[postgis-users] Measure distance to nearest border

Thu Oct 21 02:14:13 PDT 2010

Ok still don't get the second query. If I understand you right you could try this: CREATE TABLE borddisttest2 ASselect a.gid, a.cell, MIN(ST_Distance(a.centroid, b.the_geom)) AS distance FROM pgfinal2008 a, cshapes b, cshapes c WHERE a.gwcode != b.gwcode AND b.gwsyear <= 2008 AND b.gweyear >= 2008 and a.gwcode = c.gwcode and st_intersects(b.the_geom, c.the_geom)GROUP BY a.gid LIMIT 1;
Here I assume that every country only have one representation in cshape.What I do is that I try to only get the countries that intersects with the country that has the same gwcode as the cell you are calculating.  Here it is essential to have spatial index on the country-polygons and the centriod-pointYou should also have index on gwsyear, gweyear and gwcode in both pgfinal2008 and cshapes. I guess your limit 1 is just for testing purposes Probably you could make it quite a lot faster if you could trick it to inly do the intersection process once per country and not once per cell. Right now I don't see any good way to do that. Hopefully someone else sees a solution to that, or I will return if it comes to me. But maybe, if I recall right there were some discussions here that postgresql can cache results from one row to another. If soo it should work by itself or at least if you order your query by the gwcode so it takes all grid cells from one country first. /Nicklas 
2010-10-21 Andreas Forø Tollefsen wrote:

Ok. I will try to explain more in detail.>
I have a vector grid consisting of 64818 grid cells. Each of these cells are stored in the pgfinal2008 table, with two geometries variables: centroid (point) and cell (polygon).>
Each grid cell have a gwcode variable which is the country code. >

>>
In addition i have a table with all countries represented by polygons, and a table with all countries represented by their boundaries (line).>

>>
What i want to do is to measure the distance from each cell to the closest border..>

>>
While a simple measure of ST_Distance(pgfinal2008.centroid, cshapes.the_geom) WHERE pgfinal2008.gwcode != cshapes.gwcode will measure the distances, it will measure to all the polygons where the gwcode is different.>
However, I can apply a MIN() function to select the nearest of these, though this will take considerable time since every cell need to be measured to every country.>

>>
What i need to solve is how to limit the distance to measuring only to neighbouring countries of the country the cell is located.>

>>
Thanks for your help Nicklas.>

>>

>
>>
2010/10/21 Nicklas Avén <nicklas.aven at jordogskog.no>
> >
>hmm, I have to admit I don't really understand what you are trying to do, but there are some thingsI don't think you really mean.>
> >select a.gid, MIN(ST_Distance(a.centroid, b.the_geom)) AS distance FROM pgfinal2008 a, cshapes b, cshapes c WHERE a.gwcode != b.gwcode AND b.gwsyear <= 2008 AND b.gweyear >= 2008 GROUP BY a.gid LIMIT 1;>
> Why do you have, cshapes c, there. if you have 200 countries in it taking away that will reduce the querytime to 1/200>What you get here is theclosest country from your first grid cell.>
> >>
SELECT pgfinal2008.cell, borddisttest2.gid, MIN(borddisttest2.distance) FROM borddisttest2, pgfinal2008 WHERE borddisttest2.gid = pgfinal2008.gidGROUP BY pgfinal2008.cell, borddisttest2.gid;
>>This I don't get. Do you have many rows in pgfinal2008 with the same value in pgfinal2008.cell but different pgfinal2008.gid. What is "cell"? > >/Nicklas>
>
 >
> >
> 2010-10-21 Andreas Forø Tollefsen wrote:
>
> Thanks.>>
I have found a way to do this calculation.> >
However, since i gave 64818 cells and almost 200 countries, the distance from every cell to every country is calculated before the minimum distance is selected.> >
This takes awful lot of time.> >

> >> >
Any ideas on how to increase the performance?> >

> >> >
> >
DROP TABLE IF EXISTS borddisttest2;> >
CREATE TABLE borddisttest2 AS> >
select a.gid, MIN(ST_Distance(a.centroid, b.the_geom)) AS distance FROM pgfinal2008 a, cshapes b, cshapes c WHERE a.gwcode != b.gwcode AND b.gwsyear <= 2008 AND b.gweyear >= 2008 GROUP BY a.gid LIMIT 1;> >

> >> >
DROP TABLE IF EXISTS borddist2008;> >
CREATE TABLE borddist2008 AS> >
SELECT pgfinal2008.cell, borddisttest2.gid, MIN(borddisttest2.distance) FROM borddisttest2, pgfinal2008 WHERE borddisttest2.gid = pgfinal2008.gid> >
GROUP BY pgfinal2008.cell, borddisttest2.gid;
> >> >
2010/10/13 Nicklas Avén <nicklas.aven at jordogskog.no>
> > > >
> >Hi> > > >You can not cast to geography that way (if I haven't missed something essential)
> > And if you could you would mix geometry and geograph type in ST_Distance, which wouldn't work.> > > >St_Distance for geometry vs geography uses totally differnt algorithms. The geography calculations is far more advanced. > > > >ST_Shortestline only works for geometry type.> > > >/Nicklas> >
> >
 > >
> > > >
> > 2010-10-13 Andreas Forø Tollefsen wrote:
> >
> > Hi>> >

> > >> > >
Thanks,> > >
I will give it a try.> > >
The reason I used the geography types in the distance query, were to get the distance in meters instead of degrees.> > >

> > >> > >
Andreas
> > >
> > >> > >
> >
> >
 > >
2010/10/13 Nicklas Avén <nicklas.aven at jordogskog.no>
> > > > > >
> > >Hi> > > > > >As I understand you you want to get the distance from each cell to the closest neighbour country. Then, why don't you just query the distance from your cell to closest country with other gwcode?> > >Maybe something like:> > > > > >Create table borddisttest as> > >Select ST_Distance(st_collect(b.the_geom), c.centroid)> > >from cshapes a, cshapes b, pgfinal2008 c> > >where a.gwcode=c.gwcode and b.gwcode != c.gwcode and st_intersects(a.the_geom, b.the_geom)> > >group by c.gwcode and c.the_geom;
> > >> > >With some tweaking like ordering the cells by gwcode I don't think the intersection calculation have to be done for each cell. Otherwise this is a very bad approach if there are many grid-cells.> > > > > >indexing on gwcode on both tables and spatial index on the country geoms for the intersection will also be of importance.> > > > > >I don't understand:> > >geography(pgfinal2008.centroid)> > > > > >what does geography means here?> > > > > >I would also go for calculating the centroids on the fly. It should be fast and you will not get the trouble of updating the centroid column if adjustuing the grid.> > > > > >About doing calculations comparing geometries inside a dataset you use self join. > > > > > >You can join a table with itself as long as you put an alias so you can identify them like I did above with a, b and c> > > > > >Well, I am not sure I answered the right question but anyway...> > >HTH> > >Nicklas> > >
> > >
 > > >
> > >
> > > 2010-10-13 Andreas Forø Tollefsen wrote:
> > >
> > > >> > >
Hi all,> > > >

> > > >> > > >
I have two datasets. On is a quadrate grid, and the other is the boundaries of all countries.> > > >
What i want to do is to measure the distance from each centroid of the grid cells to the nearest border, but not all borders. Only the international.> > > >

> > > >> > > >
First i convert my country polygon dataset into line features:> > > >
> > > >
select gwcode, gwsyear, gweyear, ST_Boundary(the_geom) AS line Into cshapes_line from cshapes;> > > >

> > > >> > > >
Next i calculate the distance from the centroid of each cell to the nearest border where the gwcode (country code) of the cell is the same as the gwcode in the line feature.> > > >
drop table if exists borddisttest;> > > >
select pgfinal2008.cell, pgfinal2008.gwcode, ST_ShortestLine(pgfinal2008.centroid, st_boundary(cshapes.the_geom)) > > > >
AS shortestline, ST_Distance(geography(pgfinal2008.centroid), st_boundary(cshapes.the_geom))/1000 AS borddist > > > >
into borddisttest > > > >
from cshapes, pgfinal2008 > > > >
where cshapes.gwcode = pgfinal2008.gwcode;> > > >

> > > >> > > >
The problem here is that when using ST_Boundary, it converts all boundaries into lines, while i only want to get the boundaries where two polygons with different gwcodes meet.> > > >
The lines where two countries meet are overlapping, meaning these borders have two gwcodes but in different line entries.> > > >

> > > >> > > >
Is there a way to calculate the overlaps within one dataset? If i could do this, then i could put a clause so the measurement only measures to lines with 2 or more gwcodes.> > > >

> > > >> > > >
Thanks.
> > >_______________________________________________
> > > postgis-users mailing list
> > >postgis-users at postgis.refractions.net
> >> >
>postgis.refractions.net/mailman/listinfo/postgis-users
> > >
> > >
> > >
> >_______________________________________________
> > postgis-users mailing list
> >postgis-users at postgis.refractions.net
> >postgis.refractions.net/mailman/listinfo/postgis-users
> >
> >
> >
>_______________________________________________
> postgis-users mailing list
>postgis-users at postgis.refractions.net
>postgis.refractions.net/mailman/listinfo/postgis-users
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20101021/e0e8509e/attachment.html>