Forum Discussion
Why does this formula triggers a cross join?
I have a test dimension table and a fact table. I try to count the number of ids in the fact table. However, my DAX formula triggers a cross join between the tables. I want to know why. D_id is from the dimension table. F_id is the id in the fact table.
If I just use "COUNT()" without the IF statement to test blank values. It does not trigger the cross join. I am very confused.
Anonymous
I have to agree that this behaviour is confusing!
See this page for an explanation from the experts:
https://www.sqlbi.com/articles/autoexist-and-normalization/
When columns of the same table are added to a visual, the AutoExist feature ensures only existing combinations of values from those columns are displayed.
However, when columns of different tables are added to a visual (as in your case), even if there is a relationship between the tables, all combinations including "non-existing" combinations are displayed.
On top of this, any "rows" where all measures evaluate to blank are hidden.
Since you have constructed a measure that always returns a non-blank value, all possible combinations of D_id and F_id are displayed, effectively giving you a crossjoin of the two sets of values.
When you use COUNT() without IF, the non-existing combinations are hidden because the measure returns blank.
Did you have a particular output you were trying to create, or were you just curious about the behaviour?
Regards,
Owen
2 Replies
- OwenAuger
Super User
Anonymous
I have to agree that this behaviour is confusing!
See this page for an explanation from the experts:
https://www.sqlbi.com/articles/autoexist-and-normalization/
When columns of the same table are added to a visual, the AutoExist feature ensures only existing combinations of values from those columns are displayed.
However, when columns of different tables are added to a visual (as in your case), even if there is a relationship between the tables, all combinations including "non-existing" combinations are displayed.
On top of this, any "rows" where all measures evaluate to blank are hidden.
Since you have constructed a measure that always returns a non-blank value, all possible combinations of D_id and F_id are displayed, effectively giving you a crossjoin of the two sets of values.
When you use COUNT() without IF, the non-existing combinations are hidden because the measure returns blank.
Did you have a particular output you were trying to create, or were you just curious about the behaviour?
Regards,
Owen
Thanks for the link and explanation. I kept running through my head that the row context was correct and the filter context was correct. I was banging my head on the wall.
"auto exists" - this is the first time I heard of it.
I have a report that needs to turn blanks cells into zero. Do you have any good idea how to do it?
Again thank you very much.