June 22, 2010 at 4:42 pm
Let me try to explain the situation a little better.
If a user #1 is connected from 1:00 PM - 6:00 PM
and user #2 is connected from 1:01 PM - 7:59 PM
and user #3 is connected from 3:00 PM - 8:00 PM
and user #4 is connected from 5:15 PM - 5:59 PM
my hourly totals would look like this...
Hour | Total Users Connected
12 | 0
13 | 2
14 | 2
15 | 3
16 | 3
17 | 4
18 | 3
19 | 2
20 | 1
21 | 0
I also just realized that the data that I furnished was a little off. The overlap will only ever occur with the start time, not the stop time. Notice record 683 starts on the 31st and ends on the 1st.
--===== Insert the test data into the test table
INSERT INTO #whosonwhen ([ID], [Start], [Stop])
SELECT '6876','2010-06-01 06:13:54.000','2010-06-01 16:00:26.000' UNION ALL
SELECT '6919','2010-06-01 07:56:59.000','2010-06-01 17:57:00.000' UNION ALL
SELECT '6863','2010-06-01 08:37:00.000','2010-06-01 15:30:52.000' UNION ALL
SELECT '6851','2010-06-01 08:18:00.000','2010-06-01 15:00:26.000' UNION ALL
SELECT '6852','2010-06-01 09:02:17.000','2010-06-01 15:00:29.000' UNION ALL
SELECT '683','2010-05-31 23:02:17.000','2010-06-01 02:00:29.000'
So to answer your questions Jeff...
Jeff Moden (6/22/2010)
Are you saying that you want an hourly count by day? And what do you want to show if you have a time span of more than 1 day?
I want an hourly count, but just the current day. The current day will always be the day of the stop time. I want to force the start time to be the current day if the start time is a previous day. Maybe just force it to be one second after midnight.
Thanks!
June 23, 2010 at 5:48 am
From, what I can see, the query I gave you does not return running totals but retuns the count of loggins for each hour for period of time during the same day.
Base on the last sample you gave, this query will return your expected results.
Can you please provide the test data setup to point out where the query does wrong.
Or, may be you wanted something else? For example the count of different users who logged into the system for each hour during the given period, so if the user logged into the system twice during the same hour, it does count as 1 not as 2?
Then the query must be modified to do the following (together with relevant test data setup):
--===== If the test table already exists, drop it
IF OBJECT_ID('TempDB..#whosonwhen','U') IS NOT NULL
DROP TABLE #whosonwhen
--===== Create the test table
CREATE TABLE #whosonwhen
(
[ID] INT,
[Start] DATETIME,
[Stop] DATETIME
)
--===== Insert the test data into the test table
INSERT INTO #whosonwhen ([ID], [Start], [Stop])
SELECT '6876','2010-06-01 13:00:00.000','2010-06-01 18:00:00.000' UNION ALL
SELECT '6876','2010-06-01 18:10:00.000','2010-06-01 18:20:00.000' UNION ALL -- user logged back during 18th hour of day
SELECT '6919','2010-06-01 13:01:00.000','2010-06-01 19:59:00.000' UNION ALL
SELECT '6863','2010-06-01 15:00:00.000','2010-06-01 20:00:00.000' UNION ALL
SELECT '6851','2010-06-01 17:15:00.000','2010-06-01 17:59:00.000'
declare @periodStart datetime
declare @periodEnd datetime
-- they must be inside of the same day, otherwise counting hours from 0 to 23 is not appropriate!!!!
set @periodStart = '2010-06-01 03:15:00.000'
set @periodEnd = '2010-06-01 21:20:00.000'
-- Given Period is taken as inclusive of given hours in the input (eg. 15:25:30 will be taken as 15:00:00)
-- frist make sure that the minutes, seconds and milliseconds are removed from input range for clarity
set @periodStart = dateadd(hh, datepart(hh,@periodStart), convert(varchar(12),@periodStart,112))
set @periodEnd = dateadd(hh, datepart(hh,@periodEnd), convert(varchar(12),@periodEnd,112))
-- you may create this CTE by other ways (eg. from permanent Tally table)...
;with dh
as
(
select top 24
DATEADD(hour,ROW_NUMBER() OVER (ORDER BY [Object_id])-1,convert(varchar(12),@periodStart,112)) as HoDstart
,DATEADD(hour,ROW_NUMBER() OVER (ORDER BY [Object_id]),convert(varchar(12),@periodStart,112)) as HoDend
,ROW_NUMBER() OVER (ORDER BY Object_id)-1 as DayHour
from sys.columns -- or any other (not very big) table which have more than 24 raws, just remamber to change
-- [Object_id] in OVER (ORDER BY [Object_id]... to some existing column
)
select d.DayHour, count(distinct w.ID) as RecCount -- basically, count only distinct users
from dh d
left join #whosonwhen w
on w.[Start] < d.HoDend
and w.[Stop] >= d.HoDstart
where d.HoDstart between @periodStart and @periodEnd
group by d.DayHour
order by d.DayHour
June 23, 2010 at 9:45 am
Here's my solution. This first part uses the test data you provided in your first post, with an additional row showing the startdate overlap.
--===== If the test table already exists, drop it
IF OBJECT_ID('TempDB..#whosonwhen','U') IS NOT NULL
DROP TABLE #whosonwhen
--===== Create the test table
CREATE TABLE #whosonwhen
(
[ID] INT,
[Start] DATETIME,
[Stop] DATETIME
)
--===== Insert the test data into the test table
INSERT INTO #whosonwhen ([ID], [Start], [Stop])
SELECT '6876','2010-06-01 06:13:54.000','2010-06-01 16:00:26.000' UNION ALL
SELECT '6919','2010-06-01 07:56:59.000','2010-06-01 17:57:00.000' UNION ALL
SELECT '6863','2010-06-01 08:37:00.000','2010-06-01 15:30:52.000' UNION ALL
SELECT '6851','2010-06-01 08:18:00.000','2010-06-01 15:00:26.000' UNION ALL
SELECT '6852','2010-06-01 09:02:17.000','2010-06-01 15:00:29.000' UNION ALL
SELECT '683','2010-06-01 20:02:17.000','2010-06-02 02:00:29.000' UNION ALL
SELECT '6831','2010-05-31 23:02:17.000','2010-06-01 02:00:29.000'
Now, my solution:
declare @starttime datetime
set @starttime = DATEADD(day, DateDiff(day, 0, '20100601'), 0)
;WITH
-- Use the small, built-in tally table in the master db.
-- Use it to build a tally table of hours,
-- with the corresponding StartTime and EndTime for that hour.
-- (EndTimes will be used with a "<", so it's actually the start of the next day.)
TallyHour AS (SELECT [Hour] = number,
StartTime = DATEADD(hour, number, @starttime),
EndTime = DATEADD(hour, number+1, @StartTime)
FROM master.dbo.spt_values
WHERE number < 24
AND [type] = 'P'),
-- This subquery gets a count by hour of entries in #whosonwhen
Sub1 AS (select [Hour], Qty = COUNT(*)
from #whosonwhen t
JOIN TallyHour th
ON (t.Start >= th.StartTime AND t.Start < th.EndTime)
OR th.StartTime between t.Start and t.Stop
GROUP BY [Hour])
-- get entries by hour
SELECT [Hour], Qty
FROM Sub1
UNION ALL
-- combine it with the remaining hours with qty = 0
SELECT [Hour], 0
FROM TallyHour
EXCEPT
SELECT [Hour], 0
FROM Sub1
Note that this handles overlaps, be they from the previous day, or into the next day.
Wayne
Microsoft Certified Master: SQL Server 2008
Author - SQL Server T-SQL Recipes
July 14, 2015 at 1:38 pm
What about something like this?
SELECT month(DateField) as DteFld_Month, DAY(DateField) AS DteFld_Day, { fn HOUR(DateField) } AS DteFld_Hour, COUNT(MeasMyID) AS DupCount
FROM MyTable
WHERE (DateField >= CONVERT(DATETIME, '2015-07-13 00:00:00', 102))
GROUP BY month(DateField), DAY(DateField), { fn HOUR(DateField) }
order by DteFld_Month, DteFld_Day, DteFld_Hour
July 14, 2015 at 7:56 pm
daveb-840383 (6/22/2010)
I want an hourly count, but just the current day. The current day will always be the day of the stop time. I want to force the start time to be the current day if the start time is a previous day. Maybe just force it to be one second after midnight.
To be honest, I think you're incorrectly trying to fit an answer to what you think is the correct method. As you properly stated in your original post...
I have a record with start time and stop time. I want to determine an hourly count of records that exist during the time span.
The bottom line is that you want to know how many "records" are active in at any given hour no matter when they logged in or out.
I have to take a look at some of the other's code to make sure I'm not setting up to a duplicate effort. If I'm not, I'll be back.
--Jeff Moden
Change is inevitable... Change for the better is not.
July 14, 2015 at 8:00 pm
laudena (7/14/2015)
What about something like this?
SELECT month(DateField) as DteFld_Month, DAY(DateField) AS DteFld_Day, { fn HOUR(DateField) } AS DteFld_Hour, COUNT(MeasMyID) AS DupCount
FROM MyTable
WHERE (DateField >= CONVERT(DATETIME, '2015-07-13 00:00:00', 102))
GROUP BY month(DateField), DAY(DateField), { fn HOUR(DateField) }
order by DteFld_Month, DteFld_Day, DteFld_Hour
I guess my first question would be, why do you feel it necessary to use ODBC temporal functions?
--Jeff Moden
Change is inevitable... Change for the better is not.
July 14, 2015 at 8:10 pm
Sorry... made a pretty bad mistake. Post withdrawn.
--Jeff Moden
Change is inevitable... Change for the better is not.
July 17, 2015 at 12:31 pm
laudena (7/14/2015)
What about something like this?
SELECT month(DateField) as DteFld_Month, DAY(DateField) AS DteFld_Day, { fn HOUR(DateField) } AS DteFld_Hour, COUNT(MeasMyID) AS DupCount
FROM MyTable
WHERE (DateField >= CONVERT(DATETIME, '2015-07-13 00:00:00', 102))
GROUP BY month(DateField), DAY(DateField), { fn HOUR(DateField) }
order by DteFld_Month, DteFld_Day, DteFld_Hour
You do realize that this thread is over FIVE YEARS OLD. I hope that the OP has resolved this issue in the intervening five years, and the OP has almost certainly forgotten about this issue in the intervening five years.
Drew
J. Drew Allen
Business Intelligence Analyst
Philadelphia, PA
July 18, 2015 at 9:56 am
drew.allen (7/17/2015)
laudena (7/14/2015)
What about something like this?
SELECT month(DateField) as DteFld_Month, DAY(DateField) AS DteFld_Day, { fn HOUR(DateField) } AS DteFld_Hour, COUNT(MeasMyID) AS DupCount
FROM MyTable
WHERE (DateField >= CONVERT(DATETIME, '2015-07-13 00:00:00', 102))
GROUP BY month(DateField), DAY(DateField), { fn HOUR(DateField) }
order by DteFld_Month, DteFld_Day, DteFld_Hour
You do realize that this thread is over FIVE YEARS OLD. I hope that the OP has resolved this issue in the intervening five years, and the OP has almost certainly forgotten about this issue in the intervening five years.
Drew
Why does everyone think that a post is only for the OP? Five years old or not, additional information never hurts.
--Jeff Moden
Change is inevitable... Change for the better is not.
July 20, 2015 at 10:21 am
Jeff Moden (7/18/2015)
Why does everyone think that a post is only for the OP? Five years old or not, additional information never hurts.
The vast majority of the time, people respond to old posts due to sloppiness in checking dates, and this sloppiness tends to be reflected in their responses as well.
Additional information may never hurt, but it doesn't always help, either.
Drew
J. Drew Allen
Business Intelligence Analyst
Philadelphia, PA
July 20, 2015 at 10:56 am
drew.allen (7/20/2015)
Jeff Moden (7/18/2015)
Why does everyone think that a post is only for the OP? Five years old or not, additional information never hurts.The vast majority of the time, people respond to old posts due to sloppiness in checking dates, and this sloppiness tends to be reflected in their responses as well.
Additional information may never hurt, but it doesn't always help, either.
Drew
That would sometimes be true even on current posts. 😉 Citing the fact that a post is old would fall into that category for me.
--Jeff Moden
Change is inevitable... Change for the better is not.
July 20, 2015 at 12:17 pm
daveb-840383 (6/22/2010)
Let me try to explain the situation a little better.If a user #1 is connected from 1:00 PM - 6:00 PM
and user #2 is connected from 1:01 PM - 7:59 PM
and user #3 is connected from 3:00 PM - 8:00 PM
and user #4 is connected from 5:15 PM - 5:59 PM
my hourly totals would look like this...
Hour | Total Users Connected
12 | 0
13 | 2
14 | 2
15 | 3
16 | 3
17 | 4
18 | 3
19 | 2
20 | 1
21 | 0
I also just realized that the data that I furnished was a little off. The overlap will only ever occur with the start time, not the stop time. Notice record 683 starts on the 31st and ends on the 1st.
--===== Insert the test data into the test table
INSERT INTO #whosonwhen ([ID], [Start], [Stop])
SELECT '6876','2010-06-01 06:13:54.000','2010-06-01 16:00:26.000' UNION ALL
SELECT '6919','2010-06-01 07:56:59.000','2010-06-01 17:57:00.000' UNION ALL
SELECT '6863','2010-06-01 08:37:00.000','2010-06-01 15:30:52.000' UNION ALL
SELECT '6851','2010-06-01 08:18:00.000','2010-06-01 15:00:26.000' UNION ALL
SELECT '6852','2010-06-01 09:02:17.000','2010-06-01 15:00:29.000' UNION ALL
SELECT '683','2010-05-31 23:02:17.000','2010-06-01 02:00:29.000'
So to answer your questions Jeff...
Jeff Moden (6/22/2010)
Are you saying that you want an hourly count by day? And what do you want to show if you have a time span of more than 1 day?
I want an hourly count, but just the current day. The current day will always be the day of the stop time. I want to force the start time to be the current day if the start time is a previous day. Maybe just force it to be one second after midnight.
Thanks!
Apologies for the delay but you might find it's worth it when the requirements change sometime in the future.
First, let's build a handy little tool to make various aspects of coding a bit easier. The following code builds one of the many renditions of an on-the-fly "Tally" function.
CREATE FUNCTION [dbo].[fnTally]
/**********************************************************************************************************************
Purpose:
Return a column of BIGINTs from @ZeroOrOne up to and including @MaxN with a max value of 1 Trillion.
As a performance note, it takes about 00:02:10 (hh:mm:ss) to generate 1 Billion numbers to a throw-away variable.
Usage:
--===== Syntax example (Returns BIGINT)
SELECT t.N
FROM dbo.fnTally(@ZeroOrOne,@MaxN) t
;
Notes:
1. Based on Itzik Ben-Gan's cascading CTE (cCTE) method for creating a "readless" Tally Table source of BIGINTs.
Refer to the following URLs for how it works and introduction for how it replaces certain loops.
http://www.sqlservercentral.com/articles/T-SQL/62867/
http://sqlmag.com/sql-server/virtual-auxiliary-table-numbers
2. To start a sequence at 0, @ZeroOrOne must be 0 or NULL. Any other value that's convertable to the BIT data-type
will cause the sequence to start at 1.
3. If @ZeroOrOne = 1 and @MaxN = 0, no rows will be returned.
5. If @MaxN is negative or NULL, a "TOP" error will be returned.
6. @MaxN must be a positive number from >= the value of @ZeroOrOne up to and including 1 Billion. If a larger
number is used, the function will silently truncate after 1 Billion. If you actually need a sequence with
that many values, you should consider using a different tool. ;-)
7. There will be a substantial reduction in performance if "N" is sorted in descending order. If a descending
sort is required, use code similar to the following. Performance will decrease by about 27% but it's still
very fast especially compared with just doing a simple descending sort on "N", which is about 20 times slower.
If @ZeroOrOne is a 0, in this case, remove the "+1" from the code.
DECLARE @MaxN BIGINT;
SELECT @MaxN = 1000;
SELECT DescendingN = @MaxN-N+1
FROM dbo.fnTally(1,@MaxN);
8. There is no performance penalty for sorting "N" in ascending order because the output is explicity sorted by
ROW_NUMBER() OVER (ORDER BY (SELECT NULL))
Revision History:
Rev 00 - Unknown - Jeff Moden
- Initial creation with error handling for @MaxN.
Rev 01 - 09 Feb 2013 - Jeff Moden
- Modified to start at 0 or 1.
Rev 02 - 16 May 2013 - Jeff Moden
- Removed error handling for @MaxN because of exceptional cases.
Rev 03 - 22 Apr 2015 - Jeff Moden
- Modify to handle 1 Trillion rows for experimental purposes.
**********************************************************************************************************************/
(@ZeroOrOne BIT, @MaxN BIGINT)
RETURNS TABLE WITH SCHEMABINDING AS
RETURN WITH
E1(N) AS (SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL
SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL
SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL
SELECT 1) --10E1 or 10 rows
, E4(N) AS (SELECT 1 FROM E1 a, E1 b, E1 c, E1 d) --10E4 or 10 Thousand rows
,E12(N) AS (SELECT 1 FROM E4 a, E4 b, E4 c) --10E12 or 1 Trillion rows
SELECT N = 0 WHERE ISNULL(@ZeroOrOne,0)= 0 --Conditionally start at 0.
UNION ALL
SELECT TOP(@MaxN) N = ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) FROM E12 -- Values from 1 to @MaxN
;
Ok... now, let's test against a more substantial amount of data. The following code will build a million row test table according to what I perceive from your posts. You can also use this to test other folks' code for performance, as well. It uses the "fnTally" function as a pseudo-cursor to avoid the slothfulness of a loop or recursive CTE to do the same.
Details are in the code in the form of comments to explain certain assumptions on my part and to explain indexes that either are already on your table or should be.
--===== If it already exists, drop the test table to make reruns easier in SSMS
IF OBJECT_ID('tempdb..#WhosOnWhen','U') IS NOT NULL
DROP TABLE #WhosOnWhen
;
GO
--===== Create a million rows of data where the START date/time is somewhere in 2014 or 2015.
-- The Stop datetime will be up to and not including 3 days later than the Start datetime.
-- Don't let the million rows scare you. This takes just several seconds.
WITH
cteStartDates AS
(
SELECT TOP 1000000
Start = ISNULL(RAND(CHECKSUM(NEWID()))*DATEDIFF(dd,'2014','2016')+CAST('2014' AS DATETIME),0)
FROM dbo.fnTally(1,1000000) t
)
SELECT ID = ISNULL(ROW_NUMBER()OVER(ORDER BY Start),0) --ID is realistically in order by Start and is not null
,Start
,Stop = Start + RAND(CHECKSUM(NEWID()))*3 --Stop is a random amount of up to 3 days time after the start
INTO #WhosOnWhen
FROM cteStartDates
;
--===== Add the expected Clustered Index and the expected PK.
-- In this case, they are different. Note how I guaranteed the CI to be unique.
CREATE UNIQUE CLUSTERED INDEX CI_#WhosOnWhen
ON #WhosOnWhen (Start,ID)
;
ALTER TABLE #WhosOnWhen
ADD PRIMARY KEY NONCLUSTERED (ID)
;
Here's one fairly high performance method for solving your problem for a single day.
--===== Setup variables for the day range.
-- The dates may be changed to include
-- many more days, if required. They
-- will also resolve to only whole days
-- even if a time is included.
DECLARE @pStartDate DATETIME
,@pStopDate DATETIME
;
SELECT @pStartDate = '12 Jul 2015'
,@pStopDate = '12 Jul 2015'
;
--===== This solution uses a high speed on-the-fly expansion of dates.
WITH
cteParms AS
(
SELECT StartDate = DATEDIFF(dd, 0,@pStartDate) --Ensure that we start on a whole date
,StopDate = DATEDIFF(dd,-1,@pStopDate) --Ensure that we end on next on a whole date
)
,cteExpandHours AS
( --=== Expand all date ranges by whole hour
SELECT HourPeriodStart = DATEADD(hh,DATEDIFF(hh,0,DATEADD(hh,t.N,d.Start)),0)
FROM #WhosOnWhen d
CROSS APPLY dbo.fnTally(0,DATEDIFF(hh,d.Start,d.Stop)) t
CROSS JOIN cteParms p
WHERE d.Stop >= p.StartDate --These relations correctly picks up any overlaps
AND d.Start < p.StopDate
)
SELECT HourPeriodStart
,HourPeriod = DATEPART(hh,HourPeriodStart)
,CountPerHour = COUNT(*)
FROM cteExpandHours
CROSS JOIN cteParms p
WHERE HourPeriodStart >= p.StartDate --Criteria repeated to limit output of report
AND HourPeriodStart < p.StopDate
GROUP BY HourPeriodStart
ORDER BY HourPeriodStart
;
And here are the results from one of the runs. Keep in mind that the test data is randomly constructed and your counts will vary because of that.
HourPeriodStart HourPeriod CountPerHour
----------------------- ----------- ------------
2015-07-12 00:00:00.000 0 2021
2015-07-12 01:00:00.000 1 2025
2015-07-12 02:00:00.000 2 2028
2015-07-12 03:00:00.000 3 2040
2015-07-12 04:00:00.000 4 2036
2015-07-12 05:00:00.000 5 2027
2015-07-12 06:00:00.000 6 2023
2015-07-12 07:00:00.000 7 2029
2015-07-12 08:00:00.000 8 2048
2015-07-12 09:00:00.000 9 2052
2015-07-12 10:00:00.000 10 2073
2015-07-12 11:00:00.000 11 2067
2015-07-12 12:00:00.000 12 2078
2015-07-12 13:00:00.000 13 2088
2015-07-12 14:00:00.000 14 2083
2015-07-12 15:00:00.000 15 2084
2015-07-12 16:00:00.000 16 2074
2015-07-12 17:00:00.000 17 2058
2015-07-12 18:00:00.000 18 2058
2015-07-12 19:00:00.000 19 2063
2015-07-12 20:00:00.000 20 2069
2015-07-12 21:00:00.000 21 2070
2015-07-12 22:00:00.000 22 2063
2015-07-12 23:00:00.000 23 2069
(24 row(s) affected)
And, yes... I absolutely realize that the OP may no longer be paying attention to a 5 year old post but someone finding this code might also find it useful. 😉
--Jeff Moden
Change is inevitable... Change for the better is not.
November 30, 2016 at 9:32 am
Thank you Jeff. Excellent!
This solution really helped me at my current issue.
Thank you again!
Viewing 13 posts - 16 through 27 (of 27 total)
You must be logged in to reply to this topic. Login to reply