[postgis-users] RE: Question: How can I improve the performance of the function DISTANCE_SPHERE?

[Reichle, Florian]

Hi,

i have tried this at my DB, but the query is even slow:

SELECT distinct point.field1, point.field2
FROM tab_point point LEFT JOIN
            (SELECT point.field1 as pfield1, point.field2 as pfield2  FROM tab_point point, tab_line line WHERE
 Expand(line.the_geom, 0.01)  && point.the_geom  AND
(DISTANCE_SPHEROID(point.the_geom, StartPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00 OR DISTANCE_SPHEROID(point.the_geom, Centroid(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00 OR DISTANCE_SPHEROID(point.the_geom, EndPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00) ) As ce 
           ON (ce.pfield1 = point.field1 and ce.pfield2 = point.field2)
WHERE ce.pfield1 IS NULL;

I'm thinking you are looking for
"All points whose closest line is > 50 meters away" which doesn't require a slow cartesian product. --> Yes, that's right. I' am searching for all points whose closest line is  > 50 meters. But how is the trick do not use the
cartesian product?

I have many datas in the DB and it takes moren than 10 hours. 


Date: Mon, 17 Sep 2007 15:38:08 -0400
From: "Obe, Regina" <robe.dnd at cityofboston.gov>
Subject: RE: [postgis-users] Question: How can I improve the
        performance of  thefunction DISTANCE_SPHERE?
To: "PostGIS Users Discussion" <postgis-users at postgis.refractions.net>
Message-ID:
        <53F9CF533E1AA14EA1F8C5C08ABC08D201979F8A at ZDND.DND.boston.cob>
Content-Type: text/plain; charset="iso-8859-1"

Florian,

Two points
1) You have a cartesian product here which is very slow with the tab_point point, tab_line line but it sounds like from your description of your desired result that may be intentional but its hard to tell.   Just thought I would point it out.

I'm thinking you are looking for
"All points whose closest line is > 50 meters away" which doesn't require a slow cartesian product.

But I could read your question a couple of ways.

2)
In general NOT IN is intuitive but much slower than doing a left join.  Also I think you want to  do a compound check.  Try doing the following instead


The below query should give you "all points who are > 50 meters away from the closest starting point the psuedo closest line".  Note I added an Expand of a percent of a degree because it is not guaranteed your point will be in the bounding box of the line and can still be within 50 meters but take that out if you want.  Also just going by the startpoint is faulty - it might be better to use Centroid() instead of startpoint or you could do (centroid or startpoint or endpoint) combination check - see the second example below. 

If you could transform to a non-degree projection, that would be a lot more efficient too and then you can use distance which would give you distance from closest point on the line. 

SELECT distinct point.field1, point.field2
FROM tab_point point LEFT JOIN
            (SELECT point.field1 as pfield1, point.field2 as pfield2  FROM tab_point point, tab_line line WHERE
 Expand(line.the_geom, 0.01)  && point.the_geom  AND
DISTANCE_SPHEROID(point.the_geom, StartPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00)  As ce 
           ON (ce.pfield1 = point.field1 and ce.pfield2 = point.field2)
WHERE ce.pfield1 IS NULL;

---What I mean by combination check

SELECT distinct point.field1, point.field2
FROM tab_point point LEFT JOIN
            (SELECT point.field1 as pfield1, point.field2 as pfield2  FROM tab_point point, tab_line line WHERE
 Expand(line.the_geom, 0.01)  && point.the_geom  AND
(DISTANCE_SPHEROID(point.the_geom, StartPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00 OR DISTANCE_SPHEROID(point.the_geom, Centroid(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00 OR DISTANCE_SPHEROID(point.the_geom, EndPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00) ) As ce 
           ON (ce.pfield1 = point.field1 and ce.pfield2 = point.field2)
WHERE ce.pfield1 IS NULL;



Caveat - the above examples wors reliably only if field1 and field2 are never null of your point table are never null. It would be better to use the primary key of the table.




________________________________

From: postgis-users-bounces at postgis.refractions.net on behalf of Reichle, Florian
Sent: Mon 9/17/2007 11:04 AM
To: postgis-users at postgis.refractions.net
Subject: [postgis-users] Question: How can I improve the performance of thefunction DISTANCE_SPHERE?



Hi,

I use the function DISTANCE_SPHERE to get a metric unit how far is a point away from a polygon. I have two tables, one for the points and one for the polygons. The polygons are MULTILINES.
The result of the query must be a list of all points, which are to far a away from any polygon. So I reduce in the first step the list with && which drop all points which are intersects with the geom of any polygon. After that i must proof if the distance of the rest to far away. Therefor I use the DISTANCE_SPHERE function. But at here I have the problem with the speed of the query, because the function take to much time to compare alle elements in the tables. Have someone any better idea?

Here my example:

SELECT distinct point.field1, point.field2, line.field1, line.field2 FROM tab_point point, tab_line line WHERE
point.field1 NOT IN (SELECT distinct point.field1, point.field2, line.field1, line.field2 FROM tab_point point, tab_line line WHERE
point.the_geom && line.the_geom AND
DISTANCE_SPHEROID(point.the_geom, StartPoint(line.the_geom), 'SPHEROID["WGS 84",6378137,298.257223563]') < 50.00 GROUP BY point.field1, point.field2, line.field1, line.field2 from );

Greetz

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 7401 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20070924/79778674/attachment.bin>