Hannu Krosing <hannu@tm.ee> writes:
> But can't clearing up the index be left for "later" ?
Based on what? Are you going to store the information about what has to
be cleaned up somewhere else, and if so where?
> Indexscan has to check the data tuple anyway, at least for visibility.
> would adding the check for field sameness in index and data tuples be
> too big performance hit ?
It does pretty much suck, especially when you think about functional
indexes on expensive functions.
regards, tom lane