Sorry, I don't understand your exact question about pg_stats. In any case I cannot make strict assumptions about data, because that greatly varies from project to project (it's a SaaS) and over time. To give an idea the table has some tens of millions of rows, each project has from a few thousands to a few millions of rows and each project has its own tags that the customer can define (unlimited tags for each row, but usually only 1 - 10 actually used)
Marco Colli <collimarco91@gmail.com> writes:
> Unfortunately I don't get actual improvements. I use PG 11 and I run the
> following commands:
> ALTER TABLE subscriptions ALTER tags SET STATISTICS 1000;
> ANALYZE subscriptions;
> However the bias remains pretty much the same (slightly worse after). Any
> idea?
So what have you got in the pg_stats fields I asked about?
How big is this table anyway (how many rows, how many different tag
values)?
regards, tom lane