Re: SQL:2011 application time

+ <para>
+ If the last column is marked with <literal>PERIOD</literal>,
+ it is treated in a special way.
+ While the non-<literal>PERIOD</literal> columns are treated normally
+ (and there must be at least one of them),
+ the <literal>PERIOD</literal> column is not compared for equality.
+ Instead the constraint is considered satisfied
+ if the referenced table has matching records
+ (based on the non-<literal>PERIOD</literal> parts of the key)
+ whose combined <literal>PERIOD</literal> values completely cover
+ the referencing record's.
+ In other words, the reference must have a referent for its
entire duration.
+ Normally this column would be a range or multirange type,
+ although any type whose GiST opclass has a "contained by" operator
+ and a <literal>referenced_agg</literal> support function is allowed.
+ (See <xref linkend="gist-extensibility"/>.)
+ In addition the referenced table must have a primary key
+ or unique constraint declared with <literal>WITHOUT PORTION</literal>.
+ </para>

typo "referenced_agg", in the gist-extensibility.html page is "referencedagg"
<literal>WITHOUT PORTION</literal> should be <literal>WITHOUT OVERLAPS</literal>

+ While the non-<literal>PERIOD</literal> columns are treated normally
+ (and there must be at least one of them),
+ the <literal>PERIOD</literal> column is not compared for equality.
the above sentence didn't say what is "normally"?
maybe we can do the following:
+ While the non-<literal>PERIOD</literal> columns are treated
+ normally for equality
+ (and there must be at least one of them),
+ the <literal>PERIOD</literal> column is not compared for equality.

+<programlisting>
+Datum
+my_range_agg_transfn(PG_FUNCTION_ARGS)
+{
+ MemoryContext aggContext;
+ Oid rngtypoid;
+ ArrayBuildState *state;
+
+ if (!AggCheckCallContext(fcinfo, &amp;aggContext))
+ elog(ERROR, "range_agg_transfn called in non-aggregate context");
+
+ rngtypoid = get_fn_expr_argtype(fcinfo-&gt;flinfo, 1);
+ if (!type_is_range(rngtypoid))
+ elog(ERROR, "range_agg must be called with a range");
+
+ if (PG_ARGISNULL(0))
+ state = initArrayResult(rngtypoid, aggContext, false);
+ else
+ state = (ArrayBuildState *) PG_GETARG_POINTER(0);
+
+ /* skip NULLs */
+ if (!PG_ARGISNULL(1))
+ accumArrayResult(state, PG_GETARG_DATUM(1), false, rngtypoid,
aggContext);
+
+ PG_RETURN_POINTER(state);
+}
+
+Datum
+my_range_agg_finalfn(PG_FUNCTION_ARGS)
+{
+ MemoryContext aggContext;
+ Oid mltrngtypoid;
+ TypeCacheEntry *typcache;
+ ArrayBuildState *state;
+ int32 range_count;
+ RangeType **ranges;
+ int i;
+
+ if (!AggCheckCallContext(fcinfo, &amp;aggContext))
+ elog(ERROR, "range_agg_finalfn called in non-aggregate context");
+
+ state = PG_ARGISNULL(0) ? NULL : (ArrayBuildState *) PG_GETARG_POINTER(0);
+ if (state == NULL)
+ /* This shouldn't be possible, but just in case.... */
+ PG_RETURN_NULL();
+
+ /* Also return NULL if we had zero inputs, like other aggregates */
+ range_count = state-&gt;nelems;
+ if (range_count == 0)
+ PG_RETURN_NULL();
+
+ mltrngtypoid = get_fn_expr_rettype(fcinfo-&gt;flinfo);
+ typcache = multirange_get_typcache(fcinfo, mltrngtypoid);
+
+ ranges = palloc0(range_count * sizeof(RangeType *));
+ for (i = 0; i &lt; range_count; i++)
+ ranges[i] = DatumGetRangeTypeP(state-&gt;dvalues[i]);
+
+ PG_RETURN_MULTIRANGE_P(make_multirange(mltrngtypoid,
typcache-&gt;rngtype, range_count, ranges));
+}

my_range_agg_transfn error message is inconsistent?
`elog(ERROR, "range_agg_transfn called in non-aggregate context");`
`elog(ERROR, "range_agg must be called with a range");`
maybe just `my_range_agg_transfn`, instead of mention
{range_agg_transfn|range_agg}
similarly my_range_agg_finalfn error is also inconsistent.

my_range_agg_finalfn need `type_is_multirange(mltrngtypoid)`?