How can I join two tables with a many-to-one relationship ordered by date? - sql

Question

I need to join two tables and get the most recent record only. Here is the basic form:
table1.id | table1.region | table1.important_col1
1 | NORTH AMERICA | abc
2 | CHINA | def
2 | NORTH AMERICA | hij
table2.id | table2.region | table2.transaction_date | table2.important_col2
1 | NORTH AMERICA | 2/13/2019 | xyz
1 | NORTH AMERICA | 1/13/2019 | zzz
1 | NORTH AMERICA | 12/13/2018 | xxx
desired result:
1 | NORTH AMERICA | 2/13/2019 | abc | xyz
I wanted to use this answer but it seems like I can't use it if I need to group by and then order by descending date. I will need information in multiple columns on the right hand side, but do not want duplicate rows on the left hand side.
The right hand side may have up to 100s of records per id, but I just need something that works for now. Thanks in advance.
edit: I also need to filter the right hand side on other criteria so a simple MAX(table2.transaction_date) won't work.

Answer 1

You can filter your table using internal window function, I used LAG for this example, but you can use ROW_NUMBER and filter several records. Using sliding windows does not change the number of records or counted as SQL aggregation, i.e. you filter using where rather than with having.
SELECT
t1.id
,t2.transaction_date
,t1.region
,t1.col1
,t2.important_col2
FROM table1 AS t1
OUTER APPLY (
SELECT
id
,transaction_date
,LAG(transaction_date,1) over (partition by id order by transaction_date desc) as prev_td
,important_col2
FROM table2
-- WHERE filter_by_col=1 -- additonal "right side" filtering
) as t2
where t1.id = t2.id
and t2.prev_td is null
Output:
1 2019-02-13 00:00:00.000 NORTH AMERICA abc xyz
I used this to test the above query:
create table table1
(id int,
region varchar(30),
col1 varchar(100));
insert into table1
values (1 ,'NORTH AMERICA' ,'abc'),
(2,'CHINA','def'),
(2,'NORTH AMERICA','hij');
create table table2
(id int,
region varchar(30),
transaction_date datetime,
important_col2 varchar(100))
insert into table2
values
(1 ,'NORTH AMERICA',convert(datetime, '02/13/19', 1),'xyz'),
(1 ,'NORTH AMERICA',convert(datetime, '01/13/19',1),'zzz'),
(1 ,'NORTH AMERICA',convert(datetime, '12/13/18',1),'xxx')

Answer 2

Try in this way:
select table11.id, table1.region, max(table2.transaction_date) transaction_date
from table1
inner join table2
on table1.id = table2.id
group by table1.id, table1.region

Answer 3

If there are more columns in table2 (other than transaction date) that you want to display as well, then aggregation alone cannot solve your question.
In MySQL 8.0 you can use window function ROW_NUMBER() to identify the most recent transaction record, as follows :
SELECT x.*
FROM (
SELECT
t1.*,
t2.*,
ROW_NUMBER() OVER(PARTITION BY t2.region ORDER BY t2.transaction_date DESC) rn
FROM table1 t1
INNER JOIN table2 t2 ON t1.region = t2.region
) x
WHERE x.rn = 1
In earlier versions of MySQL, one solution is to add a NOT EXISTS with a correlated subquery that ensures that we are joining with the most recent transaction for the current region :
SELECT t1.*, t2.*
FROM table1 t1
INNER JOIN table2 t2
ON t1.region = t2.region
AND NOT EXISTS (
SELECT 1
FROM table2
WHERE region = t2.region AND transaction_date > t2.transaction_date
)