RE: [PATCH] Use optimized single-datum tuplesort in ExecSort - Mailing list pgsql-hackers

From: David Rowley <dgrowleyml@gmail.com>
> On Wed, 21 Jul 2021 at 13:39, James Coleman <jtc331@gmail.com> wrote:
> > Thanks for doing the math measuring how much we could impact things.
> >
> > I'm +lots on getting this committed as is.
> 
> Ok good. I plan on taking a final look at the v10 patch tomorrow morning NZ
> time (about 12 hours from now) and if all is well, I'll push it.
> 
> If anyone feels differently, please let me know before then.
Hi,

I noticed a minor thing about the v10 patch.

-
-        for (;;)
+        if (node->datumSort)
         {
-            slot = ExecProcNode(outerNode);
-
-            if (TupIsNull(slot))
-                break;
-
-            tuplesort_puttupleslot(tuplesortstate, slot);
+            for (;;)
+            {
+                slot = ExecProcNode(outerNode);
+
+                if (TupIsNull(slot))
+                    break;
+                slot_getsomeattrs(slot, 1);
+                tuplesort_putdatum(tuplesortstate,
+                                   slot->tts_values[0],
+                                   slot->tts_isnull[0]);
+            }
+        }
+        else
+        {
+            for (;;)
+            {
+                slot = ExecProcNode(outerNode);
+
+                if (TupIsNull(slot))
+                    break;
+                tuplesort_puttupleslot(tuplesortstate, slot);
+            }

The above seems can be shorter like the following ?

for (;;)
{
    slot = ExecProcNode(outerNode);
    if (TupIsNull(slot))
        break;
    if (node->datumSort)
    {
        slot_getsomeattrs(slot, 1);
        tuplesort_putdatum(tuplesortstate,
                    slot->tts_values[0],
                    slot->tts_isnull[0]);
    }
    else
        tuplesort_puttupleslot(tuplesortstate, slot);
}

Best regards,
houzj