问题
My table looks for example like this:
Name date result
A 2012-01-01 1
A 2012-02-01 2
B 2013-01-01 1
...
For a full example: http://sqlfiddle.com/#!3/0226b/1
At the moment I have a working query that counts the rows by person and year: http://sqlfiddle.com/#!3/0226b/3
This is perfect, but what I want is some extra information for 2014. i need to count how many rows I have for every result. something like this:
NAME 1 2 3 2014 2013 2012 TOTAL
Person B 4 0 2 6 2 2 10
Person A 2 1 1 4 3 4 11
Person C 1 1 1 3 1 0 4
Even better would be that I give the result-columns a good name (1 = lost, 2= draw, 3=won):
NAME lost draw won 2014 2013 2012 TOTAL
Person B 4 0 2 6 2 2 10
Person A 2 1 1 4 3 4 11
Person C 1 1 1 3 1 0 4
I tried to add some extra code, like:
select @colsResult
= STUFF((SELECT ',' + QUOTENAME(result)
from list
group by result
order by result
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
I have as result:
,[1]
,[2]
,[3]
But if I run the whole code I get an error, invallid column name...
回答1:
Since you have two columns that you now want to PIVOT, you'll first have to unpivot those columns and then convert those values into the new columns.
Starting in SQL Server 2005, you could use CROSS APPLY to unpivot the columns. The basic syntax will be similar to:
select
name,
new_col,
total
from
(
select name,
dt = year(date),
result,
total = count(*) over(partition by name)
from list
) d
cross apply
(
select 'dt', dt union all
select 'result', result
) c (old_col_name, new_col)
See SQL Fiddle with Demo. This query gets you a list of names, with the "new columns" and then the Total entries for each name.
| NAME | NEW_COL | TOTAL |
|----------|---------|-------|
| Person A | 2012 | 11 |
| Person A | 1 | 11 |
| Person A | 2012 | 11 |
| Person A | 2 | 11 |
You'll see that the dates and the results are now both stored in "new_col". These values will now be used as the new column names. If you have a limited number of columns, then you would simply hard-code the query:
select name, lost = [1],
draw=[2], won = [3],
[2014], [2013], [2012], Total
from
(
select
name,
new_col,
total
from
(
select name,
dt = year(date),
result,
total = count(*) over(partition by name)
from list
) d
cross apply
(
select 'dt', dt union all
select 'result', result
) c (old_col_name, new_col)
) src
pivot
(
count(new_col)
for new_col in([1], [2], [3], [2014], [2013], [2012])
) piv
order by [2014];
See SQL Fiddle with Demo
Now since your years are dynamic, then you'll need to use dynamic sql. But it appears that you have 3 results and potentially multiple years - so I'd use a combination of static/dynamic sql to make this easier:
DECLARE @cols AS NVARCHAR(MAX),
@query AS NVARCHAR(MAX),
@orderby nvarchar(max)
select @cols
= STUFF((SELECT ',' + QUOTENAME(year(date))
from list
group by year(date)
order by year(date) desc
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
select @orderby = 'ORDER BY ['+cast(year(getdate()) as varchar(4)) + '] desc'
set @query = 'SELECT name, lost = [1],
draw=[2], won = [3],' + @cols + ', Total
from
(
select
name,
new_col,
total
from
(
select name,
dt = year(date),
result,
total = count(*) over(partition by name)
from list
) d
cross apply
(
select ''dt'', dt union all
select ''result'', result
) c (old_col_name, new_col)
) x
pivot
(
count(new_col)
for new_col in ([1], [2], [3],' + @cols + ')
) p '+ @orderby
exec sp_executesql @query;
See SQL Fiddle with Demo. This gives a result:
| NAME | LOST | DRAW | WON | 2014 | 2013 | 2012 | TOTAL |
|----------|------|------|-----|------|------|------|-------|
| Person B | 7 | 1 | 2 | 6 | 2 | 2 | 10 |
| Person A | 5 | 3 | 3 | 4 | 3 | 4 | 11 |
| Person C | 2 | 1 | 1 | 3 | 1 | 0 | 4 |
If you want to only filter the result columns for the current year, then you can perform this filtering a variety of ways but the easiest you be to include a filter in the unpivot. The hard-coded version would be:
select name, lost = [1],
draw=[2], won = [3],
[2014], [2013], [2012], Total
from
(
select
name,
new_col,
total
from
(
select name,
dt = year(date),
result,
total = count(*) over(partition by name)
from list
) d
cross apply
(
select 'dt', dt union all
select 'result', case when dt = 2014 then result end
) c (old_col_name, new_col)
) src
pivot
(
count(new_col)
for new_col in([1], [2], [3], [2014], [2013], [2012])
) piv
order by [2014] desc;
See SQL Fiddle with Demo. Then the dynamic sql version would be:
DECLARE @cols AS NVARCHAR(MAX),
@query AS NVARCHAR(MAX),
@orderby nvarchar(max),
@currentYear varchar(4)
select @currentYear = cast(year(getdate()) as varchar(4))
select @cols
= STUFF((SELECT ',' + QUOTENAME(year(date))
from list
group by year(date)
order by year(date) desc
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
select @orderby = 'ORDER BY ['+ @currentYear + '] desc'
set @query = 'SELECT name, lost = [1],
draw=[2], won = [3],' + @cols + ', Total
from
(
select
name,
new_col,
total
from
(
select name,
dt = year(date),
result,
total = count(*) over(partition by name)
from list
) d
cross apply
(
select ''dt'', dt union all
select ''result'', case when dt = '+@currentYear+' then result end
) c (old_col_name, new_col)
) x
pivot
(
count(new_col)
for new_col in ([1], [2], [3],' + @cols + ')
) p '+ @orderby
exec sp_executesql @query;
See SQL Fiddle with Demo. This version will give a result:
| NAME | LOST | DRAW | WON | 2014 | 2013 | 2012 | TOTAL |
|----------|------|------|-----|------|------|------|-------|
| Person B | 4 | 0 | 2 | 6 | 2 | 2 | 10 |
| Person A | 2 | 1 | 1 | 4 | 3 | 4 | 11 |
| Person C | 1 | 1 | 1 | 3 | 1 | 0 | 4 |
来源：https://stackoverflow.com/questions/25784437/extend-current-query-calculated-columns