Re: [postgis-users] DE-9IM question
Martin Davis mtnclimb at telus.net writes: Have a look at: http://www.vividsolutions.com/jts/tests/index.html This is a visual interface to the JTS Test Suite, which has a very complete collection of DE-9IM cases. Interesting! Seems exhaustive to me. Two comments: Did you prove exhaustiveness - seems like there should be a way to make a combinatorial argument about how many possible Clementini matrices are possible (9^4 == 6561, less impossible combinations)? Also - it might be easier to do confirm exhaustiveness with simpler shapes, at least for part of the tests. Still, looks great, and comforts me that you are testing that hard. As was mentioned, you can use the JTS TestBuilder to build and inspect test cases. You can drag-and-drop Test Suite XML files onto the TestBuilder, so it's easy to inspect the cases in the XML tests. That is cool! I'm not sure if Wikipedia would appreciate being flooded with DE-9IM test cases - it doesn't seem like quite the right place for it. The Vivid site is a good reference as long as it stays up. I will continue to think about this. I probably don't have time to do fiddle with this in any systematic way, but a table with every possible matrix, a picture, and simple code to generate it might be pretty useful. ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
No, we didn't prove exhaustiveness. But as you can see from the documentation in the test cases we did try and formalize a system for describing geometric configurations, with the idea that it might lead to a way of enumerating all possible distinct situations. It would be interesting to see this pushed through to a proof. On 5/9/2012 9:32 AM, fork wrote: Martin Davismtnclimbat telus.net writes: Have a look at: http://www.vividsolutions.com/jts/tests/index.html This is a visual interface to the JTS Test Suite, which has a very complete collection of DE-9IM cases. Interesting! Seems exhaustive to me. Two comments: Did you prove exhaustiveness - seems like there should be a way to make a combinatorial argument about how many possible Clementini matrices are possible (9^4 == 6561, less impossible combinations)? Also - it might be easier to do confirm exhaustiveness with simpler shapes, at least for part of the tests. Still, looks great, and comforts me that you are testing that hard. As was mentioned, you can use the JTS TestBuilder to build and inspect test cases. You can drag-and-drop Test Suite XML files onto the TestBuilder, so it's easy to inspect the cases in the XML tests. That is cool! I'm not sure if Wikipedia would appreciate being flooded with DE-9IM test cases - it doesn't seem like quite the right place for it. The Vivid site is a good reference as long as it stays up. I will continue to think about this. I probably don't have time to do fiddle with this in any systematic way, but a table with every possible matrix, a picture, and simple code to generate it might be pretty useful. ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users - No virus found in this message. Checked by AVG - www.avg.com Version: 2012.0.1913 / Virus Database: 2425/4987 - Release Date: 05/09/12 ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
[postgis-users] DE-9IM question
Does anyone know of an exhaustive reference of DE-9IM matrices, with pictures? Context: I just finished classifying a bunch of polygons based on their relationship to enveloping poly's in another layer (census blocks and census city boundaries, respectively). The regular old st_within() and friends weren't enough, so I calculated all the DE-9IM's, counted them (yay for GROUP BY and real databases!), and made sure I knew what was going on for each of the matrices returned (7 total). Then I classified my block polys based on the strings. I found it CRAZY how many different types of DE-9IM's could be returned, even with just polygons and very well aligned boundaries (credit to the US Census Bureau's geography division). Only a few of the matrices in my data were actually described in the few docs I could find. Besides the seven in my data, there are at least four more that can be generated by various combinations of overlapping polygons. It would have been much easier to do this project if I had been able to reference a list of possibilities along with pictures. It would be even better if I could be confident that such a list was exhaustive. Does anyone have a reference? If not, I could continue to work on an atlas of DE-9IM matrices, at least for polygons, and submit it somewhere if it were helpful. Any ideas for where that might go? (I don't have a blog and don't want to start one). Maybe just the wikipedia entry? For those who don't know, a DE-9IM matrix is the output of st_relate(geo1, geo2), and there is an excellent intro in the docs. Fascinating stuff really. Tx ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
On Tue, May 08, 2012 at 03:46:36PM +, fork wrote: Does anyone know of an exhaustive reference of DE-9IM matrices, with pictures? http://en.wikipedia.org/wiki/DE-9IM http://postgis.org/documentation/manual-svn/using_postgis_dbmanagement.html#DE-9IM --strk; ,--o-. | __/ |Delivering high quality PostGIS 2.0 ! | / 2.0 |http://strk.keybit.net - http://vizzuality.com `-o--' ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
Sandro Santilli strk at keybit.net writes: http://en.wikipedia.org/wiki/DE-9IM http://postgis.org/documentation/manual-svn/using_postgis_dbmanagement.html#DE-9IM Those either aren't exhaustive (postgis) or don't have pictures (wiki). I would like to be able to look up FF2F1F212 and see a picture of it. I think a difference is that I am not interested in using a matrix to implement a function like st_within(), but rather I am querying a database looking for relationships and how to deal with them. In my case, also, the query will never return a string with wildcards in it, because it will give me the exact relationship. (Maybe the above links are exhaustive if you count wildcards, though...) Sorry -- I tried to make clear that I had already read the docs that are easily available. Thanks for the explicit links though. ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
fork forkandwait at gmail.com writes: I found it CRAZY how many different types of DE-9IM's could be returned, even with just polygons and very well aligned boundaries (credit to the US Census Bureau's geography division). Only a few of the matrices in my data were actually described in the few docs I could find. For a flavor, look at Table 3 here: http://www.cise.ufl.edu/~mschneid/Research/papers/BS01ER.pdf Note that these matrices are boolean for the intersection, and there are 32 of them. One can tweak them so that the max dimension on the boundaries is either 0 (a point) or 1 (a line) and create more matrices using the full definition of DE-91M; see #33. I am mostly just entertaining myself with mathematical patterns, but I think the topological matrices might be useful in trying to figure out how to simplify slivers and weird things. I also think one's intuition about within might not stand up to closer scrutiny via mathematical formalisms like this. The only way I could figure out my blocks-in-cities problem was to abandon st_within() etc and go for a list of specific DE-9IM's based on the data. If I feel bored/ ambitious I will try to create a useful atlas. I don't think one exists... ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
On 9 May 2012 03:46, fork forkandw...@gmail.com wrote: Does anyone know of an exhaustive reference of DE-9IM matrices, with pictures? A really handy visual tool to work with DE-9IM is JTS TestBuilder. You can interact with the geometries and the results, which makes it a pretty good educational tool. You can copy/paste WKB directly into the geometry inputs, which makes it work nice with PostGIS. Take a look at http://gis.stackexchange.com/a/12119/1872 -Mike ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users
Re: [postgis-users] DE-9IM question
Hi fork On 2012/5/8 fork forkandw...@gmail.com wrote: Does anyone know of an exhaustive reference of DE-9IM matrices, with pictures? I don't think I would call it a reference but take a look http://labs.geometa.info/postgisterminal/?xapi=node[tourism=zoo] : There is a menu item Relate exemplified (bool) which displays parameters as as list of pictures (disregarding of the current query). Choose the two geometry types and the DE-9IM pattern, then click on [+] and look at the prerendered pictures. Pls. give feedback if you have any suggestions. -S. Context: I just finished classifying a bunch of polygons based on their relationship to enveloping poly's in another layer (census blocks and census city boundaries, respectively). The regular old st_within() and friends weren't enough, so I calculated all the DE-9IM's, counted them (yay for GROUP BY and real databases!), and made sure I knew what was going on for each of the matrices returned (7 total). Then I classified my block polys based on the strings. I found it CRAZY how many different types of DE-9IM's could be returned, even with just polygons and very well aligned boundaries (credit to the US Census Bureau's geography division). Only a few of the matrices in my data were actually described in the few docs I could find. Besides the seven in my data, there are at least four more that can be generated by various combinations of overlapping polygons. It would have been much easier to do this project if I had been able to reference a list of possibilities along with pictures. It would be even better if I could be confident that such a list was exhaustive. Does anyone have a reference? If not, I could continue to work on an atlas of DE-9IM matrices, at least for polygons, and submit it somewhere if it were helpful. Any ideas for where that might go? (I don't have a blog and don't want to start one). Maybe just the wikipedia entry? For those who don't know, a DE-9IM matrix is the output of st_relate(geo1, geo2), and there is an excellent intro in the docs. Fascinating stuff really. Tx ___ 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
Re: [postgis-users] DE-9IM question
Have a look at: http://www.vividsolutions.com/jts/tests/index.html This is a visual interface to the JTS Test Suite, which has a very complete collection of DE-9IM cases. It is meant to be exhaustive - if you find something missing, send it in! It would be interesting to see the 7 situations that you found. The test suite is here: http://jts-topo-suite.svn.sourceforge.net/viewvc/jts-topo-suite/trunk/jts/testxml/ The validate directory contains the DE-9IM tests. As was mentioned, you can use the JTS TestBuilder to build and inspect test cases. You can drag-and-drop Test Suite XML files onto the TestBuilder, so it's easy to inspect the cases in the XML tests. I'm not sure if Wikipedia would appreciate being flooded with DE-9IM test cases - it doesn't seem like quite the right place for it. The Vivid site is a good reference as long as it stays up. On 5/8/2012 8:46 AM, fork wrote: Does anyone know of an exhaustive reference of DE-9IM matrices, with pictures? Context: I just finished classifying a bunch of polygons based on their relationship to enveloping poly's in another layer (census blocks and census city boundaries, respectively). The regular old st_within() and friends weren't enough, so I calculated all the DE-9IM's, counted them (yay for GROUP BY and real databases!), and made sure I knew what was going on for each of the matrices returned (7 total). Then I classified my block polys based on the strings. I found it CRAZY how many different types of DE-9IM's could be returned, even with just polygons and very well aligned boundaries (credit to the US Census Bureau's geography division). Only a few of the matrices in my data were actually described in the few docs I could find. Besides the seven in my data, there are at least four more that can be generated by various combinations of overlapping polygons. It would have been much easier to do this project if I had been able to reference a list of possibilities along with pictures. It would be even better if I could be confident that such a list was exhaustive. Does anyone have a reference? If not, I could continue to work on an atlas of DE-9IM matrices, at least for polygons, and submit it somewhere if it were helpful. Any ideas for where that might go? (I don't have a blog and don't want to start one). Maybe just the wikipedia entry? For those who don't know, a DE-9IM matrix is the output of st_relate(geo1, geo2), and there is an excellent intro in the docs. Fascinating stuff really. Tx ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users - No virus found in this message. Checked by AVG - www.avg.com Version: 2012.0.1913 / Virus Database: 2425/4985 - Release Date: 05/08/12 ___ postgis-users mailing list postgis-users@postgis.refractions.net http://postgis.refractions.net/mailman/listinfo/postgis-users