Luca Ferrari <fluca1978@infinito.it> writes:
> I'm curious to know why, if a table is empty, it seems that an ANALYZE
> of the table does not insert any stats in the pg_stats table,
Uh ... because there are no stats to insert. The only available
information is that the table is empty, which is something we track in
pg_class not pg_stats.
> Second, if I execute an EXPLAIN
> on an empty table, even after an ANALYZE of the table, I got an explain that
> proposes me a row numbers and size that I cannot understand (since it seems to
> be different depending on the table definition).
That's the planner not wanting to bet on an empty table staying empty.
regards, tom lane