> In general I don't think it even makes sense to think of making
> executor rollback non-transactional.

Agreed.

I would not want to rollback some statements and not others within a
transaction. I would like a complete rollback to happen, but only when
the exception propogates out of the exception block unhandled, not when
the exception is first thrown. Maybe if I could get into the TRY
section of the PG_CATCH()/PG_TRY() construct without an intervening
elog(ERROR) then I'd have a chance ...

> Seems like your choices are
> ...
> implementing a separate expression evaluator
> ...
> make the restriction "read-only database access"
> ...

Thanks for all these ideas. I'm just getting to know the PG code tree,
and I appreciate the guidance.