Hannu Krosing <hannu@skype.net> writes:
> On R, 2005-09-02 at 10:31 -0400, Tom Lane wrote:
>> Yes. However, I don't understand how you got that result; AFAIK the
>> planner should have eliminated the duplicate subclauses.
> the query was similar to this:
> [snip]
Oh, the OR arms are actually *not* equivalent because of the log_tableid
subclauses. So that explains why canonicalize_qual() didn't eliminate
them.
[ experiments... ] It looks like 8.0 and up are smart enough to
consolidate the identical extracted indexquals, but not 7.4.
regards, tom lane
