Re: [HACKERS] [PATCH] Incremental sort - Mailing list pgsql-hackers

On Wed, Nov 15, 2017 at 7:42 AM, Alexander Korotkov
<a.korotkov@postgrespro.ru> wrote:
> Sure, please find rebased patch attached.

+ /*
+  * Check if first "skipCols" sort values are equal.
+  */
+ static bool
+ cmpSortSkipCols(IncrementalSortState *node, TupleTableSlot *a,
+                                                             TupleTableSlot *b)
+ {
+     int n, i;
+
+     Assert(IsA(node->ss.ps.plan, IncrementalSort));
+
+     n = ((IncrementalSort *) node->ss.ps.plan)->skipCols;
+
+     for (i = 0; i < n; i++)
+     {
+         Datum datumA, datumB, result;
+         bool isnullA, isnullB;
+         AttrNumber attno = node->skipKeys[i].attno;
+         SkipKeyData *key;
+
+         datumA = slot_getattr(a, attno, &isnullA);
+         datumB = slot_getattr(b, attno, &isnullB);
+
+         /* Special case for NULL-vs-NULL, else use standard comparison */
+         if (isnullA || isnullB)
+         {
+             if (isnullA == isnullB)
+                 continue;
+             else
+                 return false;
+         }
+
+         key = &node->skipKeys[i];
+
+         key->fcinfo.arg[0] = datumA;
+         key->fcinfo.arg[1] = datumB;
+
+         /* just for paranoia's sake, we reset isnull each time */
+         key->fcinfo.isnull = false;
+
+         result = FunctionCallInvoke(&key->fcinfo);
+
+         /* Check for null result, since caller is clearly not expecting one */
+         if (key->fcinfo.isnull)
+             elog(ERROR, "function %u returned NULL", key->flinfo.fn_oid);
+
+         if (!DatumGetBool(result))
+             return false;
+     }
+     return true;
+ }

Is there some reason not to use ApplySortComparator for this?  I think
you're missing out on lower-overhead comparators, and in any case it's
probably good code reuse, no?

Embarrassingly, I was unaware of this patch and started prototyping
exactly the same thing independently[1].  I hadn't got very far and
will now abandon that, but that's one thing I did differently.  Two
other things that may be different: I had a special case for groups of
size 1 that skipped the sorting, and I only sorted on the suffix
because I didn't put tuples with different prefixes into the sorter (I
was assuming that tuplesort_reset was going to be super efficient,
though I hadn't got around to writing that).  I gather that you have
determined empirically that it's better to be able to sort groups of
at least MIN_GROUP_SIZE than to be able to skip the comparisons on the
leading attributes, but why is that the case?