在多个数据库表上按日期进行计数和分组

Question

Working on a report page that contains counts for certain columns from the last 10 ten days I got this to work in c# after multiple queries to the database but ran in to trouble on the date columns so I figured I'd try and write a stored procedure or a single query that returned all the results I needed but I've stressed myself enough and I'm not a SQL guru (this is not homework btw!)

I have 2 different tables in different databases that only have the date in (common) (sort -of) but the column names are different.

db1.order  
-----------  
c1    c2      c3      requestedshipping    dateadded
------------------------------------------------------
1    1.00    .95      ground            2013-05-15 18:34:48.867<-has time stamp in date
2    1.50    1.56     express           2013-05-15 18:34:48.867
3    .50     75       ground            2013-05-14 18:34:48.867
4    .50     75       ground            2013-05-14 18:34:48.867

db2.ship
-----------  
c1    c2         service               dateshipped
------------------------------------------------------
.25    1.00      ground            2013-05-15 00:00:00<-time stamp ?
2.3    1.50      express           2013-05-15 00:00:00
.36    .50       ground            2013-05-14 00:00:00

Trying to get results like

date       tot_order  tot_shipped  tot_express  tot_order_c3   tot_ship_c1   tot_ship_c2
--------------------------------------------------------------------------------------
2013-05-15   2           2            1           2.51           2.28          2.50
2013-05-14   2           1            0           150            .36           .50

THINGS I'VE TRIED I was using these queries to get totals then comparing the dates in c# but with the time stamp numbers weren't matching up

SELECT TOP (10) DATEADD(dd, DATEDIFF(dd, 0, db1order.dateadded), 0) od,   COUNT(db1order.dateadded)
   FROM [db1].[dbo].[order] db1order 
   GROUP BY DATEADD(dd, DATEDIFF(dd, 0, db1order.dateadded), 0) 
   ORDER BY od DESC 



SELECT TOP (10) DATEADD(dd, DATEDIFF(dd, 0, db2ship.dateshipped), 0) od,   COUNT(db2ship.dateshipped)
   FROM [db2].[dbo].[ship] db2ship 
   GROUP BY DATEADD(dd, DATEDIFF(dd, 0, db2ship.dateshipped), 0) 
   ORDER BY od DESC 

I saw a article about using UNION ALL but this unions the totals together.

select TOP (10) dt, count(*)
   from (select DATEADD(dd, DATEDIFF(dd, 0, dateadded), 0) AS dt from [db1].[dbo].   [order] union all
      select DATEADD(dd, DATEDIFF(dd, 0, dateshipped), 0) AS dt from [db2].[dbo].[ship]) v
group by dt
ORDER BY dt DESC 

I seen a couple of articles about creating a temp table but no examples of how to get information like I want into the table.

Answer 1

You can union the results of the two tables in the same format within a Common Table Expression http://technet.microsoft.com/en-us/library/ms190766(v=sql.105).aspx (aka CTE) , and add a "rowType" column to allow you to differentiate between an order vs. a ship row, and then perform your aggregates based on this.

;WITH cteCombined as
(
    SELECT 
          rowType = 0 --orders
        , [date] = convert(date, o.dateadded) 
        , o.c1
        , o.c2
        , o.c3
        , o.requestedshipping
    FROM [db1].[dbo].[order] o 
    UNION ALL
    SELECT 
          rowType = 1 --shipments
        , [date] = convert(date, s.dateshipped)
        , s.c1, s.c2
        , c3 = null
        , requestedshipping = s.[service]
    FROM [db2].[dbo].[ship] s
)
SELECT c.[date]
    , tot_order = sum(case when c.rowtype = 0 then 1 else 0 end)
    , tot_shipped = sum(case when c.rowtype = 1 then 1 else 0 end)
    --EDIT: based on expected results then tot_express appears to be only based on orders (with assumption that you will have an order prior to a shipment)
    , tot_express = sum(case when c.rowType = 0 AND c.requestedshipping = 'EXPRESS' then 1 else 0 end)
    , tot_order_c3 =  sum(case when c.rowtype = 0 then c.c3 else 0 end)
    , tot_ship_c1 =  sum(case when c.rowtype = 1 then c.c1 else 0 end)
    , tot_ship_c2 =  sum(case when c.rowtype = 1 then c.c2 else 0 end)
    /*etc...*/
FROM cteCombined c
GROUP BY c.[date]
;

Answer 2

You could use two CTEs, one for the orders and one for the shipments, and then join the two. I wasn't sure how you wanted to handle express though -- where does that value come from? In the example below, I used the service column from the shipments table, not the orders table, because based on your example you're only summing one of them.

See sqlFiddle

;with orders AS (
  SELECT CAST(dateadded AS DATE) as [date],
        COUNT(*) AS tot_order, 
        SUM(o.c3) AS tot_order_c3
  FROM tblOrder o
  GROUP BY CAST(dateadded AS DATE)
),
shipments AS (
  SELECT CAST(dateshipped AS DATE) as [date],
         COUNT(*) AS tot_shipped,
         SUM(CASE WHEN service = 'express' THEN 1 ELSE 0 END) AS tot_express,
         SUM(c1) AS tot_ship_c1,
         SUM(c2) AS tot_ship_c2
  FROM tblShip
  GROUP BY CAST(dateshipped AS DATE)
)
SELECT ISNULL(s.[date], o.[date]) as [date],
       ISNULL(o.tot_order, 0) as tot_order,
       ISNULL(s.tot_shipped, 0) as tot_shipped,
       ISNULL(s.tot_express, 0) as tot_express,
       ISNULL(o.tot_order_c3, 0) as tot_order_c3,
       ISNULL(s.tot_ship_c1, 0) as tot_ship_c1,
       ISNULL(s.tot_ship_c2, 0) as tot_ship_c2
FROM shipments s
FULL JOIN orders o ON s.[date] = o.[date]