SQL Join two tables, only get latest entry of second table

Question

I have these two tables:

items

itemname  description  belongs
A1        some_text    user1
A2        some_text    user1
A3        some_text    user1
A4        some_text    user1
A5        some_text    user1
A1        some_text    user2
B2        some_text    user2

movements

itemname  start_date  end_date    belongs
A1        2013-02-01  2014-01-12  user1
A1        2014-08-14  NULL        user1
A1        2014-10-15  2015-01-01  user1
A2        2013-08-03  2014-08-14  user1     
A2        2014-08-14  NULL        user1    
A3        2013-08-02  2014-08-20  user1
A3        2013-12-05  2014-01-07  user1
A4        2013-07-15  2014-09-13  user1
A4        2014-09-13  NULL        user1
A5        2013-07-15  2014-09-13  user1
A5        2015-03-11  2016-03-12  user1
A5        2016-03-12  2016-04-13  user1
A1        2015-08-01  2015-08-12  user2
B2        2015-08-13  2015-08-23  user2

I was playing around with joins and max(date) but didn't come to a working solution.

A result for user1 should looks like this:

itemname  description  belongs  start_date  end_date
A1        some_text    user1    2014-08-14  NULL        
A2        some_text    user1    2014-08-14  NULL       
A3        some_text    user1    2013-08-02  2014-08-20      
A4        some_text    user1    2014-09-13  NULL              
A5        some_text    user1    2016-03-12  2016-04-13       

I need the line (movement) with the highest (latest, newest) end_date if there is no line where:

end_date = NULL

If there is a line where end_date = NULL, I need this line for that item.

Difficulty here is that sorting for max(start_date) would not work since sometimes there is a timeperiod inside another timeperiod for one item.

I hope you could understand my problem.

Greetings from germany :)

Answer 1

You need something like this

Find the most recent start date for each itemname & belongs combination then join the result back with max start date to get the result

SELECT i.itemname,i.description,i.belongs,m.start_date,m.end_date
FROM   items i
       JOIN movements m
         ON i.itemname = m.itemname
            AND i.belongs = m.belongs
       JOIN (SELECT itemname,
                    belongs,
                    Max(COALESCE(end_date, start_date)) AS max_dt,
                    Max(end_date)                       AS max_end_dat,
                    Max(start_date)                       AS max_start_dat
             FROM   movements
             GROUP  BY itemname,
                       belongs) m1
         ON m1.itemname = m.itemname
            AND m1.belongs = m.belongs
            AND ( ( m.end_date = m1.max_dt
                    AND m1.max_dt = m1.max_start_dat )
                   OR ( m1.max_dt = COALESCE(end_date, m.start_date)
                        AND m1.max_start_dat <> m1.max_dt )
                   OR ( m1.max_dt = m.start_date
                        AND m1.max_end_dat <> m1.max_dt ) )
ORDER  BY i.belongs,
          i.itemname 

• SQL FIDDLE DEMO

Update :

SELECT i.itemname,
       i.description,
       i.belongs,
       m.start_date,
       m.end_date
FROM   items i
       JOIN movements m
         ON i.itemname = m.itemname
            AND i.belongs = m.belongs
       JOIN (SELECT itemname,
                    belongs,
                    Max(start_date) AS max_dat,
                    'st'            AS indi
             FROM   movements
             WHERE  end_date IS NULL
             GROUP  BY itemname,
                       belongs
             UNION ALL
             SELECT itemname,
                    belongs,
                    Max(end_date) AS max_dat,
                    'ed'
             FROM   movements m
             WHERE  NOT EXISTS (SELECT 1
                                FROM   movements m1
                                WHERE  m.itemname = m1.itemname
                                       AND m.belongs = m1.belongs
                                       AND end_date IS NULL)
             GROUP  BY itemname,
                       belongs) m1
         ON m1.itemname = m.itemname
            AND m1.belongs = m.belongs
            AND ( ( m1.max_dat = m.start_date
                    AND indi = 'st' )
                   OR ( m1.max_dat = m.end_date
                        AND indi <> 'st' ) )
ORDER  BY i.belongs,
          i.itemname 

• SQL FIDDLE DEMO

This will be really easy if your RDBMS supports ROW_NUMBER window function or APPLY operator

Answer 2

to select a record from a group of records: IN ALL the Methods you can join with your Items table to get extra information.

Method-1 use self join:

Select a.*  
from yourTable as a
left join YourTable as b
  on b.belongs = a.Belongs
 and b.Item_Name=a.Item_Name
 and b.Start_date>a.Start_Date
where b.Start_Date is null

Method-2, use MaxDate.

with MaxDate as(
Select belongs, Item_Name, Max(Start_Date) as MaxDate
from yourTable
group by belongs, Item_Number
) select * from MaxDate as a
  inner join YourTable as b 
     on b.belongs=a.belongs 
    and b.Item_Number=a.Item_Number 
    and b.Start_Date=a.MaxDate

Method-3, use Row_Number to flag the last record in the group

with LastRec as(
select *,
Row_Number() over(partition by belongs, Item_number order by Start_date desc) as RN
From yourTable
) select * from LastRec
  where rn=1

Method-4, use sub-query to get the last record in the group.

Select * from your table as a
Where a.StartDate=(Select Max(Start_Date) from yourTable as b Where b.belongs=a.belongs and b.Item_Name=a.Item_Number)

Answer 3

MySQL currently doesn't seem to have a ROW_NUMBER() window function.

But it can be simulated through variables.

In the example below, the @rn variable is calculated.

In MS SQL Server that would be done via a ROW_NUMBER() OVER (PARTITION BY itemname,belongs ORDER BY coalesce(end_date,getdate()) desc, start_date desc) instead.

Then because of the WHERE clause rn = 1 , it will only return the movements with the maximum end_date, with null considered as max.

select m.itemname, i.description, m.belongs, m.start_date, m.end_date, m.rn
from (
  select itemname, belongs, start_date, end_date,
  @rn := if(@itemname = itemname and @belongs = belongs, @rn+1,1) as rn,
  @itemname := itemname as i,
  @belongs := belongs as b
  from movements
  order by itemname, belongs, coalesce(end_date,'9999-12-31') desc, start_date desc
) m
left join items i on (m.itemname = i.itemname and m.belongs = i.belongs)
where m.rn = 1 
order by m.itemname, m.belongs