Reputation: 1214
I have a table itemsInShippment
with the following data:
itemid shippmentid qty
10 1 100
20 1 200
10 2 300
10 3 1000
and table shippments
shippmentid date shippmentstatus supplierid
1 2015-01-12 OK 5000
2 2015-01-17 OK 5000
3 2015-01-17 Cancelled 5000
I need to write a query that shows this details about specific shippment say shipmentid 1
. My given parameters are supplierid
and date
. together they related to one shipment (unique).
For supplierid=5000
and date=2015-01-12
I want to get:
itemid qty qtyInOtherShipments
10 100 300 //1000 is canceled.
20 200 0
My query works fine without considering the cancelled:
SELECT cte.*
FROM
(SELECT
a.itemid, b.date, a.qty,
(coalesce( SUM(a.qty) OVER (PARTITION BY a.itemid), 0) -
coalesce( SUM(a.qty) OVER (PARTITION BY a.itemid, a.shipmentid) ,0) ) AS qtyInOtherShipments,
FROM
itemsInShippment a
LEFT JOIN
shippments b using (shippmentid)
WHERE
b.supplierid = 5000) AS cte
WHERE
cte.date = '2015-01-12'
the cte
must be this way as in qtyInOtherShipments
I Sum the total qty
and then remove my own qty
. In order to sum the total qty
I can't do WHERE d.date=...
inside I must do that outside.
This query gives:
itemid qty qtyInOtherShipments
10 100 1300
20 200 0
I'm having trouble taking under consideration the cancelled
shipments.
if I change the Where
to :
where b.supplierid = 5000 and b.shippmentstatus not like 'cancelled'
it works... I will see:
itemid qty qtyInOtherShipments
10 100 300
20 200 0
but if I run the query on cancelled shipments (supplierid=5000 and date=2015-01-17
) I will get:
itemid qty qtyInOtherShipments
nothing
what I should have get is:
itemid qty qtyInOtherShipments
10 1000 300
so my problem is that I don't want to sum itemid
that is related to cancelled but I still want to see this rows.
How do I get the correct result?
Upvotes: 2
Views: 1922
Reputation: 324405
A more efficient variant of Florian's answer exists for PostgreSQL 9.4, the filter
clause for an aggregate.
SUM (a.qty) FILTER (WHERE b.shippmentstatus <> 'cancelled') OVER (PARTITION BY ...
See FILTER
in the docs for aggregates. It's basically a mini-WHERE
clause that applies only for that aggregate.
Thanks to @a_horse_with_no_name for pointing it out earlier.
Upvotes: 3
Reputation: 1057
Try Below query
create table #itemsInShippment (itemid int, shippmentid int, qty int)
insert into #itemsInShippment (itemid, shippmentid, qty)
SELECT 10 as itemid, 1 as shippmentid, 100 as qty UNION
SELECT 20 , 1, 200 UNION
SELECT 10 , 2, 300 UNION
SELECT 10 , 3, 1000
CREATE TABLE #shippments (shippmentid int , dt date, shippmentstatus varchar(50), supplierid int)
insert into #shippments (shippmentid, dt, shippmentstatus,supplierid)
SELECT 1 as shippmentid, '2015-01-12' as dt, 'OK' as shippmentstatus , 5000 as supplierid UNION ALL
SELECT 2, '2015-01-17', 'OK' , 5000 UNION ALL
SELECt 3, '2015-01-17' , 'Cancelled' , 5000
SELECT cte.*
FROM (
select a.itemid,b.dt,a.qty,
(coalesce( SUM(case when shippmentstatus <> 'Cancelled' then a.qty else 0 end) OVER (PARTITION BY a.itemid) ,0) -
coalesce( SUM(case when shippmentstatus <> 'Cancelled' then a.qty else 0 end) OVER (PARTITION BY a.itemid,a.shippmentid) ,0) )
AS qtyInOtherShipments
from #itemsInShippment a
left join #shippments b on a.shippmentid = b.shippmentid
where b.supplierid = 5000 --and shippmentstatus = 'Cancelled'
) as cte
where cte.dt='2015-01-12'
Upvotes: 0
Reputation: 17643
You want to exclude canceled items only from sums. So, do not filter them with where, just filter them on sums:
SUM(case when b.shippmentstatus <> 'cancelled' then a.qty end) OVER (PARTITION BY ...
Sum does not take in consideration null, that's why the above works. (When status is canceled the case expression will return null.)
Upvotes: 3