来自MS查询计算和分组问题的Excel 2010数据透视表

Question

First post here, you all have helped me so much.

My issue is that I am building a Sales Board in Excel 2010 using ODBC connection to a SQL Server 2008 R2 database. I have two row groups and two column groups. Groups are in parent child order.

Row groups are Product Grade and Product. Column groups are Date and Region(East and West). Value is product price by region. I'm trying to add a subtotal after each region group which gives (East price - West price). Sometimes only one region is reported so if it is West the subtotal shouldn't exist.

I've tried a combination of calculated items, which almost works but adds a column for each region. Or doing the work in SQL Server which almost works except the Product Grade grouping spreads all products in every parent group. I'd prefer to make this work in SQL Server for obvious reasons and I am new to building data in Excel. I'm sure it is something very simple I'm missing. Any help would be much appreciated.

;with t as 
(  
   select  
      info.protype,  
      info.product,  
      case when info.pricecode = 'x' then 'East' else 'West' end as Region, 
      SUM(isnull(info.price,0)) as Price,
      case when r1.product=r2.product then SUM(x.price-y.price) else 0 end as Diff,
      info.effectdate 
   from   
      (select 
          protype, product, pricecode, price, effectdate    
      from 
          prc_table     
      where 
          pricecode in ('x','y')    
          and protype = 'z') as info  
   left join 
      prc_table r1 on info.protype = r1.protype 
                   and info.product = r1.product 
                   and info.effectdate = r1.effectdate  
                   and r1.pricecode = 'x'  
   left join 
      prc_table r2 on info.protype = r2.protype 
                   and info.product = r2.product 
                   and info.effectdate = r2.effectdate  
                   and r2.pricecode = 'y'      
   where 
      info.effectdate >= DATEADD(MM, -3, GETDATE()) 
      and info.effectdate <= GETDATE()  
   group by     
      info.effectdate, info.protype, info.product,
      r1.product, r2.product, info.pricecode, r1.price, r2.price
)          
select   
   c.codedesc as [Grade],   
   r.product as [Product],  
   r.Region as [Region],  
   r.Price as [Price],
   r.Diff as [E-W],
   r.effectdate as [Date]
from 
   t r 
inner join 
   pro_item i on r.protype = i.protype and r.product = i.product   
inner join 
   pro_duct p on i.protype = p.protype and i.product = p.product and i.1 = p.1
(product grade join)
inner join 
   xxx_codes c on p.desc3 = c.code and c.prefix = 'xxx'    
where   
    i.protype = 'z'    
    and i.loc = 'loc'    
    and p.desc4 = ''    
    and i.branch = 'm'  
order by 
    r.effectdate desc, codedesc, product

Answer 1

I get what you're trying to do. And you're on the right track to calculate the differences in your sql, because it's not possible to show/hide the subtotals in the pivot table depending on whether it's east/west or just east. So doing it in sql is good. Where I think you have a bit of a design flaw in your query is trying to put the difference in a separate column.

The problem with that is, how do you present it in your pivot table? If you have both Price and Diff as your value field in your pivot table, it will botch the layout completely. You want only one value field - Price. And you want East, West, AND Diff to be your column headings, so they should all be Regions in your data.

So, in your sql, get the grade, product, region, price, and date like you're doing, put it in a temp table, then INSERT INTO that temp table the difference records, with region='Diff' and Price= the actual difference value, where both an east and a west record exist for that product and effective date.

Your output should look something like this:

Grade   Product Region  Price   Effdt                               
A       P1      East    1.5     31-Oct-14                                   
A       P2      East    3       31-Oct-14                                   
B       P3      East    5       31-Oct-14                                   
B       P4      East    2.44    31-Oct-14                                   
C       P5      East    3.67    31-Oct-14                                   
C       P6      East    10.3    31-Oct-14                                   
B       P3      West    5.5     31-Oct-14                                   
B       P4      West    2.7     31-Oct-14                                   
C       P5      West    3.8     31-Oct-14                                   
C       P6      West    10.7    31-Oct-14                                   
A       P1      East    1.5     30-Nov-14                                   
A       P2      East    3       30-Nov-14                                   
B       P3      East    5       30-Nov-14                                   
B       P4      East    2.44    30-Nov-14                                   
C       P5      East    3.67    30-Nov-14                                   
C       P6      East    10.3    30-Nov-14                                   
B       P3      Diff    -0.5    31-Oct-14                                   
B       P4      Diff    -0.26   31-Oct-14                                   
C       P5      Diff    -0.13   31-Oct-14                                   
C       P6      Diff    -0.4    31-Oct-14   

You only insert difference rows where applicable, so when you pivottable this data, it will show either east+west+diff or just east but not east+diff.

It looks like you're pretty comfortable with sql to write the above query so you should be able to implement that. If you are having trouble with it post where you get to and i'll have a look.

Cheers