Oracle组按日期排序并总计总和

Question

My Orders table looks like:

order_id     (number)
order_total  (number)
created_date (timestamp)
status       (varchar2)

My goal is to get a set of rows where each row represents all orders on that date, so I'm trying to group the orders by date and get the sum of the order_total . I'm also limiting the results by only selecting orders from the last 30 days.

To clarify, for example if there were 30 orders in the last 30 days all on unique days then I would get 30 rows in the result. Another example: if there were 10 orders on 30th July, and only 1 order on 31st July then I'm aiming to get 2 rows in the result set, with order_total summed for all 10 orders in the first row, and the second row would of course have the order_total of the single order on the 31st.

My attempt so far:

select
  sum(order_total) total_amount,
  to_char(created_date, 'DD/MM/YYYY') grouped_date
from
  orders
where
  status = 'Complete' and
  created_date >= (sysdate-30)
group by
  to_char(created_date, 'DD'), to_char(created_date, 'MM'), to_char(created_date, 'YYYY')
order by
  created_date asc

This gives an error:

ORA-00936: missing expression

I have tried to use the solution from this question but I don't think it quite fits for my scenario (this is where my group by expression has come from).

Answer 1

Assuming order_id should not be there, and that created_date has a time component (which seems likely as it's a timestamp ), you need to truncate the date to remove the time when doing the aggregation:

select
  sum(order_total) as total_amount,
  to_char(trunc(created_date), 'DD/MM/YYYY') as grouped_date
from
  orders
where
  status = 'Complete' and
  created_date >= trunc(sysdate-30)
group by
  trunc(created_date)
order by
  trunc(created_date) asc

I've also applied trunc to the where clause, otherwise it would ignore any orders 30 days ago between midnight and whatever time you ran the query today. And I've used the trunc'd date directly in the order by , rather than the column alias, so that the order is right when you go across a month-end - ordering by the DD/MM/YYYY string value would put 01/07/2013 before 30/06/2013, for example.

Quick SQL Fiddle .