Обсуждение: bug in create table with serial

In PostgreSQL 7.3b2, if I create a table like so:

create table customers (id serial primary key);

The id field will be created of type 32 bit integer, yet the upper limit
for the sequence will be relative to a 64 bit integer.

\d customers                                 Table "public.customers"    Column    |     Type     |
   Modifiers 
-------------+--------------+-----------------------------------------------------------  id          | integer      |
notnull default 
nextval('public.customers_id_seq'::text)



shop-devel=# select * from customers_id_seq;
-[ RECORD 1 ]-+--------------------
sequence_name | customers_id_seq
last_value    | 1
increment_by  | 1
max_value     | 9223372036854775807
min_value     | 1
cache_value   | 1
log_cnt       | 1
is_cycled     | f
is_called     | f



shop-devel=# select version ();                                version
---------------------------------------------------------------------  PostgreSQL 7.3b2 on i386-portbld-freebsd4.7,
compiledby GCC 2.95.4 

Cristóvão Dalla Costa <cbraga@bsi.com.br> writes:
> In PostgreSQL 7.3b2, if I create a table like so:
> create table customers (id serial primary key);

> The id field will be created of type 32 bit integer, yet the upper limit
> for the sequence will be relative to a 64 bit integer.

So?

AFAICS, you will get a failure when you reach 2^31 items.  Whether the
failure is "can't fit the value into int4" or "reached sequence maximum"
doesn't seem very exciting to me ...
        regards, tom lane

On Wed, 16 Oct 2002, Tom Lane wrote:

> Cristóvão Dalla Costa <cbraga@bsi.com.br> writes:
> > In PostgreSQL 7.3b2, if I create a table like so:
> > create table customers (id serial primary key);
>
> > The id field will be created of type 32 bit integer, yet the upper limit
> > for the sequence will be relative to a 64 bit integer.
>
> So?
>
> AFAICS, you will get a failure when you reach 2^31 items.  Whether the
> failure is "can't fit the value into int4" or "reached sequence maximum"
> doesn't seem very exciting to me ...
>

Unless of course the column's data type was int8.

Kris Jurka