Thread

Re: [PATCH] Rebuild CHECK constraints after generated column SET EXPRESSION

jian he <jian.universality@gmail.com> — 2026-05-15T02:52:07Z

On Fri, May 15, 2026 at 1:06 AM Ayush Tiwari
<ayushtiwari.slg01@gmail.com> wrote:
>
> I've two minor follow-up comments/questions:
>
> 1. The "-- indedx with whole-row reference need rebuild" comment is
>    still in the new SQL/expected blocks for both generated_stored and
>    generated_virtual:
>        "-- index with whole-row reference needs rebuild"
>
Thanks for finding the typo!

I did another round of code cleanup and cosmetic refactoring. Main
idea still the same: loop through
pg_constraint, pg_index, and pg_policy to locate all objects
containing whole-row
references, then error reporting or remember them for recreation.

The commit message was also updated.
Later, I will add this thread to
https://wiki.postgresql.org/wiki/PostgreSQL_19_Open_Items

> 2. The new policy error message:
>
>        errmsg("ALTER TABLE / SET EXPRESSION is not supported for
>                generated columns in tables that are part of a policy
>                definition"),
>        errdetail("%s contains whole row references.", ...)
>
>    I still find this wording a bit awkward.  Quoting "ALTER TABLE /
>    SET EXPRESSION" as a syntactic form in the message is unusual for
>    tablecmds.c, and the sentence is long.  Could we keep it closer to
>    the surrounding style.
>

+                    if (subtype == AT_SetExpression)
+                        ereport(ERROR,
+                                errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+                                errmsg("cannot alter generation
expression of table %s because %s uses its row type",
+                                       RelationGetRelationName(rel),
+                                       getObjectDescription(&pol_obj, false)),
+                                errdetail("You might need to drop %s
first.", getObjectDescription(&pol_obj, false)));
What do you think?



--
jian
https://www.enterprisedb.com/