Re: pgsql: Generalize hash and ordering support in amapi - Mailing list pgsql-committers

From Peter Eisentraut
Subject Re: pgsql: Generalize hash and ordering support in amapi
Date
Msg-id 5d98c343-4d8e-4c10-bbfe-7f122065fc18@eisentraut.org
Whole thread Raw
In response to Re: pgsql: Generalize hash and ordering support in amapi  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-committers
On 07.03.25 19:50, Tom Lane wrote:
> Peter Eisentraut <peter@eisentraut.org> writes:
>> I have committed fixes for these issues along the lines you suggested.
> 
> Thanks.  There is a typo in hashhandler:
> 
> -   amroutine->amcancrosscompare = true;
> +   amroutine->amconsistentequality = true;
> +   amroutine->amconsistentequality = false;
> 
> The second line should be setting amconsistentordering = false.
> 
> Also, may I suggest one more thing?  I think the test in
> comparison_ops_are_compatible should be just
> 
> -           if (amroutine->amcanorder && amroutine->amconsistentordering)
> +           if (amroutine->amconsistentordering)
> 
> (and the comment for it needs adjustment too).  To my mind,
> amconsistentordering is a static declaration that operators
> within one of the AM's opfamilies are expected to have this
> property.  That could be true whether or not the AM is capable
> of returning tuples in order.  So although these flags might
> commonly be set together, I think they are independent
> properties.

Agreed, done.




pgsql-committers by date:

Previous
From: Peter Eisentraut
Date:
Subject: pgsql: Make amcanorder independent of amconsistentordering
Next
From: Peter Geoghegan
Date:
Subject: pgsql: Make parallel nbtree index scans use an LWLock.