Re: Restrict publishing of partitioned table with a foreign table as partition

public inbox for [email protected]  
help / color / mirror / Atom feed

From: Álvaro Herrera <[email protected]>
To: Shlok Kyal <[email protected]>
Cc: vignesh C <[email protected]>
Cc: Sergey Tatarintsev <[email protected]>
Cc: [email protected]
Subject: Re: Restrict publishing of partitioned table with a foreign table as partition
Date: Mon, 24 Mar 2025 16:47:31 +0100
Message-ID: <[email protected]> (raw)
In-Reply-To: <CANhcyEXbKwQ4ySy0fSzX46N7sauJsvteWMTpYMci_Q03R_ewQA@mail.gmail.com>

One thing that bothers me a bit about this is that there's no single
code comment where this restriction it documented in full; in fact it
doesn't seem documented anywhere, only in the commit message.

I think check_foreign_tables() is a good place to add an explanatory
comment; other places can reference that.  For instance, add something
like

/*
 * Protect against including foreign tables that are partitions of
 * partitioned tables published by the given publication.  This would
 * not work properly, because , so we disallow
 * the case here and in all DDL commands that would end up creating
 * such a case indirectly.
 */

Then for instance in check_publication_add_relation() and
ATExecAttachPartition() you comment would say /* if the would-be
partition is a foreign table, verify that the partitioned table is not
in a publication with publish_via_root=false.  See check_foreign_tables
for details */

Also, surely we should document this restriction in the SGML docs
somewhere.

Would it be better if check_partrel_has_foreign_table() used
RelationGetPartitionDesc(omit_detached=true) instead of
find_all_inheritors()?

I'm wary of all those accesses of subscription/publication catalogs in
DDL code.  Maybe I worry about nothing, but I cannot but feel that we're
missing one layer of abstraction there (including possibly some caching
on top of syscache).

I think this
castNode(RangeVar, lfirst(list_head(stmt->base.inhRelations)));
is better written
linitial_node(RangeVar, stmt->base.inhRelations);

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/
"Tiene valor aquel que admite que es un cobarde" (Fernandel)

view thread (22+ messages)  latest in thread

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: [email protected]
  Cc: [email protected], [email protected], [email protected], [email protected], [email protected]
  Subject: Re: Restrict publishing of partitioned table with a foreign table as partition
  In-Reply-To: <[email protected]>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox