Обсуждение: Use of nextval, currval

I'm attempting to convert a code which uses Oracle to Postgres.

I am concerned about a couple of points in the SQL, which I should be able to do
in Postgres.

Say I have defined a sequence DataId;

The Oracle uses the sequence as follows:

insert into TABLE (id1,.......,id2)
            values (DataId.NextVal,......,DataId.Currval);

I believe that I can replace this in Postgres with:

insert into TABLE (id1,.......,id2)
            values (nextval('DataId'),.....,currval('DataId');


What I think is expected is that the values of id1 and id2 will be equal.
I have been told that Oracle will guarantee this; will that be assumption be
true for Postgres, also? (My concern is that, with multiple/many users of the
same database, there may be instances where DataId increments before the currval
happens, such that it's possible that id1 != id2, above.)

Is there a simple way to write the SQL such that I can only use nextval once,
keeping its value for both id1 and id2?  (I have no idea why the values are
duplicated.....)

For that matter, is that a realistic assumption for Oracle's treatment of
sequences?

Thanks,

Matt

Matt,

> I'm attempting to convert a code which uses Oracle to Postgres.
>
> I am concerned about a couple of points in the SQL, which I should be
> able to do
> in Postgres.
>
> Say I have defined a sequence DataId;
>
> The Oracle uses the sequence as follows:
>
> insert into TABLE (id1,.......,id2)
>             values (DataId.NextVal,......,DataId.Currval);
>
> I believe that I can replace this in Postgres with:
>
> insert into TABLE (id1,.......,id2)
>             values (nextval('DataId'),.....,currval('DataId');
>
>
> What I think is expected is that the values of id1 and id2 will be
> equal.

This is correct.   Please post if you for some reason find them not
equal.

-Josh Berkus

"Josh Berkus" <josh@agliodbs.com> writes:
>> I believe that I can replace this in Postgres with:
>>
>> insert into TABLE (id1,.......,id2)
>> values (nextval('DataId'),.....,currval('DataId');
>>
>> What I think is expected is that the values of id1 and id2 will be
>> equal.

> This is correct.

Well ... it's relying on the assumption that the elements of the VALUES
list will be evaluated left-to-right.  Which is true at the moment,
and I don't really expect it to change, but it still could be called an
unsafe dependency on undocumented implementation details.  I doubt you
will find anything in the SQL spec that requires implementations to
behave that way.

            regards, tom lane