On Thu, 2003-07-31 at 17:21, Tom Lane wrote:
> Robert Treat <xzilla@users.sourceforge.net> writes:
> > On Thu, 2003-07-31 at 16:50, Tom Lane wrote:
> >> None; we just haven't gotten around to removing code that no longer
> >> pulls its weight. The no-column-name variant is just as dangerous as
> >> it was in the COPY case, IMHO.
>
> > by dangerous you mean functional right?
>
> :-)
>
> By dangerous I mean "might not restore the table correctly". There are
> scenarios involving child tables and ALTER TABLE ADD COLUMN where a
> column-name-less INSERT or COPY will dump the data in a different column
> order than pg_dump's CREATE TABLE command will create. Before 7.3 it
> was in fact not possible to dump and reload the regression-test database
> using COPY, because of this problem --- perhaps that causes me to
> overstate its importance, but there is a definite risk.
>
You do overstate it's importance IMHO, because I'm not lobbying for
permanent removal of the column names, nor making the default to not
have column names, I'm simply stating that the option to not have them
provides benefits and I'm willing to take the "risks" associated with
them.
At least the work arounds are simple... :-\
Robert Treat
--
Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL