On Fri, Aug 15, 2014 at 11:52:30PM -0400, Tom Lane wrote:
> David G Johnston <david.g.johnston@gmail.com> writes:
> > cpronovost wrote
> >> When using the <@ operator in conjunction with the NOT operator on a jsonb
> >> column, the NOT reverse the <@ operator(becomes a @>).
>
> > The presence of "NOT" does not (aside from a possible bug) change the "<@"
> > operator into the "@>" operator. "NOT" simply inverts the supplied boolean
> > value so that "not(true) := false" and vice-versa.
>
> That's what it's *supposed* to do, but somebody got the pg_operator entry
> wrong. It looks to me like pg_operator OIDs 3246 and 3250 should be
> listed as each others' commutators, not each others' negators.
>
> Sigh. So much for no initdb for 9.4beta3. I wonder whether there are
> other thinkos in the jsonb operator entries...
Just to be clear, pg_upgrade can handle system catalog changes, while it
can't handle data format changes. However, I support changing the JSONB
data format to be optimal. FYI, I have always been concerned that
pg_upgrade would inhibit data storage format improvements.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ Everyone has their own god. +
