Re: Proposal: Conflict log history table for Logical Replication

[Amit Kapila]

On Sun, Oct 4, 2026 at 2:01 AM Dilip Kumar <dilipbal...@gmail.com> wrote:
>
> Here is the failed test case
>
>  -- On Publisher:
>     CREATE TABLE tab_temporal (
>         id int4range,
>         valid_at datemultirange,
>         PRIMARY KEY (id, valid_at WITHOUT OVERLAPS)
>     );
>     INSERT INTO tab_temporal VALUES ('[1,2)', '{[2020-01-01,2021-01-01)}');
>     CREATE PUBLICATION pub_temporal FOR TABLE tab_temporal;
>
>     -- On Subscriber:
>     CREATE TABLE tab_temporal (
>         id int4range,
>         valid_at datemultirange,
>         PRIMARY KEY (id, valid_at WITHOUT OVERLAPS)
>     );
>     CREATE SUBSCRIPTION sub_temporal
>         CONNECTION '...'
>         PUBLICATION pub_temporal
>         WITH (conflict_log_destination = 'table');
>
>     -- After initial sync, delete the row locally on the subscriber:
>     DELETE FROM tab_temporal;
>
>     -- Then delete the row on the publisher to trigger a
> delete_missing conflict:
>     DELETE FROM tab_temporal;
>
> And without fix shared as a v77-0003 [1], this would error out with
> below error[2], I have explained the reason [1].
>

Okay, I see the problem but I wonder why you didn't consider using
rd_opcintype stored in index relation to solve this problem. I have
tried the attached atop v79-0001 and v79-0002, manually executed the
test you shared in this email and it fixed the issue for me.

-- 
With Regards,
Amit Kapila.

diff --git a/src/backend/replication/logical/conflict.c 
b/src/backend/replication/logical/conflict.c
index f0b71a74d89..a286f616f9c 100644
--- a/src/backend/replication/logical/conflict.c
+++ b/src/backend/replication/logical/conflict.c
@@ -1138,7 +1138,11 @@ build_index_key_json(EState *estate, Relation localrel,
                                continue;
                        }
 
-                       getTypeOutputInfo(att->atttypid, &outfuncoid, 
&typisvarlena);
+                       /*
+                        * See BuildIndexValueDescription() for why we used
+                        * rd_opcintype instead index's tupleDesc.
+                        */
+                       getTypeOutputInfo(indexDesc->rd_opcintype[i], 
&outfuncoid, &typisvarlena);
                        outputstr = OidOutputFunctionCall(outfuncoid, 
values[i]);
                        escape_json(&result, outputstr);
                        pfree(outputstr);