Re: [PATCH] Equivalence Class Filters

On 12/7/15 9:54 AM, Tom Lane wrote:
> Jim Nasby<Jim(dot)Nasby(at)BlueTreble(dot)com> writes:
>> >On 12/6/15 10:38 AM, Tom Lane wrote:
>>> >>I said "in most cases". You can find example cases to support almost any
>>> >>weird planner optimization no matter how expensive and single-purpose;
>>> >>but that is the wrong way to think about it. What you have to think about
>>> >>is average cases, and in particular, not putting a drag on planning time
>>> >>in cases where no benefit ensues. We're not committing any patches that
>>> >>give one uncommon case an 1100X speedup by penalizing every other query 10%,
>>> >>or even 1%; especially not when there may be other ways to fix it.
>> >This is a problem that seriously hurts Postgres in data warehousing
>> >applications.
> Please provide some specific examples. I remain skeptical that this
> would make a useful difference all that often in the real world ...
> and handwaving like that does nothing to change my opinion. What do
> the queries look like, and why would deducing an extra inequality
> condition help them?

I was speaking more broadly than this particular case. There's a lot of
planner improvements that get shot down because of the planning overhead
they would add. That's great for cases when milliseconds count, but
spending an extra 60 seconds (a planning eternity) to shave an hour off
a warehouse/reporting query.

There needs to be some way to give the planner an idea of how much
effort it should expend. GEQO and *_collapse_limit addresses this in the
opposite direction (putting a cap on planner effort), but I think we
need something that does the opposite "I know this query will take a
long time, so expend extra effort on planning it."
--
Jim Nasby, Data Architect, Blue Treble Consulting, Austin TX
Experts in Analytics, Data Architecture and PostgreSQL
Data in Trouble? Get it in Treble! http://BlueTreble.com