How to do conditional select using SQL Server 2005 - sql-server

I have following query that I want to improve so it displays two additional columns as Variance and Percentage_Variance.
declare #w0s datetime
declare #w0e datetime
declare #w1s datetime
declare #w1e datetime
set #w0s = dateadd(hh, datediff(hh, 0, getdate()), 0)
set #w0e = getdate()
set #w1s = dateadd(dd, datediff(dd, 0, getdate()) / 7 * 7 - 7, 0)
set #w1e = dateadd(dd, datediff(dd, 0, getdate()) / 7 * 7, 0)
select
state,
sum(case
when create_time >= #w0s and create_time < #w0e
then 1 else 0
end) as Call_Volume_This_hr,
sum(case
when create_time >= #w1s and create_time < #w1e
then 1 else 0
end) / 7.0 as Avg_Call_Volume_Past_Week1
--,(case when ((sum(case when create_time>=#w0s and create_time<#w0e then 1 else 0 end)) - (sum(case when create_time>=#w1s and create_time<#w1e then 1 else 0 end)/7.0) > 0)then 1 else 0 end) as Variance
--,(case when (((case when ((sum(case when create_time>=#w0s and create_time<#w0e then 1 else 0 end)) - (sum(case when create_time>=#w1s and create_time<#w1e then 1 else 0 end)/7.0) > 0)then 1 else 0 end)/(sum(case when create_time>=#w1s and create_time<#w1e then 1 else 0 end)/7.0)*100.0) > 0) then 1 else 0 end) as Percentage_Variance
from
ctirpts.dbo.cti_reporting
where
create_time >= #w1s
and datepart(hh, create_time) = datepart(hour, getdate())
and state is not null
group by
state
;
The condition is that
if (Call_Volume_This_hr > Avg_Call_Volume_Past_Week1)
then
Variance = Call_Volume_This_hr - Avg_Call_Volume_Past_Week1
and
Percentage_Variance = (Variance/Avg_Call_Volume_Past_Week1)*100
else
Variance = 0 and Percentage_Variance = 0
The two commented lines shows my feeble attempt at getting this done but it prints 1 under columns "Variance" and "Percentage_Variance" when "if (Call_Volume_This_hr > Avg_Call_Volume_Past_Week1)" is true. Instead I want the actual computed values as per:
Variance = Call_Volume_This_hr - Avg_Call_Volume_Past_Week1
Percentage_Variance = (Variance/Avg_Call_Volume_Past_Week1)*100
I would appreciate any help or pointers.

You could put your initial select into a common table expression, then those calculated SUM columns become much more readily available for conditional computation. You can use an APPLY operator to perform the CASE statement, returning Variance of zero or more (never negatives). Then you can perform the calculation to get the percentage variance.
-- this CTE performs the initial aggregation and makes the recordset available
-- to be used like a table:
;with PreAggregate
as (
select state
,sum(case when create_time>=#w0s and create_time<#w0e then 1 else 0 end) as Call_Volume_This_hr
,sum(case when create_time>=#w1s and create_time<#w1e then 1 else 0 end)/7.0 as Avg_Call_Volume_Past_Week1
from ctirpts.dbo.cti_reporting
where create_time>=#w1s
and datepart(hh,create_time)=datepart(hour, getdate())
and state is not null
group by state
)
-- The APPLY operator creates an inline-table-value function without the
-- "black box" query optimizer problems of using an actual UDF
-- The "CROSS APPLY" calculates a 0 or more value for Variance.
-- The Variance value can then be used to compute Percentage_Variance.
select PreAggregate.state
, PreAggregate.Call_Volume_This_hr
, PreAggregate.Avg_Call_Volume_Past_Week1
, InlineFunction.Variance
, (InlineFunction.Variance/PreAggregate.Avg_Call_Volume_Past_Week1)*100 as 'Percentage_Variance'
from PreAggregate
cross apply (
select case
when Call_Volume_This_hr < Avg_Call_Volume_Past_Week1 then 0
else Call_Volume_This_hr -Avg_Call_Volume_Past_Week1
end 'Variance'
) InlineFunction;

I think your two commented lines should be as follows:
( ( SUM(CASE WHEN create_time >= #w0s
AND create_time < #w0e THEN 1
ELSE 0
END) ) - ( SUM(CASE WHEN create_time >= #w1s
AND create_time < #w1e THEN 1
ELSE 0
END) / 7.0 ) ) AS Variance,
( ( CASE WHEN ( ( SUM(CASE WHEN create_time >= #w0s
AND create_time < #w0e THEN 1
ELSE 0
END) ) - ( SUM(CASE WHEN create_time >= #w1s
AND create_time < #w1e THEN 1
ELSE 0
END) / 7.0 ) > 0 ) THEN 1
ELSE 0
END ) / ( SUM(CASE WHEN create_time >= #w1s
AND create_time < #w1e THEN 1
ELSE 0
END) / 7.0 ) * 100.0 ) AS Percentage_Variance
However without a fiddle or some structure and data I was unable to test it beyond correct syntax. Essentially the outermost CASE statements where the problem. They needed to be removed.

Related

How to get total and percentage of values more than 10 in SQL Server

I have written a SQL statement which yields Days taken and count of how many of each.
What I’d like to accomplish is a counts and percentages of how above and below 10 days there are.
Below sample data and what I'd like to achieve in colored texts.
Your help will be much appreciated thanks heaps.
If you just want the last two days, you can use:
select (case when days_take <= 10 then '10 or less' else 'greater than 10' end) as grp,
count(*) * 100.0 / sum(count(*)) over () as percentage,
count(*) as ratio
from t
group by (case when days_take <= 10 then '10 or less' else 'greater than 10' end);
In TSQL, you can use cross apply and aggregation:
You can do conditional aggregation:
select
x.descr,
1.0 * sum(x.how_many) / sum(sum(x.how_many)) over() as how_many_ratio,
sum(x.how_many) as how_many_value
from mytable t
cross apply (values
(
'greater than 10 days'
case when days_taken > 10 then how_many else 0 end
),
(
'less than and 10 days'
case when days_taken <= 10 then how_many else 0 end
)
) as x(descr, how_many)
group by x.descr
If you are content with having all results on a single row, conditional aggregation is simpler:
select
1.0 * sum(case when days_taken > 10 then how_many else 0 end)
/ sum(how_many) as how_many_above_10_ratio,
sum(case when days_taken > 10 then how_many else 0 end) as how_many_above_10,
1.0 * sum(case when days_taken <= 10 then how_many else 0 end)
/ sum(how_many) as how_many_below_10_ratio,
sum(case when days_taken <= 10 then how_many else 0 end) as how_many_below_10
from mytable

SQL Server 2014 - Sum Working Hour group by Day/Night Time, Week/Weekend, Regular/Overtime

I have a list of tasks,
for each I have a TaskID, startTime and StopTime as milliseconds from 1-1-1970 and a list of users (#Tasks).
I need to calculate the time spent by each user on the task splitted by day/night time, week or weekend, regular/overtime considering nighttime hours from 10:00 PM to 06:00 AM.
Surely there's a better solution but so far I got this:
IF OBJECT_ID('tempdb..#Tasks') IS NULL
BEGIN
create table #Tasks
(
TaskID nvarchar(50),
DateStart bigint,
DateStop bigint,
Staff nvarchar(100)
)
insert into #Tasks values
('C001',1554181200000,1554190200000,'john,jack'),
('C002',1554202800000,1554212700000,'tom,john'),
('C003',1554228000000,1554246900000,'john,franck'),
('C004',1554613200000,1554626700000,'john')
END
GO
declare
#UserName nvarchar(50)='john',
#DateFrom datetime='2019-04-01',
#DateTo datetime='2019-04-30',
#nStart time='06:00:00',
#nStop time='22:00:00'
select
startday as [Day],
sum([WeekDay]) as [WeekDay],
sum([WeekNight]) as [WeekNight],
sum([WeekendDay]) as [WeekendDay],
sum([WeekendNight]) as [WeekendNight],
sum([TotalMinutes]) as [TotalMinutes],
0 WeekDayOverTime,
0 WeekNightOvertime,
0 WeekendDayOvertime,
0 WeekendNightOvertime,
[UserName]
,timeframe
from
(
select
iif(isWeekend=1,NightMinutes,0) WeekendNight,
iif(isWeekend=0,NightMinutes,0) WeekNight,
iif(isWeekend=1,DayMinutes,0) WeekendDay,
iif(isWeekend=0,DayMinutes,0) [WeekDay],
TotalMinutes,
username,
startday,
timeframe
from
(
select
iif(Before6>0,Before6,0)+ iif(After22>0,After22,0) NightMinutes,
TotalMinutes-iif(Before6>0,Before6,0)- iif(After22>0,After22,0) DayMinutes,
TotalMinutes,
startday,
isWeekend,
UserName,
timeframe
from
(
Select
(t.datestop-t.datestart)/60000 TotalMinutes,
datediff(n,convert(time,DATEADD(SECOND,t.DateStart/1000,'1970-01-01')),#nStart) Before6,
datediff(n,#nStop,convert(time,DATEADD(SECOND,t.DateStop/1000,'1970-01-01'))) After22,
iif((((DATEPART(DW, convert(datetime,DATEADD(SECOND,t.DateStart/1000,'1970-01-01'))) - 1 ) + ##DATEFIRST ) % 7) IN (0,6),1,0) isWeekend,
convert(varchar(10),DATEADD(SECOND,t.DateStart/1000,'1970-01-01'),126) startday,
STUFF(( SELECT distinct ' ' + convert(varchar(5),DATEADD(SECOND,t.DateStart/1000,'1970-01-01'),108)+'-'+convert(varchar(5),DATEADD(SECOND,t.DateStop/1000,'1970-01-01'),108) AS [text()]
FROM #Tasks tt
--WHERE tt.taskID=t.TaskID
FOR XML PATH('') ), 1, 1, '' ) AS [timeframe],
#UserName UserName
FROM #Tasks t
WHERE t.Staff like '%'+#UserName+'%'
and DATEADD(SECOND,t.DateStart/1000,'1970-01-01') between #DateFrom and #DateTo
) z
) zz
) zzz
group by startday,username,timeframe
order by startday
I need now to :
1) group result by day, summing up WeekDay,WeekNight,WeekendDay,WeekendNight and TotalMinutes, and concatenating timeframe so to have for example on 2nd April "05:00-07:30|11:00-13:45|18:00-23:00"
2) Not sum up time between 12:00 and 12:30 (if applicable) since it is lunch time
3) considering that after 8 hours daily it has to be calculated as overtime, I have to split the total minutes between in time and overtime, but depending if overtime is by daytime or nighttime or on the weekend
4) eventually using a holiday table
in other words we should have this:
Day TotalMinutes WeekDay WeekNight WeekendDay WeekendNight WeekDayOverTime WeekNightOvertime WeekendDayOvertime WeekendNightOvertime UserName timeframe
02/04/2019 630 420 60 0 0 45 75 0 0 john 05:00-07:30|11:00-13:45|18:00-23:00
07/04/2019 225 0 0 165 60 0 0 0 0 john 05:00-08:45
because (on 2nd April) we have:
First Task:
60 minutes of Regular NightTime
90 minutes of Regular DayTime
Second Task:
165 minutes of Regular DayTime, but have to count only 135 due to lunch time
Third Task:
240 DayTime
75 NightTime
but since with Task 1 and 2 we sum up 285 minutes, only the first 185 Minutes of Third task are Regular DayTime: the remaining 45 are Overtime DayTime, and the following 75 of NightTime are actually OvertimeNightTime
In this approach the first CTE (properDates) get the Start and Stop Datetimes, then you don't need to repeat that formula over the query.
The second CTE(splittedMinutes) is to get the same data you get in your current approach, except for the first CROSS APPLY, which is splitting the timeframes crossing with lunch time. The second CROSS APPLY gets the number of minutes and isWeekend value.
In the third CTE(qualifiedMinutes) I am using a window partition to get the accumulated minutes and generate the overtimes when applies.
At the end I used a selective SUM to separate weekdays and weekends in the aggregates
;with properDates AS (
SELECT TaskID, DATEADD(SECOND,t.DateStart/1000,'1970-01-01') as DateStart,DATEADD(SECOND,t.DateStop/1000,'1970-01-01') as DateStop, Staff
FROM #Tasks t
WHERE Staff LIKE '%' + #UserName + '%'
), splittedMinutes AS (
select
CAST(p.DateStart AS DATE) as [Day],
TotalMinutes,
SUM(TotalMinutes) OVER (PARTITION BY CAST(p.DateStart AS DATE) ORDER BY b.start) AS cumulate,
TotalMinutes - EarlyMinutes - LateMinutes as DayTime,
EarlyMinutes + LateMinutes as NightTime,
isWeekend,
CONVERT(VARCHAR(5),b.Start,108) + '-' + CONVERT(VARCHAR(5),b.Stop,108) as [timeframe]
from properdates p
cross apply (
select CAST(p.DateStart As TIME) AS Start, #bStart as Stop WHERE CAST(p.DateStart AS TIME) < #bStart and CAST(p.DateStop AS TIME) > #bStart
union
select #bStop as Start, CAST(DateStop AS TIME) AS Stop WHERE CAST(p.DateStop AS TIME) > #bStop and CAST(p.DateStart AS TIME) < #bStop
union
select CAST(p.DateStart AS TIME) AS Start, CAST(p.DateStop AS TIME) AS Stop WHERE NOT (CAST(p.DateStart AS TIME) < #bStart and CAST(p.DateStop AS TIME) > #bStart) AND NOT (CAST(p.DateStop AS TIME) > #bStop and CAST(p.DateStart AS TIME) < #bStop)
) b
cross apply (
select
DATEDIFF(Minute, b.Start, b.Stop) as TotalMinutes,
(DATEDIFF(Minute, CAST(b.Start AS TIME), #nStart) + ABS(DATEDIFF(Minute, CAST(b.Start AS TIME), #nStart))) / 2 as EarlyMinutes,
(DATEDIFF(Minute, #nStop, CAST(b.Stop AS TIME)) + ABS(DATEDIFF(Minute, #nStop, CAST(b.Stop AS TIME)))) / 2 as LateMinutes,
CASE WHEN DATEPART(DW, p.DateStart) IN (1,7) THEN 1 ELSE 0 END AS isWeekend
) c
), qualifiedMinutes As (
SELECT Day, TotalMinutes, RegularDay, RegularNight, OvertimeDay, OvertimeNight, isWeekend, timeframe
FROM splittedMinutes
OUTER APPLY (
SELECT RegularDay = CASE WHEN cumulate <= #maxTime THEN DayTime WHEN DayTime - (cumulate - TotalMinutes - #maxTime) > 0 THEN ABS(cumulate - TotalMinutes - #maxTime) ELSE 0 END
) RD
OUTER APPLY (
SELECT OvertimeDay = DayTime - RegularDay
) OWD
OUTER APPLY (
SELECT RegularNight = CASE WHEN cumulate <= #maxTime THEN NightTime WHEN (cumulate - TotalMinutes - #maxTime + RegularDay) < 0 THEN NightTime + (cumulate - TotalMinutes - #maxTime + RegularDay) ELSE 0 END
) RWN
OUTER APPLY (
SELECT OvertimeNight = NightTime - RegularNight
) OWN
)
SELECT
Day,
#UserName and UserName,
SUM(TotalMinutes) AS TotalMinutes,
SUM(CASE WHEN isWeekend = 0 THEN RegularDay ELSE 0 END) AS WeekDay,
SUM(CASE WHEN isWeekend = 0 THEN RegularNight ELSE 0 END) AS WeekNight,
SUM(CASE WHEN isWeekend = 1 THEN RegularDay ELSE 0 END) AS WeekendDay,
SUM(CASE WHEN isWeekend = 1 THEN RegularNight ELSE 0 END) AS WeekendNight,
SUM(CASE WHEN isWeekend = 0 THEN OvertimeDay ELSE 0 END) AS WeekDayOverTime,
SUM(CASE WHEN isWeekend = 0 THEN OvertimeNight ELSE 0 END) AS WeekNightOvertime,
SUM(CASE WHEN isWeekend = 1 THEN OvertimeDay ELSE 0 END) AS WeekendDayOverTime,
SUM(CASE WHEN isWeekend = 1 THEN OvertimeNight ELSE 0 END) AS WeekendNightOvertime,
STUFF((SELECT '|' + timeframe FROM qualifiedMinutes tt WHERE tt.Day = q.Day ORDER BY timeframe FOR XML PATH('') ), 1, 1, '' ) AS [timeframe]
FROM qualifiedMinutes q
GROUP BY Day

SQL Server PIVOT Function To Switch Rows, Columns

I'm trying to switch rows and columns with PIVOT (or another method). The documentation is pretty confusing to me. Thanks
DECLARE #CallCenterID Int;
DECLARE #BeginDateofReview SmallDateTime;
DECLARE #EndDateofReview SmallDateTime;
SELECT
COUNT(case when Score_Greeting = 'Yes' then Score_Greeting END) AS Score_Greeting_Passed,
SUM(CASE WHEN Score_Greeting IS NOT NULL THEN 1 ELSE 0 END) AS Score_Greeting_Reviewed,
ROUND(CONVERT(decimal(4,1), (COUNT(CASE WHEN Score_Greeting = 'Yes' THEN Score_Greeting END) * 100.0) / NULLIF(SUM(CASE WHEN Score_Greeting IS NOT NULL THEN 1 ELSE 0 END),0),0),0) AS Score_Greeting_PctngPassed,
COUNT(CASE WHEN Score_Authentication = 'Yes' THEN Score_Authentication END) AS Score_Authentication_Passed,
SUM(CASE WHEN Score_Authentication IS NOT NULL THEN 1 ELSE 0 END) AS Score_Authentication_Reviewed,
ROUND(CONVERT(decimal(4,1), (COUNT(CASE WHEN Score_Authentication = 'Yes' THEN Score_Authentication END) * 100.0) / NULLIF(SUM(CASE WHEN Score_Authentication IS NOT NULL THEN 1 ELSE 0 END), 0), 0), 0) AS Score_Authentication_PctngPassed,
FROM
Calls
WHERE
CallCenterID = #CallCenterID AND
(DateofReview >= #BeginDateofReview AND DateofReview <= #EndDateofReview)
Desired results:
Score_Greeting_Passed 5
Score_Greeting_Reviewed 9
Score_Greeting_PctngPassed 56
Score_Authentication_Passed 6
Score_Authentication_Reviewed 9
Score_Authentication_PctngPassed 67
You can use the UNPIVOT operator to transpose the rows and columns. If we assume your query above is stored in #YourData, it's going to be something like:
SELECT
UnpivotedData.ScoreType
, UnpivotedData.ScoreValue
FROM
#YourData
UNPIVOT (ScoreValue FOR ScoreType IN ( [Score_Greeting_Passed], [Score_Greeting_Reviewed], [Score_Greeting_PctngPassed],
[Score_Authentication_Passed], [Score_Authentication_Reviewed], [Score_Authentication_PctngPassed] )) AS UnpivotedData

Show value when CASE expression does not return rows

This code returns a fixed column 198 and then another one with a CASE. This CASE may return 0 rows (because there's also 0 row with TRL_ORIGQTY in the database, so it's not even evaluated)
If there are rows, the result is (tested with other organizations):
ORG SumTotal
198 98.51
If there are no rows in the database, the result shows nothing, but it should be:
ORG SumTotal
198 NULL
or whatever:
ORG SumTotal
198 00001
This is the full code:
declare #startdate varchar(30);
declare #enddate varchar(30);
declare #count int;
set #startdate = '2017-12-01'
set #enddate = '2018-01-01'
select #count = COUNT(*)
from R5TRANSLINES
where TRL_TYPE = 'STTK'
and TRL_TRANS in (select TRA_CODE from R5TRANSACTIONS
where TRA_FROMCODE in ('198')
and TRA_STATUS in ('A')
and TRA_DATE between #startdate and #enddate)
select distinct
'198' as 'ORG',
(SUM(CASE
WHEN (TRL_ORIGQTY = 0 AND TRL_ORIGQTY > 0) THEN 0
WHEN ((TRL_ORIGQTY - TRL_QTY) = 0 AND TRL_ORIGQTY = 0) THEN 100
WHEN ((TRL_ORIGQTY - TRL_QTY) > TRL_ORIGQTY) THEN ((TRL_ORIGQTY / (TRL_ORIGQTY - TRL_QTY)) * 100)
ELSE (((TRL_ORIGQTY - TRL_QTY) / TRL_ORIGQTY) * 100)
END) OVER ()/#count) AS 'SumTotal'
from
R5TRANSLINES
where
TRL_TYPE = 'STTK'
and TRL_TRANS in (select TRA_CODE from R5TRANSACTIONS
where TRA_FROMCODE in ('198')
and TRA_STATUS in ('A')
and TRA_DATE between #startdate and #enddate)
Tried with ISNULL, or NOT EXISTS, or even with another CASE = NULL, but none of them work... no rows returned.
UPDATED CODE:
select distinct
'ESREQ1' as 'ORG',
CASE COUNT(*)
WHEN 0 THEN 0
ELSE (SUM(CASE
WHEN (TRL_ORIGQTY = 0 AND TRL_ORIGQTY > 0) THEN 0
WHEN ((TRL_ORIGQTY - TRL_QTY) = 0 AND TRL_ORIGQTY = 0) THEN 100
WHEN ((TRL_ORIGQTY - TRL_QTY) > TRL_ORIGQTY) THEN ((TRL_ORIGQTY / (TRL_ORIGQTY - TRL_QTY)) * 100)
ELSE (((TRL_ORIGQTY - TRL_QTY) / TRL_ORIGQTY) * 100)
END) OVER ()/#count)
END AS 'SumTotal'
from
R5TRANSLINES
where
TRL_TYPE = 'STTK'
and TRL_TRANS in (select TRA_CODE from R5TRANSACTIONS
where UPPER(TRA_DESC) like '%AUDITESREQ1%'
and TRA_FROMCODE in ('138-05')
and TRA_STATUS in ('A')
and TRA_DATE between #startdate and #enddate)
group by
TRL_ORIGQTY, TRL_QTY
The reason you have 0 rows returning is that a SUM of 0 rows will return no rows (thus no dataset). The only exception to this is when using COUNT; which returns a 0 when there are no rows.
You could, therefore, use a further CASE expression before hand with a COUNT(*), and then put the SUM in the ELSE.
SELECT '198' AS ORG, --DISTINCT achieves nothing here, so I have removed it.
CASE COUNT(*) WHEN 0 THEN 0
ELSE ... --your SUM
END AS SumTotal
FROM R5TRANSLINES ...
Also, please think about using White Space when writing SQL. Your Code is very difficult to read to the complete lack of indentation. It makes your life easier, and anyone else that needs to read your statements. (this is why I haven't rewritten your query, as i'd spend more time formatting it than anything else).
Edit: Based on the OPs latest SQL:
SELECT 'ESREQ1' AS ORG,
CASE COUNT(*) WHEN 0 THEN 0
ELSE SUM(CASE WHEN (TRL_ORIGQTY = 0 AND TRL_ORIGQTY > 0) THEN 0
WHEN ((TRL_ORIGQTY - TRL_QTY) = 0 AND TRL_ORIGQTY = 0) THEN 100
WHEN ((TRL_ORIGQTY - TRL_QTY) > TRL_ORIGQTY) THEN ((TRL_ORIGQTY / (TRL_ORIGQTY - TRL_QTY)) * 100)
ELSE (((TRL_ORIGQTY - TRL_QTY) / TRL_ORIGQTY) * 100)
END) / #count
END AS SumTotal
FROM R5TRANSLINES
WHERE TRL_TYPE ='STTK'
AND TRL_TRANS IN (SELECT TRA_CODE
FROM R5TRANSACTIONS
WHERE UPPER(TRA_DESC) LIKE '%AUDITESREQ1%'
AND TRA_FROMCODE IN ('138-05')
AND TRA_STATUS = 'A'
AND TRA_DATE BETWEEN #startdate AND #enddate);
Note that this is untested, as I don't have any sample data to run this against. If this still doesn't work, please provide some DDL and consumable Sample data, that we can run against.
Thanks.

How To Get Non-Null,Non-0 for Average with SqlServer Select

**Note: I need to go a little further and add NULLIF(0 or 5). I wrote a short post about my answer here:
http://peterkellner.net/2013/10/13/creating-a-compound-nullif-in-avg-function-with-sqlserver/
but am not happy with my solution)
I've got a table with results where attendees type in estimated attendance to a course. If they type 0 or leave it empty, I want ignore that and get the average of values typed in. I can't figure out how to add that constraint to my AVG function without having a where clause for the entire SQL. Is that possible? My code looks like this: (EstimatedNumberAttendees is what I'm going after).
SELECT dbo.SessionEvals.SessionId,
AVG(Cast (dbo.SessionEvals.CourseAsWhole as Float)) AS CourseAsWholeAvg,
COUNT(*),
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'On Time' then 1
else null
end) AS SpeakerOnTime,
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'Late' then 1
else null
end) AS SpeakerLate,
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'NoShow' then 1
else null
end) AS SpeakerNoShow,
COUNT(case
when dbo.SessionEvals.PercentFull = '10% to 90%' then 1
else null
end) AS PercentFull10to90,
COUNT(case
when dbo.SessionEvals.PercentFull = '> 90%' then 1
else null
end) AS PercentFullGreaterThan90,
COUNT(case
when dbo.SessionEvals.PercentFull = ' < 10% Full ' then 1
else null
end) AS PercentFullLessThan10,
AVG(Cast (dbo.SessionEvals.EstimatedNumberAttendees as Float)) AS
EstimatedAttending
FROM dbo.Sessions
INNER JOIN dbo.SessionEvals ON (dbo.Sessions.Id =
dbo.SessionEvals.SessionId)
WHERE dbo.Sessions.CodeCampYearId = 8
GROUP BY dbo.SessionEvals.SessionId
AVG omits NULLs. Therefore make it treat 0s as NULLs. Use NULLIF for that:
...
AVG(NULLIF(Cast (dbo.SessionEvals.CourseAsWhole as Float), 0)) AS CourseAsWholeAvg,
...
AVG(NULLIF(Cast (dbo.SessionEvals.EstimatedNumberAttendees as Float), 0)) AS EstimatedAttending
...
You can try to use an inner query to get the same sessions but exclude zero and null:
SELECT dbo.SessionEvals.SessionId,
(
SELECT AVG(SE1.CourseAsWhole)
FROM dbo.SessionEvals SE1
WHERE SE1.SessionId = dbo.SessionEvals.SessionId
AND ISNULL(SE1.CourseAsWhole, 0) <> 0
) AS CourseAsWholeAvg,
COUNT(*),
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'On Time' then 1
else null
end) AS SpeakerOnTime,
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'Late' then 1
else null
end) AS SpeakerLate,
COUNT(case
when dbo.SessionEvals.InstructorPromptness = 'NoShow' then 1
else null
end) AS SpeakerNoShow,
COUNT(case
when dbo.SessionEvals.PercentFull = '10% to 90%' then 1
else null
end) AS PercentFull10to90,
COUNT(case
when dbo.SessionEvals.PercentFull = '> 90%' then 1
else null
end) AS PercentFullGreaterThan90,
COUNT(case
when dbo.SessionEvals.PercentFull = ' < 10% Full ' then 1
else null
end) AS PercentFullLessThan10,
AVG(Cast (dbo.SessionEvals.EstimatedNumberAttendees as Float)) AS
EstimatedAttending
FROM dbo.Sessions
INNER JOIN dbo.SessionEvals ON (dbo.Sessions.Id =
dbo.SessionEvals.SessionId)
WHERE dbo.Sessions.CodeCampYearId = 8
GROUP BY dbo.SessionEvals.SessionId
SQL AVG function will by default ignore null values so you need to only exclude the 0s. Your AVG code can be changed to below:
AVG(nullif( Cast(dbo.SessionEvals.CourseAsWhole as Float), 0) AS CourseAsWholeAvg

Resources