Open otan opened 4 years ago
I think there were more problems than listed above with the temp tables. I think we'd want an in-memory implementation for them by default, and some smarts about the locality in which they're created. There was a lot of discussion on the RFC; have you found those threads?
Nope, this (so far) isn't a definitive list (I just wanted to create the issue). Feel free to edit (or lead me to where to find them)!
Hello just curious if there's any update on this? Appreciating that this is likely more complex in the Cockroach case for various reasons I can see this being a potential question for our internal customers deciding between Cockroach and Aurora given Aurora uses local storage for temporary tables which can significantly improve read performance for some workloads.
The history here is that temp tables are used extensively in ORM test suites. We really wanted that coverage, so we implemented temp table support to be correct, but not efficient. Internally, they use real tables and all the overhead which comes with them. There is no immediate plan to implement performant temp tables.
Currently, temp tables (/views/sequences) are experimental. This issues is be a "master issue" of things remaining to do before moving it off experimental.
Things to do:
Enhancements
(feel free to edit, I couldn't find any more for the time being)
Jira issue: CRDB-5098
gz#22057