连接中不存在的行的 SQL 聚合函数

Question

I am trying to find a SQL Get aggregate as 0 for non existing row using joins for the below tables for a certain year.

Customers -> customerid, Name
product -> prodcutid, price, name
Sales -> productid,customerid,date, orders

create table product(productid int,name char(20), price int,PRIMARY KEY(productid));
create table Customers(customerid int,name char(20),PRIMARY KEY(customerid));
create table sale(productid int,customerid int,orders int, date date);

INSERT INTO product (productid,name ,price) 
VALUES (1, 'TV', 3200);

INSERT INTO product (productid,name ,price) 
VALUES (2, 'fridge', 4200);

INSERT INTO Customers (customerid,name) 
VALUES (1, 'US');

INSERT INTO Customers (customerid,name) 
VALUES (2, 'Aus');

INSERT INTO sale (productid,customerid,orders,date) 
VALUES (2, 2, 1,'2018-10-2');

INSERT INTO sale (productid,customerid,orders,date) 
VALUES (1, 2, 10,'2018-10-13');

INSERT INTO sale (productid,customerid,orders,date) 
VALUES (1, 1, 100,'2018-10-2');

INSERT INTO sale (productid,customerid,orders,date) 
VALUES (1, 1, 100,'2019-10-1');

I have got this query to find the aggregation for existing order for a products for a certain year but i also wanted to get the products which were not ordered during the same year as others:

  with cte as (select producid,customerid,sum(orders) as s from sales 
  where date_part('year', date) = 2018
  group by productid,customerid)
  select a.name,b.name,(c.s * b.price) as sales
  from customers as a 
  join cte as c on a.prodcutid=c.productid
  join product as b on b.customerid=c.customerid; 

Data what i am able to get with above query:

 -----------------------------------------------
 Product.name | Customer.name | Total_sale Value
 -----------------------------------------------
      TV       |   AUS        |  32000 (orders * price per product) 
      Fridge   |   AUS        |  4200 
      TV       |   US         |  320,000 

Data what i want is for the year 2018:

 -----------------------------------------------
 Product.name | Customer.name | Total_sale Value
 -----------------------------------------------
      TV       |   AUS        |  32000 (orders * price per product) (if there was a sale then you should get a value else then it should be 0)
      Fridge   |   AUS        |  4200 
      Fridge   |   US         |  0 (as Customer US didnt order fridge at all)
      TV       |   US         |  320,000 

I tried following the same solution provided SQL Get aggregate as 0 for non existing row using inner joins but couldn't get the answer, forgive me for the duplicate but please let me know what i did wrong. I am new to sql so please forgive me for duplicates.

Answer 1

Note: the question has been edited since this answer was posted and I'm no longer 100% sure what is being asked

I am trying to find a SQL Get aggregate as 0 for non existing row using joins

Conceptually, the sales table relates products with customers

• There is a product, and there always is a product (because you're looking at this as if "i am interested in every product and whether it sold or not")

• There is perhaps a customer (if they registered and never bought anything they could be a customer with no sales)

• And there is a sale only if a product has been bought by a customer (so customers that never bought anything aren't mentioned in sales, and we're only interested in products that have or haven't sold)

So your relation graph to answer your question needs to look like:

products -> sales -> customer

The products -> sales relation will break if a product has never been ordered, so it needs to be a LEFT JOIN to keep all the products and associates sales with products only if matching sales exist.

The sales -> customer shouldn't break and can be INNER JOIN (if you're willing to join it first) or LEFT JOIN if you don't do it as a subquery (if you make it immer join, any sales that are null because the product has never sold will not match to a customer so the unsold product will disappear from the results

As such, you need a query like:

SELECT ...
FROM
  products p
  LEFT JOIN sales s ON ...
  LEFT JOIN customer c ON ...

Or you need

SELECT ...
FROM
  products p
  LEFT JOIN 
  (SELECT * FROM sales s INNER JOIN customer c ON ...) sc

It might be simpler if you use the second form and perform the grouping inside the subquery, but it's trivial to do these anywhere

The nulls introduced for no-sales against products that never sold will participate in SUM/AVG etc as if they were 0

---

Side note, I'm not really sure what bearing the customer table has on any of your stated requirements, and it can probably be omitted

Answer 2

with cte as (
    select productid, customerid, sum(orders) as s from sale
    where year(date) = 2018
    group by productid, customerid
)
select p.name, c.name, coalesce(s.s * p.price, 0) as sales
from product as p
    cross join customer as c
    left outer join cte as s on s.productid = p.productid
                           and s.customerid = c.customerid