Feedback-ID: ie0a040ee:Fastmail
Message-ID: <6c1fbccc-85c8-40d3-b08b-4f47f2093711@eisentraut.org>
Date: Fri, 23 Feb 2024 15:34:56 +0100
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Content-Language: en-US
From: Peter Eisentraut <peter@eisentraut.org>
Subject: RangeTblEntry.inh vs. RTE_SUBQUERY
To: pgsql-hackers <pgsql-hackers@postgresql.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/6c1fbccc-85c8-40d3-b08b-4f47f2093711%40eisentraut.org>
Precedence: bulk

Various code comments say that the RangeTblEntry field inh may only be 
set for entries of kind RTE_RELATION.

For example

  *    inh is true for relation references that should be expanded to 
include
  *    inheritance children, if the rel has any.  This *must* be false for
  *    RTEs other than RTE_RELATION entries.

and various comments in other files.

(Confusingly, it is also listed under "Fields valid in all RTEs:", but 
that definitely seems wrong.)

I have been deploying some assertions to see if the claims in the 
RangeTblEntry comments are all correct, and I tripped over something.

The function pull_up_simple_union_all() in prepjointree.c sets ->inh to 
true for RTE_SUBQUERY entries:

     /*
      * Mark the parent as an append relation.
      */
     rte->inh = true;

Whatever this is doing appears to be some undocumented magic.  If I 
remove the line, then regression tests fail with plan differences, so it 
definitely seems to do something.

Is this something we should explain the RangeTblEntry comments?