Re: Planner Row Estimate with Function - Mailing list pgsql-general

From Tom Lane
Subject Re: Planner Row Estimate with Function
Date
Msg-id 8614.1262146782@sss.pgh.pa.us
Whole thread Raw
In response to Re: Planner Row Estimate with Function  (Michael Fork <mfork00@yahoo.com>)
Responses Re: Planner Row Estimate with Function  (Michael Fork <mfork00@yahoo.com>)
List pgsql-general
Michael Fork <mfork00@yahoo.com> writes:
>> Also, what happened to the type='CREDIT' condition in your query?  Is
>> that a partial index?

> Yes, this is partial index.  I should have included the index definition earlier:

Ah.  The optimizer is ignoring the index stats on the grounds that
they are not representative of the whole table.  I'm not entirely sure
offhand whether it would be safe to use them anyway if the index
predicate is known to match the query --- it seems a bit risky but on
the other hand it'd probably be better than having no stats at all.

What you can do to get a better estimate is to create a non-partial
index on parsecardidfromreferencecode(reference_code).  It's fairly
likely that if you do that, maintaining the partial index as well
is not worth your time; but that's a decision you'd have to make.

            regards, tom lane

pgsql-general by date:

Previous
From: Michael Fork
Date:
Subject: Re: Planner Row Estimate with Function
Next
From: Michael Fork
Date:
Subject: Re: Planner Row Estimate with Function