Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Clearly, we should shoot for something that leverages the existing
> SELECT code rather than hang more clauses off of COPY.
Yeah, that's a good point. COPY IN is still a special case, I think,
but seems like COPY OUT could be reimplemented as a special tuple
destination for the regular executor machinery.
Q: how much performance hit would we be taking? If this slows down
pg_dump a lot, the extra code is worth keeping.
regards, tom lane