I have an indexed datetime column - lets call it "CreatedDateUTC" in table "Entity" in MSSQL db. Column contains UTC datetime. User connecting to db has known timezone offset. I have another table - DateDim with primary key column Date. This is to support date dimension for SSAS/Tabular model/Direct query scenario.
So on model level "Entity" is defined as SQL query
SELECT CAST(fn_AdjustCurrentTimezoneOffset(CreatedDateUTC) as Date) as CreatedDate FROM dbo.Entity
The SQL query featuring DateDim from report then looks like
SELECT * FROM [Entity] e JOIN [DateDim] d on CAST(fn_AdjustCurrentTimezoneOffset(e.CreatedDateUTC) as Date) = d.Date
which is unbearably slow when [Entity] table grows large. Even if i resign on adjusting timezone (which I don't want to), it doesn't get much better - i.e. this query is slow as well:
SELECT * FROM [Entity] e JOIN [DateDim] d on CAST(e.CreatedDateUTC as Date) = d.Date
My feeling is the issue here is with
CreatedDateUTC datetime - it needs to be converted to indexed
CreatedDateDayUTC date upfront (not a problem since I query special reporting db). My problem is with the timezone - since users have different timezone offset I cannot simply precalculate this.
On SQL level I would be able to get around this by this ugly hack - precalculate three dates (
Entity table and adjusting join to something like:
SELECT * FROM [Entity] e JOIN [DateDim] d on CAST(e.CreatedDateUTC as Date) = d.Date and (e.Date1 = d.Date or e.Date2 = d.Date and e.Date3 = d.Date)
However this can't be translated to query defining "Entity" on model level (there has to be just one column which can create relationship do DateDim Date column).
Any idea how to get around this ?