Re: named queries and the wire protocol

David Welton <davidw(at)dedasys(dot)com> writes:
>> send(State, ?BIND, ["", 0, StatementName, 0, Bin1, Bin2]),
>> send(State, ?EXECUTE, ["", 0, <<0:?int32>>]),
>> send(State, ?CLOSE, [?PREPARED_STATEMENT, StatementName, 0]),
>> send(State, ?SYNC, []),

> And then the code above. So it's generating a name itself and then
> destroying it once the query is done.
> Perhaps this behavior is not a good idea and using the unnamed portal
> would be a better idea?

My point is that it *is* using the unnamed portal, AFAICS --- the ""s
in the Bind and Execute commands appear to correspond to the empty
strings that would select that portal.

The Close on the other hand is specifying closing a prepared statement,
not a portal. If you're right about the control flow around this
function, then the code is generating a prepared statement, using it
once, and destroying it. Which is dumb; you should instead use the
unnamed-statement protocol flow, which is better optimized for that
usage pattern.

regards, tom lane