Doh, you're right of course! We were having this problem in the Postgres 9.1 era, before index-only scans were a thing.
Still, it's quite inefficient maintaining an extra copy of the data that is never actually used. Though no longer multiple orders of magnitude less efficient.
However, I'd guess that programmers don't often think to add these seemingly useless fields to an index, as it feels inefficient and just wrong. But at least this offers an out in pathetic cases.
Still, it's quite inefficient maintaining an extra copy of the data that is never actually used. Though no longer multiple orders of magnitude less efficient.
However, I'd guess that programmers don't often think to add these seemingly useless fields to an index, as it feels inefficient and just wrong. But at least this offers an out in pathetic cases.