Greg Stark <gsstark@mit.edu> writes:
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>> Rather than hard-wiring a special case for any of these things, I'd much
>> rather see us implement INSERT...RETURNING and UPDATE...RETURNING as per
>> previous suggestions.
> I wonder whether the ui tools need anything more low level than that. In
> general sticking their grubby fingers in the query the user entered seems
> wrong and they would have to tack on a RETURNING clause.
That was mentioned before as a possible objection, but I'm not sure that
I buy it. The argument seems to be that a client-side driver would
understand the query and table structure well enough to know what to do
with a returned pkey value, but not well enough to understand how to
tack on a RETURNING clause to request that value. This seems a bit
bogus.
There may be some point in implementing a protocol-level equivalent of
RETURNING just to reduce the overhead on both sides, but I think we
ought to get the RETURNING functionality in place first and then worry
about that...
regards, tom lane
