On Tue, 31 Jan 2023, David G. Johnston wrote:
>
> It feels like there is room for improvement here using table statistics and the visibility map to significantly
reducethe number of pages retrieved that turn out to be all dead.
Yes, I too wish postgres was more clever for a simple unordered query. I
should also mention that this table has two btree indices, one primary key
index and one on another column. Would it be feasible to "descend" an
index in order to immediately find some live data?
> A limit without an order is not a query to optimize for of course...
A good reason is that the web interface I'm using to browse my tables (and
I guess many other too) do a simple limited+unordered SELECT when you
click on a table. Timing out was what made me look deeper. I couldn't
accept that no matter how slow my system is, it has trouble returning a
few rows.
It is also the first query somoebody will type on psql to interactively
get a clue about the contents of a table.
Regards,
Dimitris
