Yugo NAGATA <nagata@sraoss.co.jp> writes:
>> You could even argue that case 2 isn't good enough either,
>> and we should be delivering a specific error message saying
>> that an ENR can't be used in a view/matview. To do that,
>> we'd likely need to pass down the QueryEnvironment in more
>> places not fewer.
> We can raise a similar error for (not materialized) views by passing
> QueryEnv to DefineView() (or in ealier stage) , but there are other
> objects that can contain ENR in their definition, for examle, functions,
> cursor, or RLS policies. Is it worth introducing this version of error
> message for all these objects?
If it's worth checking for here, why not in other cases?
I'm not sure I like using isQueryUsingTempRelation as a model,
because its existing use in transformCreateTableAsStmt seems
like mostly a hack. (And I definitely don't love introducing
yet another scan of the query.) It seems to me that we should
think about this, for MVs as well as those other object types,
as fundamentally a dependency problem. That is, the reason
we can't allow a reference to an ENR in a long-lived object
is that we have no catalog representation for the reference.
So that leads to thinking that the issue ought to be handled
in recordDependencyOnExpr and friends. If we see an ENR while
scanning a rangetable to extract dependencies, then complain.
This might be a bit messy to produce good error messages for,
though.
Speaking of error messages, I'm not sure that it's okay to
use the phrase "ephemeral named relation" in a user-facing
error message. We don't use that term in our documentation
AFAICS, except in some SPI documentation that most users
will never have read. In the context of triggers, "transition
relation" seems to be what the docs use.
regards, tom lane
