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

On Tues, May 17, 2022 9:03 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> On Fri, May 13, 2022 at 3:11 PM wangw.fnst@fujitsu.com
> <wangw.fnst@fujitsu.com> wrote:
> >
> > Attach the patches.(Only changed the patch for HEAD.).
> >
Thanks for your comments.

>  # publications
> -{ oid => '6119', descr => 'get OIDs of tables in a publication',
> +{ oid => '6119', descr => 'get OIDs of tables in one or more publications',
>    proname => 'pg_get_publication_tables', prorows => '1000', proretset => 't',
> -  provolatile => 's', prorettype => 'oid', proargtypes => 'text',
> -  proallargtypes => '{text,oid}', proargmodes => '{i,o}',
> +  provolatile => 's', prorettype => 'oid', proargtypes => 'any',
> +  proallargtypes => '{any,oid}', proargmodes => '{i,o}',
> 
> Won't our use case (input one or more publication names) requires the
> parameter type to be 'VARIADIC text[]' instead of 'any'? I might be
> missing something here so please let me know your reason to change the
> type to 'any' from 'text'?
Yes, you are right. I improve the approach according to your suggestion.
I didn't notice the field "provariadic" in pg_proc before. And now I found we
could change the type of input from text to variadic text by specifying fields
"provariadic" and specifying 'v' in "proargmodes".
I also make corresponding changes to the processing of the input in function
pg_get_publication_tables.

BTW, in previous patch HEAD_v4, when invoke function GetPublicationByName in
function pg_get_publication_tables, I changed the second input from "false" to
"true". I changed this because when we invoke function
pg_get_publication_tables in the query in function fetch_table_list, if the
publication does not exist, it will error.
But this change will affect the compatibility of function
pg_get_publication_tables. So I revert this change in HEAD_v5, and filter the
publications that do not exist by the query in function fetch_table_list.

Attach the patches.(Only changed the patch for HEAD.)
1. Improve the approach to modify the input type of the function
   pg_get_publication_tables. [suggestions by Amit-san]
2. Free allocated memory in function fetch_table_list. [suggestions by Osumi-san]
3. Improve the approach of the handling of non-existing publications.

BTW, I rename the patch for REL14
from
"REL14_v4-0001-Fix-data-replicated-twice-when-specifying-PUBLISH.patch"
to
"REL14_v5-0001-Fix-data-replicated-twice-when-specifying-PUBLISH_patch".
Just for the version doesn't mess up between two branches and for cfbot.

Regards,
Wang wei