问题
I am trying to build a SQL query which will give me the date range for the dates with same prices. If there is a break in the prices, I expect to see it in a new line. Even if sometime during the month there are same prices, if there is change in the prices sometime in between I want to see it as two separate rows with the specific date range.
Sample Data:
Date Price
1-Jan 3.2
2-Jan 3.2
3-Jan 3.2
4-Jan 3.2
5-Jan 3.2
6-Jan 3.2
7-Jan 3.2
8-Jan 3.2
9-Jan 3.5
10-Jan 3.5
11-Jan 3.5
12-Jan 3.5
13-Jan 3.5
14-Jan 4.2
15-Jan 4.2
16-Jan 4.2
17-Jan 3.2
18-Jan 3.2
19-Jan 3.2
20-Jan 3.2
21-Jan 3.2
22-Jan 3
23-Jan 3
24-Jan 3
25-Jan 3
26-Jan 3
27-Jan 3
28-Jan 3
29-Jan 3.5
30-Jan 3.5
31-Jan 3.5
Desired Result :
Price Date Range
3.2 1-8
3.5 9-13
4.2 14-16
3.2 17-22
3 22-28
3.5 29-31
回答1:
Non-relational Solution
I don't think any of other answers are correct.
GROUP BYwon't workUsing
ROW_NUMBER()forces the data into a Record Filing System structure, which is physical, and then processes it as physical records. At a massive performance cost. Of course, in order to write such code, it forces you to think in terms of RFS instead of thinking in Relational terms.Using CTEs is the same. Iterating through the data, especially data that does not change. At a slightly different massive cost.
Cursors are definitely the wrong thing for a different set of reasons. (a) Cursors require code, and you have requested a View (b) Cursors abandon the set-processing engine, and revert to row-by-row processing. Again, not required. If a developer on any of my teams uses cursors or temp tables on a Relational Database (ie. not a Record Filing System), I shoot them.
Relational Solution
Your data is Relational, logical, the two given data columns are all that is necessary.
Sure, we have to form a View (derived Relation), to obtain the desired report, but that consists of pure SELECTs, which is quite different to processing (converting it to a file, which is physical, and then processing the file; or temp tables; or worktables; or CTEs; or ROW_Number(); etc).
Contrary to the lamentations of "theoreticians", who have an agenda, SQL handles Relational data perfectly well. And you data is Relational.
Therefore, maintain a Relational mindset, a Relational view of the data, and a set-processing mentality. Every report requirement over a Relational Database can be fulfilled using a single SELECT. There is no need to regress to pre-1970 ISAM File handling methods.
I will assume the Primary Key (the set of columns that give a Relational row uniqueness) is Date, and based on the example data given, the Datatype is DATE.
Try this:
CREATE VIEW MyTable_Base_V -- Foundation View
AS
SELECT Date,
Date_Next,
Price
FROM (
-- Derived Table: project rows with what we need
SELECT Date,
[Date_Next] = DATEADD( DD, 1, O.Date ),
Price,
[Price_Next] = (
SELECT Price -- NULL if not exists
FROM MyTable
WHERE Date = DATEADD( DD, 1, O.Date )
)
FROM MyTable MT
) AS X
WHERE Price != Price_Next -- exclude unchanging rows
GO
CREATE VIEW MyTable_V -- Requested View
AS
SELECT [Date_From] = (
-- Date of the previous row
SELECT MAX( Date_Next ) -- previous row
FROM MyTable_V
WHERE Date_Next < MT.Date
),
[Date_To] = Date, -- this row
Price
FROM MyTable_Base_V MT
GO
SELECT *
FROM MyTable_V
GO
Method, Generic
Of course this is a method, therefore it is generic, it can be used to determine the From_ and To_ of any data range (here, a Date range), based on any data change (here, a change in Price).
Here, your Dates are consecutive, so the determination of Date_Next is simple: increment the Date by 1 day. If the PK is increasing but not consecutive (eg. DateTime or TimeStamp or some other Key), change the Derived Table X to:
-- Derived Table: project rows with what we need
SELECT DateTime,
[DateTime_Next] = (
-- first row > this row
SELECT TOP 1
DateTime -- NULL if not exists
FROM MyTable
WHERE DateTime > MT.DateTime
),
Price,
[Price_Next] = (
-- first row > this row
SELECT TOP 1
Price -- NULL if not exists
FROM MyTable
WHERE DateTime > MT.DateTime
)
FROM MyTable MT
Enjoy.
Please feel free to comment, ask questions, etc.
回答2:
You can do this by adding a grouping column. A neat trick for this is the difference of two sequences of numbers -- when the difference is constant, then the price is the same.
select price, min(date), max(date)
from (select s.*,
(row_number() over (order by date) -
row_number() over (partition by price order by date)
) as grp
from sample s
) grp
group by grp, price;
Note: be careful that price is stored as a fixed decimal rather than a floating decimal. Otherwise, values that look the same might not actually be the same.
回答3:
This is what you are looking for
declare @temptbl table (price decimal(18,2), mindate date, maxdate date)
declare @price as decimal(18,2), @date as date
declare tempcur cursor for
select price, date
from YourTable
open tempcur
fetch next from tempcur
into @price, @date
while (@@fetch_status = 0)
begin
if (isnull((select price from @temptbl where maxdate = (select max(maxdate)from @temptbl)),0) <> @price)
insert into @temptbl (price,mindate,maxdate) values (@price,@date,@date)
else
update @temptbl
set maxdate = @date
where maxdate = (select max(maxdate)from @temptbl)
fetch next from tempcur
into @price, @date
end
deallocate tempcur
select price, convert(nvarchar(50), mindate) + ' to ' + convert(nvarchar(50), maxdate) as [date range] from @temptbl
回答4:
Use CTE, below is working code.
WITH grouped AS (
SELECT
Pricedate, price,
grp1= ROW_NUMBER() OVER (ORDER BY Pricedate) -
ROW_NUMBER() OVER (Partition by price ORDER BY Pricedate)
FROM yourTablewithDateAndPrice
)
SELECT
DtFrom = MIN(Pricedate),
DtTo = MAX(Pricedate),
Price = price
FROM grouped
GROUP BY Price,grp1
order by DtFrom;
The internal query will created same group till the time it find same price, else group will be incremented by one. in Final group by you will have required result.
来源：https://stackoverflow.com/questions/30455227/date-range-for-set-of-same-data