Related
I have a query where I need to show 24 hour calls for each day.
But I am getting the hours which I have calls only.
My requirement is I need to get all the hours split and 0 if there are no calls.
Please suggest
Below is my code.
select #TrendStartDate
,isd.Name
,isd.Call_ID
,isd.callType
,DATEPART(HOUR,isd.ArrivalTime)
from [PHONE_CALLS] ISD WITH (NOLOCK)
WHERE CallType = 'Incoming'
and Name not in ('DefaultQueue')
and CAST(ArrivalTime as DATe) between #TrendStartDate and #TrendEndDate
The basic idea is that you use a table containing numbers from 0 to 23, and left join that to your data table:
WITH CTE AS
(
SELECT TOP 24 ROW_NUMBER() OVER(ORDER BY ##SPID) - 1 As TheHour
FROM sys.objects
)
SELECT #TrendStartDate
,isd.Name
,isd.Call_ID
,isd.callType
,TheHour
FROM CTE
LEFT JOIN [PHONE_CALLS] ISD WITH (NOLOCK)
ON DATEPART(HOUR,isd.ArrivalTime) = TheHour
AND CallType = 'Incoming'
AND Name NOT IN ('DefaultQueue')
AND CAST(ArrivalTime as DATe) BETWEEN #TrendStartDate AND #TrendEndDate
If you have a tally table, you should use that. If not, the cte will provide you with numbers from 0 to 23.
If you have a numbers table you can use a query like the following:
SELECT d.Date,
h.Hour,
Calls = COUNT(pc.Call_ID)
FROM ( SELECT [Hour] = Number
FROM dbo.Numbers
WHERE Number >= 0
AND Number < 24
) AS h
CROSS JOIN
( SELECT Date = DATEADD(DAY, Number, #TrendStartDate)
FROM dbo.Numbers
WHERE Number <= DATEDIFF(DAY, #TrendStartDate, #TrendEndDate)
) AS d
LEFT JOIN [PHONE_CALLS] AS pc
ON pc.CallType = 'Incoming'
AND pc.Name NOT IN ('DefaultQueue')
AND CAST(pc.ArrivalTime AS DATE) = d.Date
AND DATEPART(HOUR, pc.ArrivalTime) = h.Hour
GROUP BY d.Date, h.Hour
ORDER BY d.Date, h.Hour;
The key is to get all the hours you need:
SELECT [Hour] = Number
FROM dbo.Numbers
WHERE Number >= 0
AND Number < 24
And all the days that you need in your range:
SELECT Date = DATEADD(DAY, Number, #TrendStartDate)
FROM dbo.Numbers
WHERE Number < DATEDIFF(DAY, #TrendStartDate, #TrendEndDate)
Then cross join the two, so that you are guaranteed to have all 24 hours for each day you want. Finally, you can left join to your call table to get the count of calls.
Example on DB<>Fiddle
You can use SQL SERVER recursivity with CTE to generate the hours between 0 and 23 and then a left outer join with the call table
You also use any other Method mentioned in this link to generate numbers from 0 to 23
Link to SQLFiddle
set dateformat ymd
declare #calls as table(date date,hour int,calls int)
insert into #calls values('2020-01-02',0,66),('2020-01-02',1,888),
('2020-01-02',2,5),('2020-01-02',3,8),
('2020-01-02',4,9),('2020-01-02',5,55),('2020-01-02',6,44),('2020-01-02',7,87),('2020-01-02',8,90),
('2020-01-02',9,34),('2020-01-02',10,22),('2020-01-02',11,65),('2020-01-02',12,54),('2020-01-02',13,78),
('2020-01-02',23,99);
with cte as (select 0 n,date from #calls union all select 1+n,date from cte where 1+n <24)
select distinct(cte.date),cte.n [Hour],isnull(ca.calls,0) calls from cte left outer join #calls ca on cte.n=ca.hour and cte.date=ca.date
I have a table which has the following columns: DeskID *, ProductID *, Date *, Amount (where the columns marked with * make the primary key). The products in use vary over time, as represented in the image below.
Table format on the left, and a (hopefully) intuitive representation of the data on the right for one desk
The objective is to have the sum of the latest amounts of products by desk and date, including products which are no longer in use, over a date range.
e.g. using the data above the desired table is:
So on the 1st Jan, the sum is 1 of Product A
On the 2nd Jan, the sum is 2 of A and 5 of B, so 7
On the 4th Jan, the sum is 1 of A (out of use, so take the value from the 3rd), 5 of B, and 2 of C, so 8 in total
etc.
I have tried using a partition on the desk and product ordered by date to get the most recent value and turned the following code into a function (Function1 below) with #date Date parameter
select #date 'Date', t.DeskID, SUM(t.Amount) 'Sum' from (
select #date 'Date', t.DeskID, t.ProductID, t.Amount
, row_number() over (partition by t.DeskID, t.ProductID order by t.Date desc) as roworder
from Table1 t
where 1 = 1
and t.Date <= #date
) t
where t.roworder = 1
group by t.DeskID
And then using a utility calendar table and cross apply to get the required values over a time range, as below
select * from Calendar c
cross apply Function1(c.CalendarDate)
where c.CalendarDate >= '20190101' and c.CalendarDate <= '20191009'
This has the expected results, but is far too slow. Currently each desk uses around 50 products, and the products roll every month, so after just 5 years each desk has a history of ~3000 products, which causes the whole thing to grind to a halt. (Roughly 30 seconds for a range of a single month)
Is there a better approach?
Change your function to the following should be faster:
select #date 'Date', t.DeskID, SUM(t.Amount) 'Sum'
FROM (SELECT m.DeskID, m.ProductID, MAX(m.[Date) AS MaxDate
FROM Table1 m
where m.[Date] <= #date) d
INNER JOIN Table1 t
ON d.DeskID=t.DeskID
AND d.ProductID=t.ProductID
and t.[Date] = d.MaxDate
group by t.DeskID
The performance of TVF usually suffers. The following removes the TVF completely:
-- DROP TABLE Table1;
CREATE TABLE Table1 (DeskID int not null, ProductID nvarchar(32) not null, [Date] Date not null, Amount int not null, PRIMARY KEY ([Date],DeskID,ProductID));
INSERT Table1(DeskID,ProductID,[Date],Amount)
VALUES (1,'A','2019-01-01',1),(1,'A','2019-01-02',2),(1,'B','2019-01-02',5),(1,'A','2019-01-03',1)
,(1,'B','2019-01-03',4),(1,'C','2019-01-03',3),(1,'B','2019-01-04',5),(1,'C','2019-01-04',2),(1,'C','2019-01-05',2)
GO
DECLARE #StartDate date=N'2019-01-01';
DECLARE #EndDate date=N'2019-01-05';
;WITH cte_p
AS
(
SELECT DISTINCT DeskID,ProductID
FROM Table1
WHERE [Date] <= #EndDate
),
cte_a
AS
(
SELECT #StartDate AS [Date], p.DeskID, p.ProductID, ISNULL(a.Amount,0) AS Amount
FROM (
SELECT t.DeskID, t.ProductID
, MAX(t.Date) AS FirstDate
FROM Table1 t
WHERE t.Date <= #StartDate
GROUP BY t.DeskID, t.ProductID) f
INNER JOIN Table1 a
ON f.DeskID=a.DeskID
AND f.ProductID=a.ProductID
AND f.[FirstDate]=a.[Date]
RIGHT JOIN cte_p p
ON p.DeskID=a.DeskID
AND p.ProductID=a.ProductID
UNION ALL
SELECT DATEADD(DAY,1,a.[Date]) AS [Date], t.DeskID, t.ProductID, t.Amount
FROM Table1 t
INNER JOIN cte_a a
ON t.DeskID=a.DeskID
AND t.ProductID=a.ProductID
AND t.[Date] > a.[Date]
AND t.[Date] <= DATEADD(DAY,1,a.[Date])
WHERE a.[Date]<#EndDate
UNION ALL
SELECT DATEADD(DAY,1,a.[Date]) AS [Date], a.DeskID, a.ProductID, a.Amount
FROM cte_a a
WHERE NOT EXISTS(SELECT 1 FROM Table1 t
WHERE t.DeskID=a.DeskID
AND t.ProductID=a.ProductID
AND t.[Date] > a.[Date]
AND t.[Date] <= DATEADD(DAY,1,a.[Date]))
AND a.[Date]<#EndDate
)
SELECT [Date], DeskID, SUM(Amount)
FROM cte_a
GROUP BY [Date], DeskID;
Initial Question
Given the following dataset paired with a dates table:
MembershipId | ValidFromDate | ValidToDate
==========================================
0001 | 1997-01-01 | 2006-05-09
0002 | 1997-01-01 | 2017-05-12
0003 | 2005-06-02 | 2009-02-07
How many Memberships were open on any given day or timeseries of days?
Initial Answer
Following this question being asked here, this answer provided the necessary functionality:
select d.[Date]
,count(m.MembershipID) as MembershipCount
from DIM.[Date] as d
left join Memberships as m
on(d.[Date] between m.ValidFromDateKey and m.ValidToDateKey)
where d.CalendarYear = 2016
group by d.[Date]
order by d.[Date];
though a commenter remarked that There are other approaches when the non-equijoin takes too long.
Followup
As such, what would the equijoin only logic look like to replicate the output of the query above?
Progress So Far
From the answers provided so far I have come up with the below, which outperforms on the hardware I am working with across 3.2 million Membership records:
declare #s date = '20160101';
declare #e date = getdate();
with s as
(
select d.[Date] as d
,count(s.MembershipID) as s
from dbo.Dates as d
join dbo.Memberships as s
on d.[Date] = s.ValidFromDateKey
group by d.[Date]
)
,e as
(
select d.[Date] as d
,count(e.MembershipID) as e
from dbo.Dates as d
join dbo.Memberships as e
on d.[Date] = e.ValidToDateKey
group by d.[Date]
),c as
(
select isnull(s.d,e.d) as d
,sum(isnull(s.s,0) - isnull(e.e,0)) over (order by isnull(s.d,e.d)) as c
from s
full join e
on s.d = e.d
)
select d.[Date]
,c.c
from dbo.Dates as d
left join c
on d.[Date] = c.d
where d.[Date] between #s and #e
order by d.[Date]
;
Following on from that, to split this aggregate into constituent groups per day I have the following, which is also performing well:
declare #s date = '20160101';
declare #e date = getdate();
with s as
(
select d.[Date] as d
,s.MembershipGrouping as g
,count(s.MembershipID) as s
from dbo.Dates as d
join dbo.Memberships as s
on d.[Date] = s.ValidFromDateKey
group by d.[Date]
,s.MembershipGrouping
)
,e as
(
select d.[Date] as d
,e..MembershipGrouping as g
,count(e.MembershipID) as e
from dbo.Dates as d
join dbo.Memberships as e
on d.[Date] = e.ValidToDateKey
group by d.[Date]
,e.MembershipGrouping
),c as
(
select isnull(s.d,e.d) as d
,isnull(s.g,e.g) as g
,sum(isnull(s.s,0) - isnull(e.e,0)) over (partition by isnull(s.g,e.g) order by isnull(s.d,e.d)) as c
from s
full join e
on s.d = e.d
and s.g = e.g
)
select d.[Date]
,c.g
,c.c
from dbo.Dates as d
left join c
on d.[Date] = c.d
where d.[Date] between #s and #e
order by d.[Date]
,c.g
;
Can anyone improve on the above?
If most of your membership validity intervals are longer than few days, have a look at an answer by Martin Smith. That approach is likely to be faster.
When you take calendar table (DIM.[Date]) and left join it with Memberships, you may end up scanning the Memberships table for each date of the range. Even if there is an index on (ValidFromDate, ValidToDate), it may not be super useful.
It is easy to turn it around.
Scan the Memberships table only once and for each membership find those dates that are valid using CROSS APPLY.
Sample data
DECLARE #T TABLE (MembershipId int, ValidFromDate date, ValidToDate date);
INSERT INTO #T VALUES
(1, '1997-01-01', '2006-05-09'),
(2, '1997-01-01', '2017-05-12'),
(3, '2005-06-02', '2009-02-07');
DECLARE #RangeFrom date = '2006-01-01';
DECLARE #RangeTo date = '2006-12-31';
Query 1
SELECT
CA.dt
,COUNT(*) AS MembershipCount
FROM
#T AS Memberships
CROSS APPLY
(
SELECT dbo.Calendar.dt
FROM dbo.Calendar
WHERE
dbo.Calendar.dt >= Memberships.ValidFromDate
AND dbo.Calendar.dt <= Memberships.ValidToDate
AND dbo.Calendar.dt >= #RangeFrom
AND dbo.Calendar.dt <= #RangeTo
) AS CA
GROUP BY
CA.dt
ORDER BY
CA.dt
OPTION(RECOMPILE);
OPTION(RECOMPILE) is not really needed, I include it in all queries when I compare execution plans to be sure that I'm getting the latest plan when I play with the queries.
When I looked at the plan of this query I saw that the seek in the Calendar.dt table was using only ValidFromDate and ValidToDate, the #RangeFrom and #RangeTo were pushed to the residue predicate. It is not ideal. The optimiser is not smart enough to calculate maximum of two dates (ValidFromDate and #RangeFrom) and use that date as a starting point of the seek.
It is easy to help the optimiser:
Query 2
SELECT
CA.dt
,COUNT(*) AS MembershipCount
FROM
#T AS Memberships
CROSS APPLY
(
SELECT dbo.Calendar.dt
FROM dbo.Calendar
WHERE
dbo.Calendar.dt >=
CASE WHEN Memberships.ValidFromDate > #RangeFrom
THEN Memberships.ValidFromDate
ELSE #RangeFrom END
AND dbo.Calendar.dt <=
CASE WHEN Memberships.ValidToDate < #RangeTo
THEN Memberships.ValidToDate
ELSE #RangeTo END
) AS CA
GROUP BY
CA.dt
ORDER BY
CA.dt
OPTION(RECOMPILE)
;
In this query the seek is optimal and doesn't read dates that may be discarded later.
Finally, you may not need to scan the whole Memberships table.
We need only those rows where the given range of dates intersects with the valid range of the membership.
Query 3
SELECT
CA.dt
,COUNT(*) AS MembershipCount
FROM
#T AS Memberships
CROSS APPLY
(
SELECT dbo.Calendar.dt
FROM dbo.Calendar
WHERE
dbo.Calendar.dt >=
CASE WHEN Memberships.ValidFromDate > #RangeFrom
THEN Memberships.ValidFromDate
ELSE #RangeFrom END
AND dbo.Calendar.dt <=
CASE WHEN Memberships.ValidToDate < #RangeTo
THEN Memberships.ValidToDate
ELSE #RangeTo END
) AS CA
WHERE
Memberships.ValidToDate >= #RangeFrom
AND Memberships.ValidFromDate <= #RangeTo
GROUP BY
CA.dt
ORDER BY
CA.dt
OPTION(RECOMPILE)
;
Two intervals [a1;a2] and [b1;b2] intersect when
a2 >= b1 and a1 <= b2
These queries assume that Calendar table has an index on dt.
You should try and see what indexes are better for the Memberships table.
For the last query, if the table is rather large, most likely two separate indexes on ValidFromDate and on ValidToDate would be better than one index on (ValidFromDate, ValidToDate).
You should try different queries and measure their performance on the real hardware with real data. Performance may depend on the data distribution, how many memberships there are, what are their valid dates, how wide or narrow is the given range, etc.
I recommend to use a great tool called SQL Sentry Plan Explorer to analyse and compare execution plans. It is free. It shows a lot of useful stats, such as execution time and number of reads for each query. The screenshots above are from this tool.
On the assumption your date dimension contains all dates contained in all membership periods you can use something like the following.
The join is an equi join so can use hash join or merge join not just nested loops (which will execute the inside sub tree once for each outer row).
Assuming index on (ValidToDate) include(ValidFromDate) or reverse this can use a single seek against Memberships and a single scan of the date dimension. The below has an elapsed time of less than a second for me to return the results for a year against a table with 3.2 million members and general active membership of 1.4 million (script)
DECLARE #StartDate DATE = '2016-01-01',
#EndDate DATE = '2016-12-31';
WITH MD
AS (SELECT Date,
SUM(Adj) AS MemberDelta
FROM Memberships
CROSS APPLY (VALUES ( ValidFromDate, +1),
--Membership count decremented day after the ValidToDate
(DATEADD(DAY, 1, ValidToDate), -1) ) V(Date, Adj)
WHERE
--Members already expired before the time range of interest can be ignored
ValidToDate >= #StartDate
AND
--Members whose membership starts after the time range of interest can be ignored
ValidFromDate <= #EndDate
GROUP BY Date),
MC
AS (SELECT DD.DateKey,
SUM(MemberDelta) OVER (ORDER BY DD.DateKey ROWS UNBOUNDED PRECEDING) AS CountOfNonIgnoredMembers
FROM DIM_DATE DD
LEFT JOIN MD
ON MD.Date = DD.DateKey)
SELECT DateKey,
CountOfNonIgnoredMembers AS MembershipCount
FROM MC
WHERE DateKey BETWEEN #StartDate AND #EndDate
ORDER BY DateKey
Demo (uses extended period as the calendar year of 2016 isn't very interesting with the example data)
One approach is to first use an INNER JOIN to find the set of matches and COUNT() to project MemberCount GROUPed BY DateKey, then UNION ALL with the same set of dates, with a 0 on that projection for the count of members for each date. The last step is to SUM() the MemberCount of this union, and GROUP BY DateKey. As requested, this avoids LEFT JOIN and NOT EXISTS. As another member pointed out, this is not an equi-join, because we need to use a range, but I think it does what you intend.
This will serve up 1 year's worth of data with around 100k logical reads. On an ordinary laptop with a spinning disk, from cold cache, it serves 1 month in under a second (with correct counts).
Here is an example that creates 3.3 million rows of random duration. The query at the bottom returns one month's worth of data.
--Stay quiet for a moment
SET NOCOUNT ON
SET STATISTICS IO OFF
SET STATISTICS TIME OFF
--Clean up if re-running
DROP TABLE IF EXISTS DIM_DATE
DROP TABLE IF EXISTS FACT_MEMBER
--Date dimension
CREATE TABLE DIM_DATE
(
DateKey DATE NOT NULL
)
--Membership fact
CREATE TABLE FACT_MEMBER
(
MembershipId INT NOT NULL
, ValidFromDateKey DATE NOT NULL
, ValidToDateKey DATE NOT NULL
)
--Populate Date dimension from 2001 through end of 2018
DECLARE #startDate DATE = '2001-01-01'
DECLARE #endDate DATE = '2018-12-31'
;WITH CTE_DATE AS
(
SELECT #startDate AS DateKey
UNION ALL
SELECT
DATEADD(DAY, 1, DateKey)
FROM
CTE_DATE AS D
WHERE
D.DateKey < #endDate
)
INSERT INTO
DIM_DATE
(
DateKey
)
SELECT
D.DateKey
FROM
CTE_DATE AS D
OPTION (MAXRECURSION 32767)
--Populate Membership fact with members having a random membership length from 1 to 36 months
;WITH CTE_DATE AS
(
SELECT #startDate AS DateKey
UNION ALL
SELECT
DATEADD(DAY, 1, DateKey)
FROM
CTE_DATE AS D
WHERE
D.DateKey < #endDate
)
,CTE_MEMBER AS
(
SELECT 1 AS MembershipId
UNION ALL
SELECT MembershipId + 1 FROM CTE_MEMBER WHERE MembershipId < 500
)
,
CTE_MEMBERSHIP
AS
(
SELECT
ROW_NUMBER() OVER (ORDER BY NEWID()) AS MembershipId
, D.DateKey AS ValidFromDateKey
FROM
CTE_DATE AS D
CROSS JOIN CTE_MEMBER AS M
)
INSERT INTO
FACT_MEMBER
(
MembershipId
, ValidFromDateKey
, ValidToDateKey
)
SELECT
M.MembershipId
, M.ValidFromDateKey
, DATEADD(MONTH, FLOOR(RAND(CHECKSUM(NEWID())) * (36-1)+1), M.ValidFromDateKey) AS ValidToDateKey
FROM
CTE_MEMBERSHIP AS M
OPTION (MAXRECURSION 32767)
--Add clustered Primary Key to Date dimension
ALTER TABLE DIM_DATE ADD CONSTRAINT PK_DATE PRIMARY KEY CLUSTERED
(
DateKey ASC
)
--Index
--(Optimize in your spare time)
DROP INDEX IF EXISTS SK_FACT_MEMBER ON FACT_MEMBER
CREATE CLUSTERED INDEX SK_FACT_MEMBER ON FACT_MEMBER
(
ValidFromDateKey ASC
, ValidToDateKey ASC
, MembershipId ASC
)
RETURN
--Start test
--Emit stats
SET STATISTICS IO ON
SET STATISTICS TIME ON
--Establish range of dates
DECLARE
#rangeStartDate DATE = '2010-01-01'
, #rangeEndDate DATE = '2010-01-31'
--UNION the count of members for a specific date range with the "zero" set for the same range, and SUM() the counts
;WITH CTE_MEMBER
AS
(
SELECT
D.DateKey
, COUNT(*) AS MembershipCount
FROM
DIM_DATE AS D
INNER JOIN FACT_MEMBER AS M ON
M.ValidFromDateKey <= #rangeEndDate
AND M.ValidToDateKey >= #rangeStartDate
AND D.DateKey BETWEEN M.ValidFromDateKey AND M.ValidToDateKey
WHERE
D.DateKey BETWEEN #rangeStartDate AND #rangeEndDate
GROUP BY
D.DateKey
UNION ALL
SELECT
D.DateKey
, 0 AS MembershipCount
FROM
DIM_DATE AS D
WHERE
D.DateKey BETWEEN #rangeStartDate AND #rangeEndDate
)
SELECT
M.DateKey
, SUM(M.MembershipCount) AS MembershipCount
FROM
CTE_MEMBER AS M
GROUP BY
M.DateKey
ORDER BY
M.DateKey ASC
OPTION (RECOMPILE, MAXDOP 1)
Here's how I'd solve this problem with equijoin:
--data generation
declare #Membership table (MembershipId varchar(10), ValidFromDate date, ValidToDate date)
insert into #Membership values
('0001', '1997-01-01', '2006-05-09'),
('0002', '1997-01-01', '2017-05-12'),
('0003', '2005-06-02', '2009-02-07')
declare #startDate date, #endDate date
select #startDate = MIN(ValidFromDate), #endDate = max(ValidToDate) from #Membership
--in order to use equijoin I need all days between min date and max date from Membership table (both columns)
;with cte as (
select #startDate [date]
union all
select DATEADD(day, 1, [date]) from cte
where [date] < #endDate
)
--in this query, we will assign value to each day:
--one, if project started on that day
--minus one, if project ended on that day
--then, it's enough to (cumulative) sum all this values to get how many projects were ongoing on particular day
select [date],
sum(case when [DATE] = ValidFromDate then 1 else 0 end +
case when [DATE] = ValidToDate then -1 else 0 end)
over (order by [date] rows between unbounded preceding and current row)
from cte [c]
left join #Membership [m]
on [c].[date] = [m].ValidFromDate or [c].[date] = [m].ValidToDate
option (maxrecursion 0)
Here's another solution:
--data generation
declare #Membership table (MembershipId varchar(10), ValidFromDate date, ValidToDate date)
insert into #Membership values
('0001', '1997-01-01', '2006-05-09'),
('0002', '1997-01-01', '2017-05-12'),
('0003', '2005-06-02', '2009-02-07')
;with cte as (
select CAST('2016-01-01' as date) [date]
union all
select DATEADD(day, 1, [date]) from cte
where [date] < '2016-12-31'
)
select [date],
(select COUNT(*) from #Membership where ValidFromDate < [date]) -
(select COUNT(*) from #Membership where ValidToDate < [date]) [ongoing]
from cte
option (maxrecursion 0)
Pay attention, I think #PittsburghDBA is right when it says that current query return wrong result.
The last day of membership is not counted and so final sum is lower than it should be.
I have corrected it in this version.
This should improve a bit your actual progress:
declare #s date = '20160101';
declare #e date = getdate();
with
x as (
select d, sum(c) c
from (
select ValidFromDateKey d, count(MembershipID) c
from Memberships
group by ValidFromDateKey
union all
-- dateadd needed to count last day of membership too!!
select dateadd(dd, 1, ValidToDateKey) d, -count(MembershipID) c
from Memberships
group by ValidToDateKey
)x
group by d
),
c as
(
select d, sum(x.c) over (order by d) as c
from x
)
select d.day, c cnt
from calendar d
left join c on d.day = c.d
where d.day between #s and #e
order by d.day;
First of all, your query yields '1' as MembershipCount even if no active membership exists for the given date.
You should return SUM(CASE WHEN m.MembershipID IS NOT NULL THEN 1 ELSE 0 END) AS MembershipCount.
For optimal performance create an index on Memberships(ValidFromDateKey, ValidToDateKey, MembershipId) and another on DIM.[Date](CalendarYear, DateKey).
With that done, the optimal query shall be:
DECLARE #CalendarYear INT = 2000
SELECT dim.DateKey, SUM(CASE WHEN con.MembershipID IS NOT NULL THEN 1 ELSE 0 END) AS MembershipCount
FROM
DIM.[Date] dim
LEFT OUTER JOIN (
SELECT ValidFromDateKey, ValidToDateKey, MembershipID
FROM Memberships
WHERE
ValidFromDateKey <= CONVERT(DATETIME, CONVERT(VARCHAR, #CalendarYear) + '1231')
AND ValidToDateKey >= CONVERT(DATETIME, CONVERT(VARCHAR, #CalendarYear) + '0101')
) con
ON dim.DateKey BETWEEN con.ValidFromDateKey AND con.ValidToDateKey
WHERE dim.CalendarYear = #CalendarYear
GROUP BY dim.DateKey
ORDER BY dim.DateKey
Now, for your last question, what would be the equijoin equivalent query.
There is NO WAY you can rewrite this as a non-equijoin!
Equijoin doesn't imply using join sintax. Equijoin implies using an equals predicate, whatever the sintax.
Your query yields a range comparison, hence equals doesn't apply: a between or similar is required.
I have hospital patient admission data in Microsoft SQL Server r2 that looks something like this:
PatientID, AdmitDate, DischargeDate
Jones. 1-jan-13 01:37. 1-jan-13 17:45
Smith 1-jan-13 02:12. 2-jan-13 02:14
Brooks. 4-jan-13 13:54. 5-jan-13 06:14
I would like count the number of patients in the hospital day by day and hour by hour (ie at
1-jan-13 00:00. 0
1-jan-13 01:00. 0
1-jan-13 02:00. 1
1-jan-13 03:00. 2
And I need to include the hours when there are no patients admitted in the result.
I can't create tables so making a reference table listing all the hours and days is out, though.
Any suggestions?
To solve this problem, you need a list of date-hours. The following gets this from the admit date cross joined to a table with 24 hours. The table of 24 hours is calculating from information_schema.columns -- a trick for getting small sequences of numbers in SQL Server.
The rest is just a join between this table and the hours. This version counts the patients at the hour, so someone admitted and discharged in the same hour, for instance is not counted. And in general someone is not counted until the next hour after they are admitted:
with dh as (
select DATEADD(hour, seqnum - 1, thedatehour ) as DateHour
from (select distinct cast(cast(AdmitDate as DATE) as datetime) as thedatehour
from Admission a
) a cross join
(select ROW_NUMBER() over (order by (select NULL)) as seqnum
from INFORMATION_SCHEMA.COLUMNS
) hours
where hours <= 24
)
select dh.DateHour, COUNT(*) as NumPatients
from dh join
Admissions a
on dh.DateHour between a.AdmitDate and a.DischargeDate
group by dh.DateHour
order by 1
This also assumes that there are admissions on every day. That seems like a reasonable assumption. If not, a calendar table would be a big help.
Here is one (ugly) way:
;WITH DayHours AS
(
SELECT 0 DayHour
UNION ALL
SELECT DayHour+1
FROM DayHours
WHERE DayHour+1 <= 23
)
SELECT B.AdmitDate, A.DayHour, COUNT(DISTINCT PatientID) Patients
FROM DayHours A
CROSS JOIN (SELECT DISTINCT CONVERT(DATE,AdmitDate) AdmitDate
FROM YourTable) B
LEFT JOIN YourTable C
ON B.AdmitDate = CONVERT(DATE,C.AdmitDate)
AND A.DayHour = DATEPART(HOUR,C.AdmitDate)
GROUP BY B.AdmitDate, A.DayHour
This is a bit messy and includes a temp table with the test data you provided but
CREATE TABLE #HospitalPatientData (PatientId NVARCHAR(MAX), AdmitDate DATETIME, DischargeDate DATETIME)
INSERT INTO #HospitalPatientData
SELECT 'Jones.', '1-jan-13 01:37:00.000', '1-jan-13 17:45:00.000' UNION
SELECT 'Smith', '1-jan-13 02:12:00.000', '2-jan-13 02:14:00.000' UNION
SELECT 'Brooks.', '4-jan-13 13:54:00.000', '5-jan-13 06:14:00.000'
;WITH DayHours AS
(
SELECT 0 DayHour
UNION ALL
SELECT DayHour+1
FROM DayHours
WHERE DayHour+1 <= 23
),
HospitalPatientData AS
(
SELECT CONVERT(nvarchar(max),AdmitDate,103) as AdmitDate ,DATEPART(hour,(AdmitDate)) as AdmitHour, COUNT(PatientID) as CountOfPatients
FROM #HospitalPatientData
GROUP BY CONVERT(nvarchar(max),AdmitDate,103), DATEPART(hour,(AdmitDate))
),
Results AS
(
SELECT MAX(h.AdmitDate) as Date, d.DayHour
FROM HospitalPatientData h
INNER JOIN DayHours d ON d.DayHour=d.DayHour
GROUP BY AdmitDate, CountOfPatients, DayHour
)
SELECT r.*, COUNT(h.PatientId) as CountOfPatients
FROM Results r
LEFT JOIN #HospitalPatientData h ON CONVERT(nvarchar(max),AdmitDate,103)=r.Date AND DATEPART(HOUR,h.AdmitDate)=r.DayHour
GROUP BY r.Date, r.DayHour
ORDER BY r.Date, r.DayHour
DROP TABLE #HospitalPatientData
This may get you started:
BEGIN TRAN
DECLARE #pt TABLE
(
PatientID VARCHAR(10)
, AdmitDate DATETIME
, DischargeDate DATETIME
)
INSERT INTO #pt
( PatientID, AdmitDate, DischargeDate )
VALUES ( 'Jones', '1-jan-13 01:37', '1-jan-13 17:45' ),
( 'Smith', '1-jan-13 02:12', '2-jan-13 02:14' )
, ( 'Brooks', '4-jan-13 13:54', '5-jan-13 06:14' )
DECLARE #StartDate DATETIME = '20130101'
, #FutureDays INT = 7
;
WITH dy
AS ( SELECT TOP (#FutureDays)
ROW_NUMBER() OVER ( ORDER BY name ) dy
FROM sys.columns c
) ,
hr
AS ( SELECT TOP 24
ROW_NUMBER() OVER ( ORDER BY name ) hr
FROM sys.columns c
)
SELECT refDate, COUNT(p.PatientID) AS PtCount
FROM ( SELECT DATEADD(HOUR, hr.hr - 1,
DATEADD(DAY, dy.dy - 1, #StartDate)) AS refDate
FROM dy
CROSS JOIN hr
) ref
LEFT JOIN #pt p ON ref.refDate BETWEEN p.AdmitDate AND p.DischargeDate
GROUP BY refDate
ORDER BY refDate
ROLLBACK
How do you create a moving average in SQL?
Current table:
Date Clicks
2012-05-01 2,230
2012-05-02 3,150
2012-05-03 5,520
2012-05-04 1,330
2012-05-05 2,260
2012-05-06 3,540
2012-05-07 2,330
Desired table or output:
Date Clicks 3 day Moving Average
2012-05-01 2,230
2012-05-02 3,150
2012-05-03 5,520 4,360
2012-05-04 1,330 3,330
2012-05-05 2,260 3,120
2012-05-06 3,540 3,320
2012-05-07 2,330 3,010
This is an Evergreen Joe Celko question.
I ignore which DBMS platform is used. But in any case Joe was able to answer more than 10 years ago with standard SQL.
Joe Celko SQL Puzzles and Answers citation:
"That last update attempt suggests that we could use the predicate to
construct a query that would give us a moving average:"
SELECT S1.sample_time, AVG(S2.load) AS avg_prev_hour_load
FROM Samples AS S1, Samples AS S2
WHERE S2.sample_time
BETWEEN (S1.sample_time - INTERVAL 1 HOUR)
AND S1.sample_time
GROUP BY S1.sample_time;
Is the extra column or the query approach better? The query is
technically better because the UPDATE approach will denormalize the
database. However, if the historical data being recorded is not going
to change and computing the moving average is expensive, you might
consider using the column approach.
MS SQL Example:
CREATE TABLE #TestDW
( Date1 datetime,
LoadValue Numeric(13,6)
);
INSERT INTO #TestDW VALUES('2012-06-09' , '3.540' );
INSERT INTO #TestDW VALUES('2012-06-08' , '2.260' );
INSERT INTO #TestDW VALUES('2012-06-07' , '1.330' );
INSERT INTO #TestDW VALUES('2012-06-06' , '5.520' );
INSERT INTO #TestDW VALUES('2012-06-05' , '3.150' );
INSERT INTO #TestDW VALUES('2012-06-04' , '2.230' );
SQL Puzzle query:
SELECT S1.date1, AVG(S2.LoadValue) AS avg_prev_3_days
FROM #TestDW AS S1, #TestDW AS S2
WHERE S2.date1
BETWEEN DATEADD(d, -2, S1.date1 )
AND S1.date1
GROUP BY S1.date1
order by 1;
One way to do this is to join on the same table a few times.
select
(Current.Clicks
+ isnull(P1.Clicks, 0)
+ isnull(P2.Clicks, 0)
+ isnull(P3.Clicks, 0)) / 4 as MovingAvg3
from
MyTable as Current
left join MyTable as P1 on P1.Date = DateAdd(day, -1, Current.Date)
left join MyTable as P2 on P2.Date = DateAdd(day, -2, Current.Date)
left join MyTable as P3 on P3.Date = DateAdd(day, -3, Current.Date)
Adjust the DateAdd component of the ON-Clauses to match whether you want your moving average to be strictly from the past-through-now or days-ago through days-ahead.
This works nicely for situations where you need a moving average over only a few data points.
This is not an optimal solution for moving averages with more than a few data points.
select t2.date, round(sum(ct.clicks)/3) as avg_clicks
from
(select date from clickstable) as t2,
(select date, clicks from clickstable) as ct
where datediff(t2.date, ct.date) between 0 and 2
group by t2.date
Example here.
Obviously you can change the interval to whatever you need. You could also use count() instead of a magic number to make it easier to change, but that will also slow it down.
General template for rolling averages that scales well for large data sets
WITH moving_avg AS (
SELECT 0 AS [lag] UNION ALL
SELECT 1 AS [lag] UNION ALL
SELECT 2 AS [lag] UNION ALL
SELECT 3 AS [lag] --ETC
)
SELECT
DATEADD(day,[lag],[date]) AS [reference_date],
[otherkey1],[otherkey2],[otherkey3],
AVG([value1]) AS [avg_value1],
AVG([value2]) AS [avg_value2]
FROM [data_table]
CROSS JOIN moving_avg
GROUP BY [otherkey1],[otherkey2],[otherkey3],DATEADD(day,[lag],[date])
ORDER BY [otherkey1],[otherkey2],[otherkey3],[reference_date];
And for weighted rolling averages:
WITH weighted_avg AS (
SELECT 0 AS [lag], 1.0 AS [weight] UNION ALL
SELECT 1 AS [lag], 0.6 AS [weight] UNION ALL
SELECT 2 AS [lag], 0.3 AS [weight] UNION ALL
SELECT 3 AS [lag], 0.1 AS [weight] --ETC
)
SELECT
DATEADD(day,[lag],[date]) AS [reference_date],
[otherkey1],[otherkey2],[otherkey3],
AVG([value1] * [weight]) / AVG([weight]) AS [wavg_value1],
AVG([value2] * [weight]) / AVG([weight]) AS [wavg_value2]
FROM [data_table]
CROSS JOIN weighted_avg
GROUP BY [otherkey1],[otherkey2],[otherkey3],DATEADD(day,[lag],[date])
ORDER BY [otherkey1],[otherkey2],[otherkey3],[reference_date];
select *
, (select avg(c2.clicks) from #clicks_table c2
where c2.date between dateadd(dd, -2, c1.date) and c1.date) mov_avg
from #clicks_table c1
Use a different join predicate:
SELECT current.date
,avg(periods.clicks)
FROM current left outer join current as periods
ON current.date BETWEEN dateadd(d,-2, periods.date) AND periods.date
GROUP BY current.date HAVING COUNT(*) >= 3
The having statement will prevent any dates without at least N values from being returned.
assume x is the value to be averaged and xDate is the date value:
SELECT avg(x) from myTable WHERE xDate BETWEEN dateadd(d, -2, xDate) and xDate
In hive, maybe you could try
select date, clicks, avg(clicks) over (order by date rows between 2 preceding and current row) as moving_avg from clicktable;
For the purpose, I'd like to create an auxiliary/dimensional date table like
create table date_dim(date date, date_1 date, dates_2 date, dates_3 dates ...)
while date is the key, date_1 for this day, date_2 contains this day and the day before; date_3...
Then you can do the equal join in hive.
Using a view like:
select date, date from date_dim
union all
select date, date_add(date, -1) from date_dim
union all
select date, date_add(date, -2) from date_dim
union all
select date, date_add(date, -3) from date_dim
NOTE: THIS IS NOT AN ANSWER but an enhanced code sample of Diego Scaravaggi's answer. I am posting it as answer as the comment section is insufficient. Note that I have parameter-ized the period for Moving aveage.
declare #p int = 3
declare #t table(d int, bal float)
insert into #t values
(1,94),
(2,99),
(3,76),
(4,74),
(5,48),
(6,55),
(7,90),
(8,77),
(9,16),
(10,19),
(11,66),
(12,47)
select a.d, avg(b.bal)
from
#t a
left join #t b on b.d between a.d-(#p-1) and a.d
group by a.d
--#p1 is period of moving average, #01 is offset
declare #p1 as int
declare #o1 as int
set #p1 = 5;
set #o1 = 3;
with np as(
select *, rank() over(partition by cmdty, tenor order by markdt) as r
from p_prices p1
where
1=1
)
, x1 as (
select s1.*, avg(s2.val) as avgval from np s1
inner join np s2
on s1.cmdty = s2.cmdty and s1.tenor = s2.tenor
and s2.r between s1.r - (#p1 - 1) - (#o1) and s1.r - (#o1)
group by s1.cmdty, s1.tenor, s1.markdt, s1.val, s1.r
)
I'm not sure that your expected result (output) shows classic "simple moving (rolling) average" for 3 days. Because, for example, the first triple of numbers by definition gives:
ThreeDaysMovingAverage = (2.230 + 3.150 + 5.520) / 3 = 3.6333333
but you expect 4.360 and it's confusing.
Nevertheless, I suggest the following solution, which uses window-function AVG. This approach is much more efficient (clear and less resource-intensive) than SELF-JOIN introduced in other answers (and I'm surprised that no one has given a better solution).
-- Oracle-SQL dialect
with
data_table as (
select date '2012-05-01' AS dt, 2.230 AS clicks from dual union all
select date '2012-05-02' AS dt, 3.150 AS clicks from dual union all
select date '2012-05-03' AS dt, 5.520 AS clicks from dual union all
select date '2012-05-04' AS dt, 1.330 AS clicks from dual union all
select date '2012-05-05' AS dt, 2.260 AS clicks from dual union all
select date '2012-05-06' AS dt, 3.540 AS clicks from dual union all
select date '2012-05-07' AS dt, 2.330 AS clicks from dual
),
param as (select 3 days from dual)
select
dt AS "Date",
clicks AS "Clicks",
case when rownum >= p.days then
avg(clicks) over (order by dt
rows between p.days - 1 preceding and current row)
end
AS "3 day Moving Average"
from data_table t, param p;
You see that AVG is wrapped with case when rownum >= p.days then to force NULLs in first rows, where "3 day Moving Average" is meaningless.
We can apply Joe Celko's "dirty" left outer join method (as cited above by Diego Scaravaggi) to answer the question as it was asked.
declare #ClicksTable table ([Date] date, Clicks int)
insert into #ClicksTable
select '2012-05-01', 2230 union all
select '2012-05-02', 3150 union all
select '2012-05-03', 5520 union all
select '2012-05-04', 1330 union all
select '2012-05-05', 2260 union all
select '2012-05-06', 3540 union all
select '2012-05-07', 2330
This query:
SELECT
T1.[Date],
T1.Clicks,
-- AVG ignores NULL values so we have to explicitly NULLify
-- the days when we don't have a full 3-day sample
CASE WHEN count(T2.[Date]) < 3 THEN NULL
ELSE AVG(T2.Clicks)
END AS [3-Day Moving Average]
FROM #ClicksTable T1
LEFT OUTER JOIN #ClicksTable T2
ON T2.[Date] BETWEEN DATEADD(d, -2, T1.[Date]) AND T1.[Date]
GROUP BY T1.[Date]
Generates the requested output:
Date Clicks 3-Day Moving Average
2012-05-01 2,230
2012-05-02 3,150
2012-05-03 5,520 4,360
2012-05-04 1,330 3,330
2012-05-05 2,260 3,120
2012-05-06 3,540 3,320
2012-05-07 2,330 3,010