SQL Group By and Join based on a weird client table - sql

Question

I have 3 tables that I want to join together and group it to get client membership info. My code works for grouping the base table together but it breaks at the join part and I can't figure out why.
BASE TABLE : sales_detail
+-------+-----------+-----------+-----------------------------------------+
| order_date | transaction_id| product_cost | payment_type | country
+-------+-----------+-----------+------------------------------------------+
| 10/1 | 12345 | 20 | mastercard | usa
| 10/1 | 12345 | 50 | mastercard | usa
| 10/5 | 82456 | 50 | mastercard | usa
| 10/9 | 64789 | 30 | visa | canada
| 10/15 | 08546 | 20 | mastercard | usa
| 10/15 | 08546 | 90 | mastercard | usa
| 10/17 | 65898 | 50 | mastercard | usa
+-------+-----------+-----------+-------------------------------------+
table : client_information
+-------+-----------+-----------+-------------------+
| other_id | client_Type | item
+-------+-----------+-----------+----------+
| 112341 | new | hola |
| 112341 | old | mango |
| 145634 | old | pine |
| 879547 | old | vip |
| 745688 | new | unio |
| 745688 | old | dog |
| 147899 | new | cat |
| 124589 | new | amigo |
+-------+-----------+-----------+-----------+
table : connector
+-------+-----------+-----------+-------------------+
| transaction_ID | other_id | item
+-------+-----------+-----------+----------+
| 12345 | 112341 | hola |
| 82456 | 145634 | pine |
| 08157 | 879547 | unio |
| 08546 | 745688 | dog |
| 65898 | 147899 | cat |
| 06587 | 124589 | amigo |
+-------+-----------+-----------+-----------+
**I want the output to look something like this: **
IDEAL OUTPUT
+-------+-----------+-----------+--------------------------------+
| order_date | transaction_ID | product_cost | client_Type|
+-------+-----------+-----------+--------------------------------+
| 10/1 | 12345 | 70 | new |
| 10/5 | 82456 | 70 | old |
| 10/15 | 08546 | 110 | old |
| 10/17 | 65898 | 50 | new |
+-------+-----------+-----------+----------------------------------+
**i am trying to join my base table to the connector table by transaction ID to get other_id and items to match to client_type **
This is the code i used but it failed to compile after adding in left joins :
select t1.transaction_id, sum(t1.product_cost), t1.order_date, t3.client_type
from sales_detail t1
left join (select DISTINCT transaction_ID, other_id, fruits from connector) t2
ON t1.transaction_ID=t2.transaction_ID
left join (select DISTINCT order_id, client_type, fruits from client information) t3
ON t2.other_id=t3.other_id and t2.item=t3.item
where t1.payment_type='mastercard' and t1.order_Date between '2020-10-01' and'2020-10-31'
and country != 'canada'
GROUP BY t1.transaction_id, t1.order_date, t3.client_type;
Thanks in advance! I am a beginner so still learning the ins and outs of sql! (am using hive)

Answer 1

I think that's joins and aggregation. For more efficiency, you can pre-aggregate in a subquery, then join:
select sd.*, ci.client_type
from (
select order_date, transaction_id, sum(product_cost) product_cost
from sales_detail
where
payment_type = 'mastercard'
and order_date >= '2020-10-01'
and order_date < '2020-11-01'
and country <> 'canada'
group by order_date, transaction_id
) sd
inner join connector c on c.transaction_id = sd.transaction_id
inner join client_information ci on ci.other_id = c.other_id
Note that I rewrote the filter on order_date to use half-open intervals rather than between. This properly handles the case when your dates have a time portion.

Answer 2

From what I have understood, your code works although not as you would like using an INNER JOIN and it fails to add a LEFT JOIN. I think what happens is a failure due to the NULL elements. to add a NULL element and not get an error, you have to use some function that changes the NULL value to 0 .
One such function is the ISNULL(yourColumn, 0) function of T-SQL.The documentation.

Answer 3

I can see that in result table you only need clients who used mastercard, so you should use inner join there so only those client who used mastercard will be considered. While the remaining query is okay i guess, but main problem was the join on client information.

Answer 4

I think on the answer with GMB you also need to join on item column otherwise you will get multiple rows output.
select sd.*, ci.client_type
from (
select order_date, transaction_id, sum(product_cost) product_cost
from sales_detail
group by order_date, transaction_id
) sd
inner join connector c on c.transaction_id = sd.transaction_id
inner join client_information ci on ci.other_id = c.other_id and ci.item = c.item
Just modify with your filters and you should be sorted.