Re: Index on immutable function call - Mailing list pgsql-general

From Tom Lane
Subject Re: Index on immutable function call
Date
Msg-id 7983.1263912455@sss.pgh.pa.us
Whole thread Raw
In response to Re: Index on immutable function call  (Dean Rasheed <dean.a.rasheed@googlemail.com>)
List pgsql-general
Dean Rasheed <dean.a.rasheed@googlemail.com> writes:
> 2010/1/19 Philippe Lang <philippe.lang@attiksystem.ch>:
>> That works just fine, but is there maybe a way of creating a slighly
>> more "generic" index? If I change the ">" with a "<" in the query, index
>> cannot of course be used. According to documentation, answer seems to be
>> "no"...

> You could create an index on the difference:
> create index long_transformation_index on indexed_table
> ((data1-this_is_a_long_transformation(data2)));

> then rewrite your queries accordingly:
> ... WHERE data1-this_is_a_long_transformation(data2) > some const (or
> < some const)

Yeah.  There's been a lot of noise in this thread about statistics,
but the real problem is this: an indexable condition compares the
indexed value to a constant (or at least something that doesn't
change across rows).  "col1 > foo(col2)" is not able to use an index
on col1, nor an index on foo(col2).  But if you indexed the whole
expression col1 - foo(col2), you could get an index search on
comparisons of that to a constant.

            regards, tom lane

pgsql-general by date:

Previous
From: Dean Rasheed
Date:
Subject: Re: Index on immutable function call
Next
From: Terry
Date:
Subject: Re: data dump help