On 12/12/2018 08:32, Richard Guo wrote: > This small revise is not expected to bring performance improvements, but > can improve the readability of the code that for empty relids, the qual > is always considered as being not-outerjoin_delayed.
I think code readability and maintainability would be improved by having fewer special cases and fast paths. In this particular case, I'd even remove the existing fast path and just let the subsequent loop run zero times. I doubt there is a performance benefit to the existing coding. And if there is no performance benefit, then it's not really a fast path, just another path.
I did not find any performance data in the commit history about the
existing fast path. Not sure about its influence.
This code was added in commit d7a6a0, by Tom.
Hi Tom, what is your opinion about the fast path in
