On Sun, Feb 4, 2018 at 3:41 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
>> It is not clear to me what is exactly your concern if we try to follow
>> #2? To me, #2 seems like a natural choice.
>
> At first, but it gives an anomaly so is not a good choice. The patch
> does behavior #5, it rechecks the conditions with the latest row.
>
> Otherwise
> WHEN MATCHED AND a=0 THEN UPDATE SET b=0
> WHEN MATCHED AND a=1 THEN UPDATE SET b=1
> would result in (a=1, b=0) in case of concurrent updates, which the
> user clearly doesn't want.
I am unable to understand this. What are you presuming the tuple was
originally?
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company