"Jie Zhang" <jzhang@greenplum.com> writes:
> This sounds great. One thing I am concern about is that this will add the
> dependency of node types into the access methods. If we still keep
> nodeBitmapIndexscan and let it do the bitmap construction for tids returned
> by amgetmulti.
No, I'm assuming the other proposal that was on the table, namely to get
rid of amgetmulti in its current form and instead have an AM call that
delivers a bitmap in one step. (Probably should rename the pg_am column
to something like amgetbitmap.) nodeBitmapIndexscan would become pretty
trivial. For the existing AMs this just means that they call
tbm_add_tuple(s) for themselves, which is no big problem, especially
considering that they probably get to save some code by not having to
stop the indexscan when the buffer array gets full.
regards, tom lane