如果您可以获得近似计数（由stuff like vacuum analyze statements维护），那么您可以使用这样的查询：

create view table_stat as 
select format('%I.%I', n.nspname, c.relname) "table name",
  reltuples "count",
  pg_size_pretty(pg_total_relation_size(c.oid)) "size",
  now()::date "date"
from pg_class c
join pg_namespace n on c.relnamespace = n.oid 
where n.nspname = 'public' 
and c.relkind  in ('r', 'm', 'p')
order by 3 desc;

然后创建一个表once并插入第一批：

create table stat_history as select * from table_stat;

然后在cron或别的什么东西中，添加新数据：

insert into stat_history select * from table_stat;

然后，最后查询显示结果：

select * from stat_history -- where "whatever you want";

请注意，使用format('%I')或quote_ident()始终可以正确转义双引号。
否则，您将需要使用动态查询，使用plpgsql，例如：

• 网址：https://www.postgresql.org/docs/15/plpgsql-control-structures.html#id-1.8.8.8.3.4
• execute format('…')，如www.example.com中的示例https://www.postgresql.org/docs/15/plpgsql-control-structures.html#PLPGSQL-RECORDS-ITERATING

查看pg_stat_all_tables系统视图：

create table test as select 1 AS a;

select n_tup_ins,n_tup_upd,n_tup_del,n_tup_hot_upd,n_live_tup,n_dead_tup 
from pg_stat_all_tables where schemaname='public' and relname='test';
-- n_tup_ins | n_tup_upd | n_tup_del | n_tup_hot_upd | n_live_tup | n_dead_tup
-------------+-----------+-----------+---------------+------------+------------
--         1 |         0 |         0 |             0 |          1 |          0

insert into test select 2; 
insert into test select 3; 
delete from test where a=1; 
update test set a=5 where a=2;

select n_tup_ins,n_tup_upd,n_tup_del,n_tup_hot_upd,n_live_tup,n_dead_tup 
from pg_stat_all_tables where schemaname='public' and relname='test';
-- n_tup_ins | n_tup_upd | n_tup_del | n_tup_hot_upd | n_live_tup | n_dead_tup
-------------+-----------+-----------+---------------+------------+------------
--         3 |         1 |         1 |             1 |          2 |          2

您可以使用cron或pg_cron每天对此视图进行快照，以便不仅监视估计的行数，还监视导致该行数的流量。
要实时查看所有流量，您可以使用pg_recvlogical接入logical streaming replication protocol。