Great content again, Reid & Denis! I love the approach to have those guidelines seen as something to aspire for but no NoGos - especially your example with the calculated columns. After many projects without any I also ended up using one recently, as my table scans for initial delivery date per product over millions of fact rows basically worked - but just when few product categories were filtered. The complete result table wouldn't render in the visual due to excessive memory usage. Moving this as a flag in a calculated column made it not only blazing fast but actually allowed to display the whole dataset in a table visual. Not that anyone would really need to see the whole table, but the user being able to create broken visuals is never good.
Just an interesting note about query folding. Sometimes if it breaks try a different way to do that function. Like if you use the upper menu to do uppercase, try using the right-click way of doing it. I've seen where one will fold and the other will not and it's not consistent.
Hmmmm, is the M code identical otherwise? Also if it's loading a cached preview of the table on screen, it doesn't always show the clickable show native query button, if you refresh the step though, it can show. Though the 100% sure way of knowing is to run something like SQL profiler to see what's being passed back to the DB :)
I want both those t-shirts! I’m about to start paraphrasing Roche and Havens in my own training sessions! Great session thanks
Awesome! Thank you!
Thank you for such great, immediately useful, content, Reid and Denis!
Glad you enjoyed it!
Great content again, Reid & Denis!
I love the approach to have those guidelines seen as something to aspire for but no NoGos - especially your example with the calculated columns. After many projects without any I also ended up using one recently, as my table scans for initial delivery date per product over millions of fact rows basically worked - but just when few product categories were filtered. The complete result table wouldn't render in the visual due to excessive memory usage. Moving this as a flag in a calculated column made it not only blazing fast but actually allowed to display the whole dataset in a table visual.
Not that anyone would really need to see the whole table, but the user being able to create broken visuals is never good.
Great example of an exception on when to use a calculated column. Thank you!
Just an interesting note about query folding. Sometimes if it breaks try a different way to do that function. Like if you use the upper menu to do uppercase, try using the right-click way of doing it. I've seen where one will fold and the other will not and it's not consistent.
Hmmmm, is the M code identical otherwise? Also if it's loading a cached preview of the table on screen, it doesn't always show the clickable show native query button, if you refresh the step though, it can show. Though the 100% sure way of knowing is to run something like SQL profiler to see what's being passed back to the DB :)