Re: speedup COPY TO for partitioned table.

On Fri, Mar 28, 2025 at 9:03 PM vignesh C <vignesh21(at)gmail(dot)com> wrote:
>
> On Fri, 28 Mar 2025 at 08:39, jian he <jian(dot)universality(at)gmail(dot)com> wrote:
> >
> > hi.
> >
> > I made a mistake.
> > The regress test sql file should have a new line at the end of the file.
>
> Couple of suggestions:
> 1) Can you add some comments here, this is the only code that is
> different from the regular table handling code:
> + scan_tupdesc = RelationGetDescr(scan_rel);
> + map = build_attrmap_by_name_if_req(tupDesc,
> scan_tupdesc, false);
>

I have added the following comments around build_attrmap_by_name_if_req.

/*
* partition's rowtype might differ from the root table's. We must
* convert it back to the root table's rowtype as we are export
* partitioned table data here.
*/

> 2) You can see if you can try to make a function add call it from both
> the partitioned table and regular table case, that way you could
> reduce the duplicate code:
> + while (table_scan_getnextslot(scandesc,
> ForwardScanDirection, slot))
> + {
> + CHECK_FOR_INTERRUPTS();
> +
> + /* Deconstruct the tuple ... */
> + if (map != NULL)
> + {
> + original_slot = slot;
> + root_slot =
> MakeSingleTupleTableSlot(tupDesc, &TTSOpsBufferHeapTuple);
> + slot =
> execute_attr_map_slot(map, slot, root_slot);
> + }
> + else
> + slot_getallattrs(slot);
> +
> + /* Format and send the data */
> + CopyOneRowTo(cstate, slot);
> +
> +
> pgstat_progress_update_param(PROGRESS_COPY_TUPLES_PROCESSED,
> +
> ++processed);
> +
> + if (original_slot != NULL)
> +
> ExecDropSingleTupleTableSlot(original_slot);
> + };
>

I consolidated it into a new function: CopyThisRelTo.