Thanks for your help, I will try to take less of your time in the following days :/
About "envelope query" the real issue is that the variable intersects is not set to null at the beginning of find method: https://github.com/nicolas-f/h2database/blob/svnmaster/h2/src/main/org/h2/index/IndexCursor.java#L76 About "Corrupt page count", I will try to create a unit test to reproduce this, but it is not easy, it may require huge index to reproduce. Is it ok to store MVRTreeMap in the DataBase.mvStore or it should work better if it is placed on a separate file ? Regards, -Nicolas Fortin GIS Workshop IRSTV FR CNRS 2488 Le jeudi 11 juillet 2013 11:20:09 UTC+2, Thomas Mueller a écrit : > > Hi, > > About "envelope query", you wrote "With index and Envelope union it take > more than 100s." - do you mean the following line in > IndexCursor.getSpatialSearchRow > is called: "v = ((ValueGeometry) v).union(vg);"? If yes, how is it > possible, because the method is only ever called with the parameter > "isIntersects" set to true. The method is only called within > IndexCursor.find, where isIntersect is set to true. > > About "Corrupt page count": I don't think it's because you directly use a > map. More likely is a bug within the MVStore, for example related to > concurrency. I do know that currently, the MVTableEngine doesn't always > work correctly when using multiple threads, so maybe there is a concurrency > problem in the MVStore that needs to be fixed. If not, then more likely is > another problem in the MVStore. I could have a look if I have a test case. > > Regards, > Thomas > > > > > On Thu, Jul 11, 2013 at 9:10 AM, Nicolas Fortin (OrbisGIS) < > [email protected] <javascript:>> wrote: > >> Hi Thomas, >> >> By using the good envelope query, in my performance test your MVRTreeMap >> is not far from JTS STRTree. >> >> In my sql request (count roads inside several areas): >> >>> select AREA.NAME,AREA.ID, COUNT(ROAD.ID) roadcount from AREA, ROAD >>> WHERE AREA.the_geom && ROAD.the_geom GROUP BY AREA.NAME,AREA.ID ORDER >>> BY AREA.NAME,AREA.ID; >> >> >> Without index, it takes ~9 s . With index and Envelope union it take more >> than 100s. With index and without envelope union it takes ~1s. >> >> I guess that Noel have done the same thing that you have done for >> interval numeric search. I will wrote a test case on that part. >> >> Currently I have the following StackTrace: >> >> 07-11 09:06:00 database: flush >>> java.lang.IllegalStateException: Corrupt page count -1 [1.3.172] >>> at >>> org.h2.mvstore.DataUtils.newIllegalStateException(DataUtils.java:632) >>> at org.h2.mvstore.MVStore.applyFreedPages(MVStore.java:1004) >>> at org.h2.mvstore.MVStore.store(MVStore.java:871) >>> at org.h2.mvstore.MVStore.store(MVStore.java:779) >>> at org.h2.mvstore.MVStore.compact(MVStore.java:1197) >>> at >>> org.h2.mvstore.db.MVTableEngine$Store.store(MVTableEngine.java:132) >>> at org.h2.engine.Database.flush(Database.java:1818) >>> at org.h2.store.WriterThread.run(WriterThread.java:87) >>> at java.lang.Thread.run(Thread.java:662) >>> >> >> I guess it is because I use directly store.openMap with main MVStore here: >> >> https://github.com/nicolas-f/h2database/blob/patch-merge/h2/src/main/org/h2/index/SpatialTreeIndex.java#L95 >> I will do the same thing as you, using Transaction, hope that it will fix >> corrupt page issue: >> >> https://github.com/nicolas-f/h2database/blob/patch-merge/h2/src/main/org/h2/mvstore/db/MVSecondaryIndex.java#L69 >> >> >> Regards, >> >> -Nicolas Fortin >> GIS Workshop >> IRSTV FR CNRS 2488 >> >> Le mercredi 10 juillet 2013 21:45:47 UTC+2, Thomas Mueller a écrit : >>> >>> Hi, >>> >>> > I have performance issues with the spatial index >>> >>> I don't think it's because of your change. Please note some features >>> are still missing or are not optimised at all. If you could post a test >>> case that would be great, I will then try to find the problem. >>> >>> Regards, >>> Thomas >>> >>> >>> On Wed, Jul 10, 2013 at 8:54 AM, Nicolas Fortin (OrbisGIS) < >>> [email protected]> wrote: >>> >>>> Hi Thomas, >>>> >>>> Ok nice, way better than a random test case. I have performance issues >>>> with the spatial index, maybe my fix introduce processing overhead. I will >>>> try to use your profiler, or the IntellliJ one.. >>>> >>>> Regards, >>>> >>>> -Nicolas Fortin >>>> GIS Workshop >>>> IRSTV FR CNRS 2488 >>>> >>>> Le lundi 8 juillet 2013 20:35:58 UTC+2, Thomas Mueller a écrit : >>>> >>>>> Hi, >>>>> >>>>> By the way, I now have a reproducible test case for the bug with the >>>>> MVRTreeMap (the one you fixed in your version). >>>>> >>>>> Regards, >>>>> Thomas >>>>> >>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "H2 Database" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to h2-database...@**googlegroups.com. >>>> To post to this group, send email to [email protected]. >>>> >>>> Visit this group at >>>> http://groups.google.com/**group/h2-database<http://groups.google.com/group/h2-database> >>>> . >>>> For more options, visit >>>> https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/opt_out> >>>> . >>>> >>>> >>>> >>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "H2 Database" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> To post to this group, send email to [email protected]<javascript:> >> . >> Visit this group at http://groups.google.com/group/h2-database. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > > -- You received this message because you are subscribed to the Google Groups "H2 Database" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/groups/opt_out.
