Re: Number of updated rows with LibPQ

On Fri, 2022-10-14 at 13:52 +0200, Dominique Devienne wrote:
> On Wed, Oct 5, 2022 at 8:17 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> > Laurenz Albe <laurenz(dot)albe(at)cybertec(dot)at> writes:
> > > On Wed, 2022-10-05 at 16:38 +0200, Dominique Devienne wrote:
> > > Yes, you have to use PQcmdTuples(), and you have to convert the string to an integer.
> > > But don't worry: the result will *not* be "INSERT 0 5", it will be just "5"
>
> Thanks. What's the leading 0 though, then?
> I guessed it might be the number of rows returned, but it isn't, see below:
>
> postgres=# create table foo (id serial primary key, v int);
> CREATE TABLE
> postgres=# insert into foo (v) values (1), (2) returning id;
>  id
> ----
>   1
>   2
> (2 rows)
>
> INSERT 0 2

That 0 is the OID of the newly inserted tuple.
Since there are no more tables WITH OIDS, the number is always 0, but
is left in the output for compatibility reasons.

> > Yeah, just applying atoi() or atol() to the result should be enough.
>
> Thanks too. Since I'm in C++, I used <charconv> instead, and
> discovered it can be empty something, not 0.
> I guess atoi() would have hidden that distinction, and worked anyway
> (returning 0).
>
> In the same topic, I've noticed an INSERT returns PGRES_COMMAND_OK,
> while an INSERT-RETURNING returns PGRES_TUPLES_OK. So there's no way
> to use the status to distinguish a SELECT from anything else? A RETURNING clause
> makes any statement supporting it an hybrid of a command and a query,
> but then how
> does one know the exact "kind" of the statement?
>
> E.g. So how does psql show INSERT in either cases? By parsing the SQL
> itself, client-side?
> Or is there a libpq API on PGresult that would allow to get the type
> of statement the result is from?

The command tag is not what you look at.

You simply check the result from an INSERT statement. If it is PGRES_TUPLES_OK,
it must have been INSERT ... RETRUNING.

Yours,
Laurenz Albe
--
Cybertec | https://www.cybertec-postgresql.com