Re: [HACKERS] Re: [SQL] prob with aggregate and group by - returns multiplesh

Don Baccus <dhogaza(at)pacifier(dot)com> writes:
> I vote for the SQL92 TEMPORARY. Let's not add a keyword that is non-standard
> just because one or another commercial database makes use of it,

You're missing the point: we are not talking about *adding* a keyword,
we're talking about *removing* one that we've already supported for
a year or so. That changes matters considerably, IMHO.

I have in fact been able to make a conflict-free grammar in which TEMP
is accepted but not reserved. It requires a certain amount of
redundancy in the productions (see below), but I think this is a
worthwhile tradeoff for not breaking existing user code.

Shall I commit this?

regards, tom lane

/*
* Redundancy here is needed to avoid shift/reduce conflicts,
* since TEMP is not a reserved word. See also OptTemp.
*
* The result is a cons cell (not a true list!) containing
* a boolean and a table name.
*/
OptTempTableName: TEMPORARY opt_table relation_name
{ $$ = lcons(makeInteger(TRUE), (List *) $3); }
| TEMP opt_table relation_name
{ $$ = lcons(makeInteger(TRUE), (List *) $3); }
| LOCAL TEMPORARY opt_table relation_name
{ $$ = lcons(makeInteger(TRUE), (List *) $4); }
| LOCAL TEMP opt_table relation_name
{ $$ = lcons(makeInteger(TRUE), (List *) $4); }
| GLOBAL TEMPORARY opt_table relation_name
{
elog(ERROR, "GLOBAL TEMPORARY TABLE is not currently supported");
$$ = lcons(makeInteger(TRUE), (List *) $4);
}
| GLOBAL TEMP opt_table relation_name
{
elog(ERROR, "GLOBAL TEMPORARY TABLE is not currently supported");
$$ = lcons(makeInteger(TRUE), (List *) $4);
}
| TABLE relation_name
{ $$ = lcons(makeInteger(FALSE), (List *) $2); }
| relation_name
{ $$ = lcons(makeInteger(FALSE), (List *) $1); }
;