SQL Server DW-使用日期维度和时间维度查询数据

Question

Before in the source systems I could do something like below

SELECT COUNT(*)
FROM FlightDeparture FD
WHERE FD.[Response_ts] BETWEEN '2013-01-01 45:01:38.000' AND '2013-01-02 15:04:20.000'

NOW we have a DW where there is a separate date dimension ( Dim_Date ) and time dimension ( Dim_time ), I tried below query but I am not getting same counts

SELECT COUNT(*)
FROM FACT_FlightDeparture FD
INNER JOIN DIM_DATE DD ON FD.Resp_DATE_SK = DD.Date_Sk
INNER JOIN DIM_TIME TT ON FD.Resp_TIME_SK = TT.TIME_Sk
AND (DD.Cal_Date >= '2013-01-01' AND TT.Time_Of_Day >= '14:01:38.000')
AND (DD.Cal_Date <= '2013-01-02' AND TT.Time_Of_Day <= '14:04:20.000')

Please help. I have reduced the issue to how I am doing the TIME portion

Thanks in advance

Answer 1

The AND statement is preventing the two predicates from hardly ever returning true. Due to the Time_Of_Day difference, there is actually just under 3 minutes per day that the condition could return true.

If my thinking is correct, you need all times on 1/1 greater than or equal to 14:01:38 and all times on 1/2 less than or equal to 14:04:20.

SELECT COUNT(*)
FROM FACT_FlightDeparture FD
INNER JOIN DIM_DATE DD ON FD.Resp_DATE_SK = DD.Date_Sk
INNER JOIN DIM_TIME TT ON FD.Resp_TIME_SK = TT.TIME_Sk
WHERE (
         (DD.Cal_Date = '2013-01-01' AND TT.Time_Of_Day >= '14:01:38.000')
      OR (DD.Cal_Date = '2013-01-02' AND TT.Time_Of_Day <= '14:04:20.000')
)

If the fact table is big and you plan to use this query pattern often, I'd put both of these columns into a single nonclustered index.

CREATE NONCLUSTERED INDEX IX_FACT_FlightDeparture_DateTime 
ON FACT_FlightDeparture (Resp_DATE_SK,Resp_TIME_SK);

Also, don't forget your partitioning, compression, and other options on the index if applicable.

Answer 2

I don't have a answer for this question. Can you check if exists more rows in DIM_TIME than in FlightDeparture ? Let´s say original table FlightDeparture has only two records

'14:01:38.000','14:04:20:000'

and DIM_TIME has another one, even duplicate

'14:01:38.000','14:04:19:000','14:04:20:000'