SQL - aggregation with groupby and partitioning

Question

I have a table that looks like this called sales

 brand  | model    | sales           
--------+----------+------
 brand1 | model11  | 100
 brand1 | model11  | 300
 brand2 | model21  | 100
 brand2 | model22  | 400 
 brand3 | model31  | 100

I was using the following SQL query

SELECT 
     brand, SUM(sales)/SUM(sales) OVER () as sales_share
FROM sales
GROUP BY brand

So as to get the sales_share for each brand as below

 brand  | sales_share          
--------+--------------
 brand1 | 0.4
 brand2 | 0.5
 brand3 | 0.1

However, I was getting error Attribute SALES.sales must be GROUPed or used in an aggregate function - any pointers ?

Answer 1

The following works

SELECT 
     brand, SUM(sales)/SUM(SUM(sales)) OVER () as sales_share
FROM sales
GROUP BY brand

What was missing in the original attempt was the aggregation function - exactly like the error message said :)

Answer 2

If you are using sql you should group by each attribute you select. So your query would be something like :

SELECT 
     brand, SUM(sales)/SUM(sales) OVER () as sales_share
FROM sales
GROUP BY brand, sales

Answer 3

You can use,

select  
     brand, SUM(sales)as brand_sales
into tmp
from sales 
group by brand

select  brand , 
        total_sales /(select SUM(b.sales) from sales b) as total_sales 
from tmp 

drop table tmp

Answer 4

Try this :

  select brand , sales_share/sum(sales_share) over () sales_share from (
    SELECT 
       brand, SUM(sales)  as sales_share
    FROM sales
       GROUP BY brand
    ) a

Answer 5

You don't really need a window function for this:

SELECT brand, 
       sum(sales) / (select sum(sales) from sales) as sales_share
FROM sales
GROUP BY brand;

Depending on the data type of sales you might need a cast to a decimal, otherwise Postgres will use integer division

SQLFiddle: http://sqlfiddle.com/#!15/548ad/1