Re: Need Help With a A Simple Query That's Not So Simple

From: "David Johnston" <polobo(at)yahoo(dot)com>
To: "'Bill Thoen'" <bthoen(at)gisnet(dot)com>, "'Postgrresql'" <pgsql-general(at)postgresql(dot)org>
Subject: Re: Need Help With a A Simple Query That's Not So Simple
Date: 2011-10-31 23:05:43
Message-ID: 027a01cc9821$9e887ef0$db997cd0$@yahoo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

From: pgsql-general-owner(at)postgresql(dot)org
[mailto:pgsql-general-owner(at)postgresql(dot)org] On Behalf Of Bill Thoen
Sent: Monday, October 31, 2011 6:51 PM
To: Postgrresql
Subject: [GENERAL] Need Help With a A Simple Query That's Not So Simple

I think this should be easy, but I can't seem to put the SQL together
correctly and would appreciate any help. (I'm using Pg 8.4 in CentOS 5.5, if
that matters.)

I have a table of Farms and a table of crops in a 1:M relationship of Farms
: Crops. There are lots of different crops to choose form but for now I'm
only interested in two crops; corn and soybeans. Some farms grow only corn
and some grow only soybeans, and some grow both. What I'd like to know is,
which Farms and how many are growing only corn, which and how many are
growing soybeans and which and how many are growing both? I can easily get
all the corn growers with:

SELECT a.*
  FROM farms a
  JOIN crops b
    ON a.farm_id=b.farm_id
 WHERE crop_cd='0041'

I can do the same with soybeans (crop_cd= '0081') and then I could subtract
the sum of these from the total of all farms that grow either corn or
soybeans to get the number of farms growing both, but having to do all those
queries sounds very time consuming and inefficient. Is there a better way to
get the farm counts or data by categories like farms growing only corn,
farms growing only soybeans, farms growing both? I'm also interested in
possibly expanding to a general case where I could select more than two
crops. and get counts of the permutations.

Here's a sketch of the relevant pieces of the data base.

Tables: 
farms                      crops
=======                    =======
farm_id  bigint (pkey)     crop_id   (pkey)
type                       farm_id    foreign key to farms
size                       crop_cd    0041 = corn 0081=soybeans
...                        year
                           ...

Any help would be much appreciated.

TIA,

- Bill Thoen
---------------------------------------------------------------

General Idea:

WITH crop_one AS (
SELECT farm_id, crop_cd AS crop_one_cd ...
), crop_two AS (
SELECT farm_id, crop_cd AS crop_two_cd
)
SELECT *
FROM crop_one
FULL OUTER JOIN crop_two USING (farm_id)
;

Records with NULL for "crop_one_cd" only grow crop 2, records with NULL for
"crop_two_cd" only grow crop 1, records where neither field is NULL grow
both.

Not sure regarding the general case. You likely want to use ARRAY_AGG to
get a result like:

Farm_id_100, { 'CROP_CD_1', 'CROP_CD_2' }

You could then probably get a query to output something like:
(crop_id, farms_exclusive, farms_shared, farms_without)
Where each of the "farms_" columns is an array of farm_ids that match the
particular conditional

= ALL (exclusive); != ALL && = ANY (shared); != ANY (without)

David J.

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message David Kerr 2011-10-31 23:54:49 Re: Server hitting 100% CPU usage, system comes to a crawl.
Previous Message Bill Thoen 2011-10-31 22:51:02 Need Help With a A Simple Query That's Not So Simple