Louis Davidson shares an approach:
Temp tables fit into my query writing process as one of those last ditch efforts to make a query execute fast enough. Of course it would have been harder if I had followed all the rules and added a lot of test cases, but I had a pretty busy week last week and am finishing this pretty late on Tuesday night, so I just gave my opinions.
I like Louis’s strategy. It’s easy to add a lot of complexity to queries out of habit, to micro-optimize performance, or because of a meandering thought process. But many times, taking a step back to think about what could make a query simpler will be helpful.
One thing I’ll cover that Louis didn’t touch on is that performance level is (or should be) a requirement. If you have a query running millions of times a day on a system, then yes, it makes sense to squeeze out every microsecond. But for an ELT job that finishes in 20 minutes and where you have a 6-hour window to get the data loaded, shaving five minutes off of the query’s runtime isn’t that important, especially if you’re in read-committed snapshot isolation or using another form of optimistic concurrency.
This is why I recommend starting with simple and only moving to more complex solutions when you need them. Now, do I always follow my own advice? Err…well, the sign pointing to Boston doesn’t have to go there itself…