Alvaro Herrera <alvherre@commandprompt.com> writes:
> Hmm, so what's happening here, I think, is that the value is getting
> assigned to the record variable without detoasting. I guess we should
> detoast the value prior to assigning it, but it seems to me that that
> would have a large performance penalty for other cases in which the
> toast table is not dropped; in fact, you can even imagine some cases in
> which the toasted value is not even accessed, so getting to the point of
> detoasting it would be a severe penalization.
Yeah, we have seen similar reports before, and concluded that forcibly
detoasting on the remote chance that the toast table wouldn't stick
around was just untenable from a performance standpoint. It's annoying
but I don't see a good compromise solution ATM.
Pavel was working recently on a hack to prevent repeated detoastings,
which is at least somewhat related to this issue. I don't recall that
anyone liked his hack though ...
regards, tom lane