* Ryan Kelly (rpkelly22@gmail.com) wrote:
> COPY ... RETURNING would certainly be useful to apply additional
> transformations to the data before finally sending it to its ultimate
> destination.
If we really think that COPY ... RETURNING is only going to be used in a
CTE or similar, then we could always only support that and forgo any
changes to the FE/BE protocol to support it. Or, at least, take the
simplest approach to supporting it which would involve cacheing the data
entirely before sending it back to the client (isn't that what we do on
a big INSERT ... VALUES ... RETURNING anyway? people can transfer in
blocks if they want to with INSERT .. VALUES or COPY .. RETURNING).
Thanks,
Stephen