Re: Conflict detection and logging in logical replication

On Tue, Aug 20, 2024 at 4:45 PM Zhijie Hou (Fujitsu)
<houzj.fnst@fujitsu.com> wrote:
>
> Here are the remaining patches.
>
> 0001 adds additional doc to explain the log format.

Thanks for the patch. Please find few comments on 001:

1)
+<literal>Key</literal> (column_name, ...)=(column_name, ...);
existing local tuple (column_name, ...)=(column_name, ...); remote
tuple (column_name, ...)=(column_name, ...); replica identity
(column_name, ...)=(column_name, ...).

-- column_name --> column_value everywhere in right to '='?

2)
+      Note that for an
+      update operation, the column value of the new tuple may be NULL if the
+      value is unchanged.

-- Shall we mention the toast value here? In all other cases, we get a
full new tuple.

3)
+      The key section in the second sentence of the DETAIL line
includes the key values of the tuple that already exists in the local
relation for insert_exists or update_exists conflicts.

-- Shall we mention the key is the column value violating a unique
constraint? Something like this:
The key section in the second sentence of the DETAIL line includes the
key values of the local tuple that violates unique constraint for
insert_exists or update_exists conflicts.

4)
Shall we give an example LOG message in the end?

thanks
Shveta