"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> pasman pasma*ski<pasman.p@gmail.com> wrote:
>> This below not works :-( Always is used merge join.
>> SELECT * FROM test1 join test2 on(test1.ctid=test2.ct)
> You're reading through the entirety of two tables matching rows
> between them. What makes you think random access would be faster
> than sequential?
FWIW, it isn't going to happen anyway, because the TID scan mechanism
doesn't support scanning based on a join condition. That hasn't gotten
to the top of the to-do list because the use case is almost vanishingly
small. ctids generally aren't stable enough for it to be useful to
store references to one table's ctids in another table.
regards, tom lane
