Re: WAL bypass for INSERT, UPDATE and DELETE?

On Thu, 2005-12-22 at 12:12 -0500, Tom Lane wrote:
> Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> > Currently, CTAS optimization requires a heap_sync during ExecEndPlan. It
> > would be easy enough to extend this so that it also works for INSERT,
> > UPDATE and DELETE.
>
> If you tried to do it that way you'd break the system completely. Not
> all updates go through the executor.
>
> I think it's a bad idea anyway; you'd be adding overhead to the lowest
> level routines in order to support a feature that would be very seldom
> used, at least in comparison to the number of times those routines are
> executed.

The current thinking seems to be that we should implement an ALTER TABLE
RELIABILITY statement that applies to COPY, INSERT, UPDATE and DELETE.

> If you tried to do it that way you'd break the system completely. Not
> all updates go through the executor.

Where would I put a heap_sync to catch all of the I, U, D cases?
(Possibly multiple places).

Or were you thinking of things like ALTER TABLE TYPE?
Or perhaps inheritance?

Best Regards, Simon Riggs