Dear Wang,
Thank you for updating patch! Following are comments form v19-0001.
01. logical-replication.sgml
I found a following statement in logical-replication.sgml. I think this may cause
mis-reading because it's OK when publishers list partitions and publish_via_root is true.
```
<para>
A subscriber node may have multiple subscriptions if desired. It is
possible to define multiple subscriptions between a single
publisher-subscriber pair, in which case care must be taken to ensure
that the subscribed publication objects don't overlap.
</para>
```
How about adding "If publications are set publish_via_partition_root as true and
they publish partitions that have same partitioned table, only a change to partitioned
table is published from the publisher."or something like that?
02. filter_partitions
IIUC this function can refactor like following to avoid "skip" flag.
How do you think?
```
@@ -209,7 +209,6 @@ filter_partitions(List *table_infos)
foreach(lc, table_infos)
{
- bool skip = false;
List *ancestors = NIL;
ListCell *lc2;
published_rel *table_info = (published_rel *) lfirst(lc);
@@ -224,13 +223,10 @@ filter_partitions(List *table_infos)
/* Is ancestor exists in the published table list? */
if (is_ancestor_member_tableinfos(ancestor, table_infos))
{
- skip = true;
+ table_infos = foreach_delete_current(table_infos, lc);
break;
}
}
-
- if (skip)
- table_infos = foreach_delete_current(table_infos, lc);
}
}
```
03. fetch_table_list
```
+ /* Get the list of tables from the publisher. */
+ if (server_version >= 160000)
```
I think boolean variable can be used to check it like check_columnlist.
How about "use_extended_function" or something?
Best Regards,
Hayato Kuroda
FUJITSU LIMITED