>
> On Sat, Mar 08, 2003 at 07:24:59PM -0500, John Siracusa wrote:
> > Just a quick question on the topic. How far are you
> willing to go (or
> > willing to let DBD authors go) to support the API you're planning?
>
> |<---- this far ---->|
>
> :)
>
> > For example, Postgres has a SERIAL psuedo-type that really just
> > creates a sequence behind the scenes for you. I currently get the
> > "last insert id" using DBD::Pg by querying what I know is the
> > auto-generated sequence name
> > ("<tablename>_<columnname>_seq") Would it be reasonable
> for DBD::Pg to do
> > this for me behind the scenes in order to service the
> proposed new API, or
> > would that be frowned upon (because it has to run another
> query, or because
> > it "guesses" the name of the sequence, or some other reason)?
>
> It's for just this reason that I suggested the last_insert_id
> method may take $table_name and $column_name parameters.
>
> For postgres the "<tablename>_<columnname>_seq" name is part
> of the core system so there's really no "guessing" going on
> (as I understand it). DBD::Oracle isn't so lucky.
Neither is DBD::ODBC, I think...unless I'm missing something obvious...(need
to re-read ODBC specs on this, because I don't believe the select
$column_name from $table_name where $column_name IS NULL works...)
Jeff
