Tom Lane writes:
> > Allowing column aliases without AS would break nothing.
>
> Now I'll admit that the postfix-operator issue is probably a red
> herring, since "SELECT a + b, ..." should get interpreted as infix +
> in any case. But I think there are other side-effects we have not
> fully analyzed yet. The one-line grammar hack was not intended to
> be actually applied, right? Let's see a complete proposal.
I have analyzed the all the grammar conflicts that the optional AS would
cause. The real killer were these two clauses:
expr ISNULL
expr NOTNULL
I don't know how to resolve that without making ISNULL and NOTNULL more
reserved than any other word, which seems unattractive.
(The only two other areas of conflict where postfix operators and trailing
multiword type names in ::-style type casts, which we already talked
about.)
--
Peter Eisentraut peter_e@gmx.net