Tom Lane wrote:
> Joe Conway <mail@joeconway.com> writes:
>>Any guidance on the preferred fix?
>
> We cannot fix this by changing ExecScanSubPlan as you suggest.
> That would amount to saying that all plans have to be run to completion,
> which destroys LIMIT to name just one unpleasant consequence.
I suspected as much.
> There is a mechanism available for functions to arrange to get cleanup
> callbacks when a containing plan is shut down early --- see
> RegisterExprContextCallback and ShutdownExprContext. It looks like this
> is not used by the existing SRF support, but I suspect it should be.
OK -- I'll take a look.
> [ scratches head ... ] Right at the moment I don't see where
> ShutdownExprContext gets called during a ReScan operation. I'm quite
> sure it once was ... there may be another bug here ...
>
Thanks, I'll keep that in mind too.
Joe