PG18: syntax & semantics behavior in Citus, part 1.

[colm-mchugh]

PG18: syntax & semantics behavior in Citus, part 1.

Includes PG18 tests for:

• OLD/NEW support in RETURNING clause of DML queries (PG commit 80feb727c)
• WITHOUT OVERLAPS in PRIMARY KEY and UNIQUE constraints (PG commit fc0438b4e)
• COLUMNS clause in JSON_TABLE (PG commit bb766cd)
• Foreign tables created with LIKE clause (PG commit 302cf1575)
• Foreign Key constraint with PERIOD clause (PG commit 89f908a6d)
• COPY command REJECT_LIMIT option (PG commit 4ac2a9bec)
• COPY TABLE TO on a materialized view (PG commit 534874fac)

Partially addresses issue #8250

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.95%. Comparing base (accd01f) to head (eba6153).
⚠️ Report is 11 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8335      +/-   ##
==========================================
+ Coverage   88.92%   88.95%   +0.02%     
==========================================
  Files         287      287              
  Lines       63151    63152       +1     
  Branches     7941     7941              
==========================================
+ Hits        56156    56175      +19     
+ Misses       4681     4668      -13     
+ Partials     2314     2309       -5     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[naisila]

I think this would only work in create time. i.e. not able to add such a primary key after creating the table, so this should probably be handled separately

[colm-mchugh]

Actually add such a primary key after creating the table does work after the table is distributed; I just added tests for the same. So adding a primary key:

ALTER TABLE temporal_rng
  ADD CONSTRAINT temporal_rng_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS);

is propagated to the table's shads:

DETAIL:  on server citus@localhost:9702 connectionId: 13
NOTICE:  issuing SELECT worker_apply_shard_ddl_command (104206, 'citus', 'ALTER TABLE temporal_rng ADD
  CONSTRAINT temporal_rng_pk PRIMARY KEY (id, valid_at WITHOUT OVERLAPS)
;')

[naisila]

Great then. It seems its part of the column name.

/*
 * AppendColumnNameList converts a list of columns into comma separated string format
 * (colname_1, colname_2, .., colname_n).
 */
void
AppendColumnNameList(StringInfo buf, List *columns)
{
	appendStringInfoString(buf, " (");

	ListCell *lc;
	bool firstkey = true;

	foreach(lc, columns)
	{
		if (firstkey == false)
		{
			appendStringInfoString(buf, ", ");
		}

		appendStringInfo(buf, "%s", quote_identifier(strVal(lfirst(lc))));
		firstkey = false;
	}

	appendStringInfoString(buf, " )");
}

I also tested distributing the table after adding a constraint, and also adding a new node, and it seems to propagate fine. Could you maybe include one of those tests here, just to be sure. Maybe include the one you commented here.

[naisila]

Sorry, just saw that you already added the tests!!!

[m3hm3t]

maybe add order by

[colm-mchugh]

There is an ORDER BY on both columns of the table - is that what you mean?

[m3hm3t]

sorry I added the comment at wrong line. There are some select queries with no order by.

[colm-mchugh]

Got it, added.

[naisila]

LGTM, thanks a lot for the extensive testing. Great to see Citus's current infrastructure seamlessly propagating all these new cool PG18 features.

[naisila]

Tested citus_add_node, propagates fine

[m3hm3t]

maybe add order by