Re: COPY TO looses on view

> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > OK, the following patch throws an error if you try to COPY TO/FROM
> > views.
>
> This is probably the wrong direction to approach it from: rather than
> disallowing wrong things one at a time, we should only allow COPY if the
> relkind is plain relation. Think about toast rels, indexes, ...

Yes, I thought about that. I tried indexes and it failed because of the
heap_open call. I wonder if heap_open() is the place for more checks,
not even COPY: Probably not.

In fact, there is no reason you shouldn't be able to COPY out a view,
except that it doesn't work. :-)

Here are the reltypes. I didn't want to start disabling them from COPY
until I was sure. SEQUENCE is invalidated in COPY FROM, but not COPY
TO. No idea why. Anyway, I guess I was trying to avoid the research to
do it right. I will add it to my list.

#define RELKIND_INDEX 'i' /* secondary index */
#define RELKIND_RELATION 'r' /* ordinary cataloged heap */
#define RELKIND_SPECIAL 's' /* special (non-heap) */
#define RELKIND_SEQUENCE 'S' /* SEQUENCE relation */
#define RELKIND_UNCATALOGED 'u' /* temporary heap */
#define RELKIND_TOASTVALUE 't' /* moved off huge values */
#define RELKIND_VIEW 'v' /* view */

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026