Re: Support NOT VALID / VALIDATE constraint options for named NOT NULL constraints

On Fri, Apr 4, 2025 at 7:54 PM Rushabh Lathia <rushabh.lathia@gmail.com> wrote:
>
> Hi Alvaro,
>
> I’ve consolidated all the changes and attached the latest version of the patch, which
> includes the updates submitted by Jian for pg_dump as well.
>
> Patch 0001 contains changes to MergeWithExistingConstraint to fix the marking on local constraints.
> Patch 0002 includes support for NOT NULL NOT VALID, corresponding pg_dump changes, test cases,
>                    and documentation updates.
>

hi.
+               /* FIXME use CompactAttribute */
                Form_pg_attribute att = TupleDescAttr(relation->rd_att, i - 1);
                if (att->attnotnull && att->attnotnullvalid &&
!att->attisdropped)
                {
                    NullTest   *ntest = makeNode(NullTest);
                    ntest->arg = (Expr *) makeVar(varno,
                                                  i,
                                                  att->atttypid,
                                                  att->atttypmod,
                                                  att->attcollation,
                                                  0);
                    ntest->nulltesttype = IS_NOT_NULL;

CompactAttribute doesn't have {atttypmod, attcollation} information,
now it is impossible to use CompactAttribute here,
so I removed this FIXME in get_relation_constraints.


i noticed that we have
"mode change 100644 => 100755 src/bin/pg_dump/t/002_pg_dump.pl"
in 0002-Support-NOT-VALID-and-VALIDATE-CONSTRAINT-for-named-.patch.

i am uncomfortable with the change in
'CREATE TABLE dump_test.test_table_generated'
so I only added 'CONSTRAINT NOT NULL / INVALID' tests in
002_pg_dump.pl.
so I only added a test case 'CONSTRAINT NOT NULL / INVALID'
to 002_pg_dump.pl.


v7-0001 commit message explains what kind of problem
MergeWithExistingConstraint is trying to fix.
v7-0002 bullet points summary about NOT NULL NOT VALID added to the
commit message.


add a test for CREATE TABLE LIKE.
CREATE TABLE LIKE will copy the invalid not-null constraint and will become
valid, i think this is what we want.


The added regress test is a little bit verbose, trying to make it less verbose.
polish comments here and there.