Bruce Momjian wrote:
>
> My guess is that the creation of the index updates the table size
> statistics.
Yes.
> However, when I see zero size, I don't know if it is accurate, or if
> someone has added rows since the last vacuum/index creation, so I think
> it is correct to use an index on a zero-length table if it is
> appropriate. If the size is 1, I will assume that number is accurate,
> and do a sequential scan.
>
> Does that make sense?
Yes. But we have to fix SeqScan for field1 = -1...
Vadim
