> > with the types of cached plan. We need to be able to differentiate > > when cached plans are not used, so a simple boolean is not > > sufficient. > Sure. But I modestly hope you would add a CachedPlanSource pointer > solely to the PlannedStmt and restructure it a little as we discussed > above. And no new structures are needed. Am I wrong?
That was my initial intention somehow to get CachedPlan available to Executor hooks. But, as you pointed out there is more value in CachedPlanSource. I know Michael opposed the idea of carrying these structures, at least CachedPlan, to Executor hooks ( or maybe just not QueryDesc?? ). It will be good to see what he think, or if others an opinion about this, about adding a pointer to CachedPlanSource in PlannedStmt vs setting a flag in PlannedStmt to track plan cache type for the current execution? The former does provide more capability for extensions, as Andrei has pointed out earlier. -- Sami
