANANA ORANGE USER_COUNT 
  b1     o1      5
  b1     o2      10

Result set:

FRUIT_BASKET_AGG

BANANA   ORANGE    USER_COUNT 
  b1     'ALL'       15
  b1      o1         5
  b1      o2         10

Tried solution in Redshift:

PREPARE executesql(VARCHAR(10),VARCHAR(10))
AS 
select $1,$2, sum(fb.user_Count) as user_count
from dv_product.fruit_basket fb
group by 1,2;

EXECUTE executesql("BANANA","ORANGE") ;
DEALLOCATE executesql;

Error: [Amazon](500310) Invalid operation: column "banana" does not exist;

Questions:

1 – Is there a way I can pass column names as parameter to a sql statement and execute the sql?

2 – What is other alternative to achieve the result set mentioned above in Redshift?

Answer

This can now be done using Stored Procedures without the need for PREPARE. “Overview of Stored Procedures in Amazon Redshift”

It seems like you are trying to emulate GROUPING SETS or ROLLUP functionality. I have added a UNION ALL to the dynamic SQL to provide this type of output.

For this example stored procedure, both column names are provided as input and a REFCURSOR is declared as output.

CREATE PROCEDURE get_fruit_sum(IN column_1 VARCHAR, IN column_2 VARCHAR, result_set INOUT REFCURSOR) AS $$
BEGIN
OPEN result_set FOR
EXECUTE 'SELECT '|| quote_ident(column_1) ||' , '|| quote_ident(column_2) 
        || ' , SUM(fb.user_Count) as user_count ' 
        || 'FROM dv_product.fruit_basket fb GROUP BY 1,2'
        || 'UNION ALL '
        || 'SELECT '|| quote_ident(column_1) ||' , ''ALL'''
        || ' , SUM(fb.user_Count) as user_count ' 
        || 'FROM dv_product.fruit_basket fb GROUP BY 1;'  
RETURN;
END;
$$ LANGUAGE plpgsql;

You specify the columns and the output REFCURSOR when calling the procedure. The column names could be retrieved from a table by another stored procedure if needed. Then fetch the output from the REFCURSOR.

BEGIN; 
CALL get_fruit_sum ( 'Banana','Orange','result_set' ); 
FETCH ALL FROM result_set; 
END;