They are all left joins in there: 1st left joined with 4 others and one of these 4 left joined with yet another. Vertical alignment is just a graphical representation. What exactly seems puzzling to you?
Thanks for your response.
Is this what tableau doing
(((((CodesWithDates Left Join 31.03.2016) Left join Amortisation) Left Join Dividend) Left Join Interest) Left Join Sheet1)
2 of 2 people found this helpful
Let's clarify few things:
- Join order does not affect the result
- In some case Tableau may choose to reshuffle join order for reasons I'd rather not dive into
- In general, yes - for cross-database joins Tableau will follow encapsulation you described however it's not driven by UI layout (tables are alphabetically sorted) but rather by the order in which tables are added to the canvas
p.s. Shameless plug: I'll be talking about some of that in my TC16 session
Need one more clarification on Joins.
Does your statement "Join order does not affect the result" hold true only in Tableau scenario as it does not support outer & right join? As Left & Inner join order does not change the result
Order of joins does not matter only in case of inner joins.
Outer joins order does matter and therefore I said that it probably depends on implementation and I believe the joins are executed in order of their appearance in the code.
I guess you are both correct and I need to rethink my answer.
My original answer referred to lack of grouping (or "layering" of you will of join statements but I can totally see now how it could provide different results.
Crawling back into my corner...
Actually, I just created 2 data sources, live connections to Redshift, both with the same exact tables, with all of the same exact inner joins, and got different results by about 300k rows.
Any idea how this can happen?
1 of 1 people found this helpful
Different [datasource] filters? Different fields used for join conditions? Different credentials used?
What does that datasource look like? Any joins?
Turns out there was some dirty data. This is actually welcome news because 1) this is NOT our production warehouse and 2) I knew that with inner joins the order of tables doesn't matter and I was baffled at the inconsistent results!!
thanks for the reply Dmitry!