Re: Parsing speed (was Re: pgstats_initstats() cost)

Joe Conway <mail(at)joeconway(dot)com> writes:
> Tom Lane wrote:
>> I'm beginning to think that was a serious omission. I'm tempted to fix
>> it, even though we're past feature freeze for 7.4. Comments?

> Seems pretty well isolated. If you're tallying votes, count me as a "yay".

Well, the early voting is running heavily to the 'yea' side. Let's talk
about what to call the thing.

Pre-7.4 we had "PQexec" (synchronous) and "PQsendQuery" (asynchronous)
to send a plain text query string (Simple Query message in new-FE-protocol-
speak). I added "PQexecParams" and "PQsendQueryParams" to do
Parse/Bind/Execute, choosing the names on the basis that being able to
pass out-of-line parameters was the primary new thing they were giving
the user.

A Bind/Execute facility would need a pair of routines with signatures
very similar to PQexecParams/PQsendQueryParams --- they'd take a
prepared statement name instead of a query string, and they'd not need
an array of parameter type OIDs, but otherwise the same. I couldn't
come up with very good names for them though. Best idea so far is
PQexecPrepared/PQsendQueryPrepared, but maybe someone has a better one?

We could also think about providing an interface to do just Parse,
although this is inessential since you can set up a prepared statement
by PQexec'ing a PREPARE command.

regards, tom lane