Forum Discussion
Which dax statement is more efficient?
Dear users.
I was wondering, while building a piece of DAX, if i could improve it or not. Below is a formula i use to calculate the sales in period 1, starting at P1 start and ending at P1 end.
P1 Sales = CALCULATE([Sales];DATESBETWEEN(DateTable[Dates];[P1 start];[P1 end]))
The second argument might lead to a table of e.g. 365 rows, if i have P1 start at Januari 1st, and P1 end at December 31st.
As i know the sales table ALWAYS ONLY has sales at the 1st of the month, i thought, why not have the second argument filtered?
Sales is a table of 500k rows by the way.
P1 Sales* = CALCULATE([Sales];filter(DATESBETWEEN(DateTable[Datum];[P1 start];[P1 end]);day(DateTable[Datum])=1))
or:
P1 Sales* = CALCULATE([Sales];DATESBETWEEN(DateTable[Datum];[P1 start];[P1 end]);day(DateTable[Datum])=1)
So this way i pass as filter to CALCULATE only 12 days, namely the 1st day of every month.
Would this speed matters up? Or is this hypothetical BS and do i just creat more overhead? :P
denpries,
In your original post, the P1 Sales* will be more efficient than P1 Sales, and it executes faster than P1 Sales. The DAX Greg_Deckler provides will spend more time to execute comparing with your original measures.
You can use SQL Server Profiler to check the query duration and DAX query plan for these measures and verify the whole process. There are some blogs for your reference.
https://insightsquest.com/2017/05/07/profiler-trace-for-power-bi-desktop/
https://www.sqlbi.com/wp-content/uploads/DAX-Query-Plans.pdf
Regards,
Lydia
5 Replies
Perhaps:
P1 Sales = VAR tmpDates = FILTER(FILTER(DateTable,DAY([Dates])=1),[Dates]>=[P1 start] && [Dates]<=[P1 end) RETURN SUMX(tmpDates,[Sales])
Greg_Deckler that also works, yes. If i understand it correctly you filter from the dates table all applicable rows (so basically 1 per month) and then sumx evaluates the [sales] for every row of that table and adds it together.
What i dont understand yet, and try to know why, if this approach is more efficient then the other.Anybody else with an opinion about this / some more in depth knowledge?
