On Thu, Nov 2, 2017 at 11:35 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Thu, Nov 2, 2017 at 8:01 AM, Craig Ringer <craig@2ndquadrant.com> wrote:
>> That forces materialization, and I'm guessing part of Tomas's goal
>> here is to prevent the need to materialize into a temp table /
>> tuplestore / etc.
>
> I get that, but if you're running a query like "SELECT * FROM
> bigtable", you don't need parallel query in the first place, because a
> single backend is quite capable of sending back the rows as fast as a
> client can read them. If you're running a query like "SELECT * FROM
> bigtable WHERE <highly selective predicate>" then that's a good use
> case for parallel query, but then materializing it isn't that bad
> because the result set is a lot smaller than the original table.
>
> I am not disputing the idea that there are *some* cases where parallel
> query is useful and materialization is still undesirable, of course.
Not seeing a code-level review of the proposed patch, I am moving it
to next CF to let the discussion move on. Nobody has registered as
reviewer yet.
--
Michael