BigQuery：以 UNION 作为 ARRAY 的子查询

Question

I have the following two example tables

Orders Table

order_id	linked_order1	linked_order2
1001	L005	null
1002	null	null
1003	L006	L007

Invoices Table

order_id	linked_order_id	charge
1001	null	4.27
1002	null	9.82
1003	null	7.42
null	L005	2.12
null	L006	1.76
null	L007	3.20

I need to join these so the charges of all the orders (linked and otherwise) can be shown as part of the single order row. My desired output is something like this.

Desired Output

order_id	linked_order1	linked_order2	invoices.charge	invoices.order_id	invoices.linked_order_id
1001	L005	null	4.27	1001	null
			2.12	null	L005
1002	null	null	9.82	null	null
1003	L006	L007	7.42	null	null
			1.76	null	L006
			3.20	null	L007

I can manage to get the main order into the table as follows.

SELECT
  orders,
  ARRAY(
  SELECT AS STRUCT * FROM `invoices_table` WHERE order=orders.order_id) AS invoice
FROM
  `orders_table` AS orders

I can run a separate query to union all of the invoice results into a single table for given order ids but I can't combine this with the above query with out getting errors.

Something like this...

SELECT
  orders,
  ARRAY(
  SELECT AS STRUCT * FROM 
(SELECT * FROM `invoices_table` WHERE order=orders.order_id
    UNION ALL SELECT * FROM `invoices_table` WHERE linked_order_id=orders.linked_order1
    UNION ALL SELECT * FROM `invoices_table` WHERE linked_order_id=orders.linked_order2)
) AS invoice
FROM
  `orders_table` AS orders

But this gives me the correlated subqueries error.

[Update]

This is much simpler than I thought. The following query gives me what I was after.

SELECT
  orders,
  ARRAY(
  SELECT AS STRUCT * FROM `invoices_table` WHERE order=orders.order_id OR linked_order_id IN(orders.linked_order1, orders.linked_order2)) AS invoice
FROM
  `orders_table` AS orders

Answer 1

Using CROSS JOINS,

SELECT o.*, ARRAY_AGG(i) invoices
  FROM Orders o, Invoices i 
 WHERE o.order_id = i.order_id 
    OR i.linked_order_id IN (o.linked_order1, o.linked_order2)
 GROUP BY 1, 2, 3;

Query results

[UPDATE]

Sometimes the query using OR conditions in WHERE clause might show poor perfomrance in large dataset. In that case you may try below query instead that generates same result.

SELECT o.*, ARRAY_AGG(i) invoices FROM (
  SELECT o, i FROM Orders o JOIN Invoices i USING (order_id)
   UNION ALL
  SELECT o, i FROM Orders o JOIN Invoices i ON i.linked_order_id IN (o.linked_order1, o.linked_order2)
) GROUP BY 1, 2, 3;

Answer 2

For the desired output table, the full outer join is the right command.

with tblA as (Select order_id, 1 linked_order1, 2   linked_order2, from unnest([1,2,3]) order_id),
tblB as (Select order_id, 109.99 charge from unnest([3,4,5]) order_id
union all select null  order_id, * from unnest([50.1,29.99]) charge
)

Select *  
  from tblA
  full join tblB
  using(order_id)

For your setting, there is the need to have several joining conditions. Therefore, the first table is used three times, for each joining key.

with tblA as (Select order_id, "L05" linked_order1, "L2"    linked_order2, from unnest(["1","2","3"]) order_id),
tblB as (Select order_id, null linked_order_id, 109.99 charge from unnest(["3","4","5"]) order_id
union all select null  order_id, "L05" , * from unnest([50.1,29.99]) charge
)

Select A.order_id,linked_order1,linked_order2, array_agg(struct(tblB.order_id,linked_order_id,charge))
  from 
  (
    Select * from tblA, unnest([order_id,linked_order1,linked_order2]) as tmp_id
  ) A
  full join tblB
  on tmp_id = ifnull(tblB.order_id,linked_order_id)
  where charge is not null #or tmp_id=A.order_id
  group by 1,2,3