Thanks for the advice. I did retile my rasters down to 100x100 simply be dropping and reloading (easy enough through a simple script.) I didn't notice an appreciable difference although I didn't check the logs for exact times.
Interesting thought on putting all tiles for all layers in the same table. I had to read that a few times to understand it. I will try that and compare to what I have now, which is a php script that queries each of my 17 raster layers and returns those results in a couple of seconds. Good enough for now. - John *********************************************** John Callahan, Research Scientist Delaware Geological Survey, University of Delaware URL: http://www.dgs.udel.edu ************************************************* On Tue, Feb 21, 2012 at 10:32 AM, Pierre Racine <pierre.rac...@sbf.ulaval.ca > wrote: > > Thank you. It works great! The following works for a single point > against a > > single raster: > > > > SELECT > > ST_Value(rast, ST_SetSRID(ST_Point(187251.0, 119422.0), 26957)) FROM > geol1 > > WHERE ST_Intersects(rast, ST_SetSRID(ST_Point(187251.0, 119422.0), > > 26957)::geometry, 1); > > It will work even faster if you retile everything to smaller tiles and > create an index on your raster column (which I have no doubt you did). You > can use the new ST_Tile() plpgsql prototype to retile everything without > reloading: > > > http://trac.osgeo.org/postgis/browser/trunk/raster/scripts/plpgsql/st_tile.sql > > It's a bit slow but can work overnight... > > > Could anyone recommend the best way to query multiple rasters using the > same > > point for each? I've seen examples pulling multiple points from the > same rasters > > but not the same point from multiple rasters. I have about 10-20 > rasters (geol2, > > geol3..) I need to find the value of at a single point. > > If you have an application that want to do that often I would recommend to > store all the tiles for all the layers in the same table with a column > identifying the semantic of each set of tile. You then just do a query like > you did, adding the point id and the semantic column. For each point you > will get as many rows as you have layers. Transforming everything in a one > multiple column row is another story: You have to JOIN the table with > itself as many time as you have layers. > > Pierre > _______________________________________________ > postgis-users mailing list > postgis-users@postgis.refractions.net > http://postgis.refractions.net/mailman/listinfo/postgis-users >
_______________________________________________ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
