Re: BUG #13830: pg_table_size and pg_indexes_size

Mr. Tom Lane
Thank you for your explanation .:)

在 2015/12/24 2:22, Tom Lane 写道:
> foggyglass(at)163(dot)com writes:
>> postgres=# create table b(id int);
>> CREATE TABLE
>> postgres=# create index b_idx on b(id);
>> CREATE INDEX
>> postgres=# select relname , pg_table_size(oid),pg_indexes_size(oid)
>> postgres-# from pg_class where relname in ('b','b_idx');
>> relname | pg_table_size | pg_indexes_size
>> ---------+---------------+-----------------
>> b | 0 | 8192
>> b_idx | 8192 | 0
>> (2 rows)
> Seems fine to me. b is empty at this point, so it has size zero.
> btree indexes, on the other hand, never have size zero because their
> metapage is created immediately. b_idx has no indexes attached to it,
> so pg_indexes_size finds nothing to report on and returns zero for
> that.
>
> If you insert any actual data, the results change:
>
> regression=# insert into b values(33);
> INSERT 0 1
> regression=# select relname , pg_table_size(oid),pg_indexes_size(oid) from pg_class where relname in ('b','b_idx');
> relname | pg_table_size | pg_indexes_size
> ---------+---------------+-----------------
> b | 8192 | 16384
> b_idx | 16384 | 0
> (2 rows)
>
> which maybe looks a bit saner.
>
> regards, tom lane
>