Bryan Rebok and Brent Ozar recently opened my eyes to something I didn’t know. When SQL Server recommends missing indexes to you it puts the columns in order in which they are found in the table. That’s it. I always thought there was more logic into it. But there isn’t. Upon reading this I had a terrible realization that I’ve made a lot of awful indexes in my time. I owe the world an apology. I hope this post can serve as that apology.
I’ve written a function that accepts the equality column list from dm_db_missing_index_details as a parameter and spits those columns back out in order by their cardinality. This won’t necessarily be the proper order for the columns in every index, but it is far more likely to be correct than the initial result from the DMV.
I’m amazed that the missing index DMV generates column names in such a simplistic manner.