RE: Data is copied twice when specifying both child and parent table in publication

On Fri, May 13, 2022 1:59 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> On Fri, May 13, 2022 at 7:32 AM wangw.fnst@fujitsu.com
> <wangw.fnst@fujitsu.com> wrote:
> >
> > Attach the patches.(Only changed the patch for HEAD.).
> >
> 
> Few comments:
> =============
Thanks for your comments.

> 1.
> @@ -1135,6 +1172,15 @@ pg_get_publication_tables(PG_FUNCTION_ARGS)
>   if (publication->pubviaroot)
>   tables = filter_partitions(tables);
>   }
> + pfree(elems);
> +
> + /*
> + * We need an additional filter for this case : A partition table is
> + * published in a publication with viaroot, and its parent or child
> + * table is published in another publication without viaroot. In this
> + * case, we should publish only parent table.
> + */
> + tables = filter_partitions(tables);
> 
> Do we need to filter partitions twice? Can't we check if any of the publications
> has 'pubviaroot' option set, if so, call filter_partitions at the end?
Improve it according to your suggestion.

> 2.  " FROM pg_class c JOIN pg_namespace n"
> + "   ON n.oid = c.relnamespace,"
> + " LATERAL pg_get_publication_tables(array[ %s ]) gst"
> 
> Here, it is better to have an alias name as gpt.
Improve it according to your suggestion.

> 3.
>   }
> + pfree(elems);
> +
> 
> An extra line between these two lines makes it looks slightly better.
Improve it according to your suggestion.

> 4. Not able to apply patch cleanly.
> patching file src/test/subscription/t/013_partition.pl
> Hunk #1 FAILED at 477.
> Hunk #2 FAILED at 556.
> Hunk #3 FAILED at 584.
> 3 out of 3 hunks FAILED -- saving rejects to file
> src/test/subscription/t/013_partition.pl.rej
> patching file src/test/subscription/t/028_row_filter.pl
> Hunk #1 succeeded at 394 (offset 1 line).
> Hunk #2 FAILED at 722.
> 1 out of 2 hunks FAILED -- saving rejects to file
> src/test/subscription/t/028_row_filter.pl.rej
> patching file src/test/subscription/t/031_column_list.pl
> Hunk #1 succeeded at 948 (offset -92 lines).
> Hunk #2 FAILED at 1050.
> 1 out of 2 hunks FAILED -- saving rejects to file
> src/test/subscription/t/031_column_list.pl.rej
New patch could apply patch cleanly now.

Attach the patches.(Only changed the patch for HEAD.).
1. Optimize the code. Reduce calls to function filter_partitions. [suggestions by Amit-san]
2. Improve the alias name in SQL. [suggestions by Amit-san]
3. Improve coding alignments. [suggestions by Amit-san] 
4. Do some optimizations for list Concatenate.

Regards,
Wang wei