MIME-Version: 1.0
References: 
 <CA+FnnTx+VJi6o9Y8HHUxb89REdNOxBtz1AVkbvM=TGMF_xa=oA@mail.gmail.com>
 <CADK3HHJLCxp+5Q48CatYqg02hBJs0qs8DS=x=Pg5Nx64rD1F7g@mail.gmail.com>
 <CA+FnnTwB1GJX-ZnO+oAJw+z+rmDVcfK7Huprj7oOj3V88Y2ZUA@mail.gmail.com>
In-Reply-To: 
 <CA+FnnTwB1GJX-ZnO+oAJw+z+rmDVcfK7Huprj7oOj3V88Y2ZUA@mail.gmail.com>
From: Igor Korot <ikorot01@gmail.com>
Date: Sat, 7 Mar 2026 17:14:00 -0600
Message-ID: 
 <CA+FnnTxF+VThVXgobZtu6-jfCc6zfqh1tC4dPtWegfov2CFW3w@mail.gmail.com>
Subject: Re: How to properly perform this?
To: Dave Cramer <davecramer@postgres.rocks>
Cc: PostgreSQL ODBC list <pgsql-odbc@postgresql.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CA%2BFnnTxF%2BVThVXgobZtu6-jfCc6zfqh1tC4dPtWegfov2CFW3w%40mail.gmail.com>
Precedence: bulk

Dave,

This is from the gdb session:

Hi, David,

On Sat, Mar 7, 2026 at 1:06=E2=80=AFPM Igor Korot <ikorot01@gmail.com> wrot=
e:
>
> Hi, Dave,
> I think its just an empty column.
>
> Is there a way to treat it as NULL?
>
> P.S.: Sorry for top-posting.
>
> On Sat, Mar 7, 2026 at 5:41=E2=80=AFAM Dave Cramer <davecramer@postgres.r=
ocks> wrote:
> >
> >
> > Dave Cramer
> > www.postgres.rocks
> >
> >
> > On Sat, 7 Mar 2026 at 01:10, Igor Korot <ikorot01@gmail.com> wrote:
> >>
> >> Hi, ALL,
> >> I'm trying to execute following query:
> >>
> >> draft=3D# SELECT c.relname AS name, ixs.tablespace, ARRAY(SELECT
> >> a.attname FROM pg_attribute a WHERE a.attrelid =3D idx.indrelid AND
> >> a.attnum =3D ANY(idx.indkey) AND a.attnum > 0 ORDER BY
> >> array_position(idx.indkey, a.attnum) OFFSET idx.indnkeyatts) AS
> >> included, c.reloptions AS storage FROM pg_index idx, pg_class c,
> >> pg_namespace n, pg_class t, pg_indexes ixs WHERE ixs.indexname =3D
> >> c.relname AND c.oid =3D idx.indexrelid AND t.oid =3D idx.indrelid AND
> >> n.oid =3D c.relnamespace AND idx.indisprimary AND n.nspname =3D 'publi=
c'
> >> AND t.relname =3D 'leagues';
> >>      name     | tablespace | included | storage
> >> --------------+------------+----------+---------
> >>  leagues_pkey |            | {}       |
> >> (1 row)
> >>
> >>
> >> My code is as follows:
> >>
> >>             while( ( ret =3D SQLGetData( m_hstmt, 4, SQL_C_WCHAR,
> >> index_param.get(), 255, &ind[3] ) ) !=3D SQL_NO_DATA )
> >>             {
> >>                 if( ret =3D=3D SQL_SUCCESS || ret =3D=3D SQL_SUCCESS_W=
ITH_INFO )
> >>                 {
> >>                     auto numBytes =3D ind[3];
> >>                     if( ind[3] =3D=3D SQL_NO_TOTAL )
> >>                         numBytes =3D 255;
> >>                     else if( ind[3] > 255 )
> >>                         numBytes =3D 255;
> >>                     str_to_uc_cpy( options, index_param.get() );
> >>                 }
> >>                 else
> >>                 {
> >>                     GetErrorMessage( errorMsg, STMT_ERROR  );
> >>                     result =3D 1;
> >>                 }
> >>             }
> >> Unfortunately it goes into the loop because apparently the driver or
> >> the server keeps seeing the last column as having data and returns
> >> SQL_SUCCESS.
> >>
> >> What is the proper way to resolve this?
> >>
> >> I'm using unixODBC with PostgreSQL 16.
> >>
> >> My odbc.ini looks like this:
> >>
> >> [code]
> >> [postgres]
> >> Driver=3DPostgres
> >> Description=3DPostgres ODBC Driver
> >> UserId=3Dpostgres
> >> Password=3DXXX
> >> Servername=3Dlocalhost
> >> Database=3Ddraft
> >> Port=3D5432
> >>
> >> [/code]
> >>
> >>
> >> What is the best way to proceed?
> >
> >
> > is the last column empty or null?
> >
> > if not null then there is data there.

 3754                while( ( ret =3D SQLGetData( m_hstmt, 3,
SQL_C_WCHAR, included.get(), 255, &ind[2] ) ) !=3D SQL_NO_DATA )
(gdb) n
3756                    if( ret =3D=3D SQL_SUCCESS || ret =3D=3D SQL_SUCCES=
S_WITH_INFO )
(gdb) p ind[2]
$1 =3D 4
(gdb) n
3758                        auto numBytes =3D ind[2];
(gdb)
3759                        if( ind[2] =3D=3D SQL_NO_TOTAL )
(gdb)
3761                        else if( ind[2] > 255 )
(gdb)
3763                        str_to_uc_cpy( includedCol, included.get() );
(gdb)
3764                    }
(gdb)
3754                while( ( ret =3D SQLGetData( m_hstmt, 3,
SQL_C_WCHAR, included.get(), 255, &ind[2] ) ) !=3D SQL_NO_DATA )
(gdb)
3771                includedCol.erase( 0, 1 );
(gdb) p ind[2]
$2 =3D 4
(gdb) p ret
$3 =3D 100
(gdb) n

This is what happens with column 3 when the array is empty.

3779            while( ( ret =3D SQLGetData( m_hstmt, pos, SQL_C_WCHAR,
index_param.get(), 255, &ind[3] ) ) !=3D SQL_NO_DATA )
(gdb)
3781                if( ret =3D=3D SQL_SUCCESS || ret =3D=3D SQL_SUCCESS_WI=
TH_INFO )
(gdb) p ind[3]
$4 =3D -1
(gdb) p ret
$5 =3D 0
(gdb) n
3783                    auto numBytes =3D ind[3];
(gdb)
3784                    if( ind[3] =3D=3D SQL_NO_TOTAL )
(gdb)
3786                    else if( ind[3] > 255 )
(gdb)
3788                    str_to_uc_cpy( options, index_param.get() );
(gdb)
3789                }
(gdb)
3779            while( ( ret =3D SQLGetData( m_hstmt, pos, SQL_C_WCHAR,
index_param.get(), 255, &ind[3] ) ) !=3D SQL_NO_DATA )
(gdb)
3781                if( ret =3D=3D SQL_SUCCESS || ret =3D=3D SQL_SUCCESS_WI=
TH_INFO )
(gdb) p ind[3]
$6 =3D -1
(gdb) p ret
$7 =3D 0
(gdb)

And this one is for column 4 of the query.

The column is NULL, so no data is there.

I think it's a bug in the driver as it should return SQL_NO_DATA on
the very first call....

Thank you.

> >
> > Dave